{
  "article_key": "llm-vram-96gb",
  "cases": [
    {
      "case_wall_seconds": 29.91674370097462,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 5704.656739645703,
          "reason": null,
          "response_union_seconds": 19.113167581963353,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 298.4673636738817,
          "trapezoid_pieces": 30,
          "watt_hours": 1.5846268721238064
        },
        "gpu_sample_count": 20,
        "host_sample_count": 19,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 6720995328,
            "mean": 6712309221.052631,
            "median": 6713368576,
            "minimum": 6703079424,
            "sample_count": 19
          },
          "gpu_board_watts": {
            "maximum": 301.0,
            "mean": 299.713,
            "median": 299.95,
            "minimum": 297.7,
            "sample_count": 20
          },
          "gpu_memory_used_bytes": {
            "maximum": 6752829440.0,
            "mean": 6752829440.0,
            "median": 6752829440.0,
            "minimum": 6752829440.0,
            "sample_count": 20
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1845.0,
            "mean": 1793.7,
            "median": 1800.0,
            "minimum": 1710.0,
            "sample_count": 20
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 95.9,
            "median": 97.0,
            "minimum": 85.0,
            "sample_count": 20
          },
          "host_mem_available_bytes": {
            "maximum": 96199720960,
            "mean": 96188027095.57895,
            "median": 96187265024,
            "minimum": 96176087040,
            "sample_count": 19
          },
          "process_pss_bytes": {
            "maximum": 819189760,
            "mean": 815475550.3157895,
            "median": 816355328,
            "minimum": 811747328,
            "sample_count": 19
          },
          "process_rss_bytes": {
            "maximum": 822177792,
            "mean": 818463582.3157895,
            "median": 819343360,
            "minimum": 814735360,
            "sample_count": 19
          }
        },
        "process_sample_count": 19,
        "response_union_seconds": 19.113167581963353,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 0
          },
          "observed_endpoint_seconds": 9.175436379155144,
          "observed_intervals": 8,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 172.7690055501098,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.119277758589535,
            "interval_count": 5,
            "interval_seconds": 7.423593470070045,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.81688827486512,
            "process_cpu_seconds": 7.41
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 2207.7645157894385,
              "reason": null,
              "response_union_seconds": 7.408736283017788,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 297.99474990762576,
              "trapezoid_pieces": 13,
              "watt_hours": 0.6132679210526218
            },
            "gpu_sample_count": 8,
            "host_sample_count": 8,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 6708359168,
                "mean": 6705936384,
                "median": 6706384896.0,
                "minimum": 6703079424,
                "sample_count": 8
              },
              "gpu_board_watts": {
                "maximum": 300.43,
                "mean": 299.49625,
                "median": 299.95,
                "minimum": 297.7,
                "sample_count": 8
              },
              "gpu_memory_used_bytes": {
                "maximum": 6752829440.0,
                "mean": 6752829440.0,
                "median": 6752829440.0,
                "minimum": 6752829440.0,
                "sample_count": 8
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1845.0,
                "mean": 1832.5,
                "median": 1833.5,
                "minimum": 1822.0,
                "sample_count": 8
              },
              "gpu_utilization_percent": {
                "maximum": 97.0,
                "mean": 94.25,
                "median": 97.0,
                "minimum": 86.0,
                "sample_count": 8
              },
              "host_mem_available_bytes": {
                "maximum": 96199720960,
                "mean": 96194545152,
                "median": 96195299328.0,
                "minimum": 96187265024,
                "sample_count": 8
              },
              "process_pss_bytes": {
                "maximum": 813074432,
                "mean": 812405760,
                "median": 812410880.0,
                "minimum": 811747328,
                "sample_count": 8
              },
              "process_rss_bytes": {
                "maximum": 816062464,
                "mean": 815393792,
                "median": 815398912.0,
                "minimum": 814735360,
                "sample_count": 8
              }
            },
            "process_sample_count": 8,
            "response_union_seconds": 7.408736283017788,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 0
              },
              "observed_endpoint_seconds": 3.05030764703406,
              "observed_intervals": 3,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 7.430938940960914,
          "response_seconds": {
            "maximum": 1.484641135961283,
            "mean": 1.4817472566035577,
            "median": 1.481124191021081,
            "minimum": 1.479376372997649,
            "sample_count": 5
          },
          "response_window_union_seconds": 7.408736283017788,
          "server_decode_tokens_per_second": {
            "maximum": 197.7636422013653,
            "mean": 197.5064926865939,
            "median": 197.53690841569204,
            "minimum": 197.21684494811652,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 0.19157884595915675,
            "mean": 0.1905686531914398,
            "median": 0.19060329400235787,
            "minimum": 0.1899017799878493,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 109.36028990279206,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.118724755561472,
            "interval_count": 5,
            "interval_seconds": 11.72354178893147,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.7991921779671,
            "process_cpu_seconds": 11.7
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 3496.8922238562645,
              "reason": null,
              "response_union_seconds": 11.704431298945565,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 298.76652137479715,
              "trapezoid_pieces": 17,
              "watt_hours": 0.9713589510711846
            },
            "gpu_sample_count": 12,
            "host_sample_count": 11,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 6720995328,
                "mean": 6716944011.636364,
                "median": 6717145088,
                "minimum": 6713245696,
                "sample_count": 11
              },
              "gpu_board_watts": {
                "maximum": 301.0,
                "mean": 299.8575,
                "median": 299.995,
                "minimum": 298.01,
                "sample_count": 12
              },
              "gpu_memory_used_bytes": {
                "maximum": 6752829440.0,
                "mean": 6752829440.0,
                "median": 6752829440.0,
                "minimum": 6752829440.0,
                "sample_count": 12
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1822.0,
                "mean": 1767.8333333333333,
                "median": 1773.5,
                "minimum": 1710.0,
                "sample_count": 12
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 97.0,
                "median": 97.5,
                "minimum": 85.0,
                "sample_count": 12
              },
              "host_mem_available_bytes": {
                "maximum": 96188547072,
                "mean": 96183286690.90909,
                "median": 96184393728,
                "minimum": 96176087040,
                "sample_count": 11
              },
              "process_pss_bytes": {
                "maximum": 819189760,
                "mean": 817708125.0909091,
                "median": 817772544,
                "minimum": 816355328,
                "sample_count": 11
              },
              "process_rss_bytes": {
                "maximum": 822177792,
                "mean": 820696157.0909091,
                "median": 820760576,
                "minimum": 819343360,
                "sample_count": 11
              }
            },
            "process_sample_count": 11,
            "response_union_seconds": 11.704431298945565,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 0
              },
              "observed_endpoint_seconds": 6.125128732121084,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 11.733725362981204,
          "response_seconds": {
            "maximum": 2.3454450629651546,
            "mean": 2.340886259789113,
            "median": 2.3407816879916936,
            "minimum": 2.3358659429941326,
            "sample_count": 5
          },
          "response_window_union_seconds": 11.704431298945565,
          "server_decode_tokens_per_second": {
            "maximum": 169.3727562261093,
            "mean": 169.06737906840763,
            "median": 169.01162004597117,
            "minimum": 168.80352064566353,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 0.834626693977043,
            "mean": 0.8324093010160141,
            "median": 0.8318316700169817,
            "minimum": 0.8301292260293849,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "baseline-8b",
      "model": "qwen3-8b-q4",
      "module": "baseline",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 333.84
          },
          {
            "device": "CUDA0",
            "reported_mib": 4455.34
          }
        ],
        "gpu_layers": 37,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 37
      },
      "record_sha256": "2db66c2c90e7ccf91ace8aaf7f6b8e1865a1de2453d6547e769e2dda7e176b21",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6373740544,
              "mean": 6365132800,
              "median": 6365132800.0,
              "minimum": 6356525056,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 15.03,
              "mean": 15.024999999999999,
              "median": 15.024999999999999,
              "minimum": 15.02,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96634912768,
              "mean": 96623800320,
              "median": 96623800320.0,
              "minimum": 96612687872,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1007707070093602,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6392254464,
              "mean": 6392254464,
              "median": 6392254464,
              "minimum": 6392254464,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 78.38,
              "mean": 78.38,
              "median": 78.38,
              "minimum": 78.38,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2325.0,
              "median": 2325.0,
              "minimum": 2325.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96599707648,
              "mean": 96599707648,
              "median": 96599707648,
              "minimum": 96599707648,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.155548368988093,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 51.318043537193375,
            "reason": null,
            "response_union_seconds": 1.2843944630003534,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 39.95504886973283,
            "trapezoid_pieces": 2,
            "watt_hours": 0.014255012093664827
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6579744768,
              "mean": 6579744768,
              "median": 6579744768,
              "minimum": 6579744768,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 50.74,
              "mean": 50.74,
              "median": 50.74,
              "minimum": 50.74,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 5259657216.0,
              "mean": 5259657216.0,
              "median": 5259657216.0,
              "minimum": 5259657216.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2280.0,
              "mean": 2280.0,
              "median": 2280.0,
              "minimum": 2280.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 76.0,
              "mean": 76.0,
              "median": 76.0,
              "minimum": 76.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96307732480,
              "mean": 96307732480,
              "median": 96307732480,
              "minimum": 96307732480,
              "sample_count": 1
            },
            "process_pss_bytes": {
              "maximum": 5364942848,
              "mean": 5364942848,
              "median": 5364942848,
              "minimum": 5364942848,
              "sample_count": 1
            },
            "process_rss_bytes": {
              "maximum": 5367930880,
              "mean": 5367930880,
              "median": 5367930880,
              "minimum": 5367930880,
              "sample_count": 1
            }
          },
          "process_sample_count": 1,
          "response_union_seconds": 1.2843944630003534,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 521.8941446628455,
            "reason": null,
            "response_union_seconds": 3.3171223630197346,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 157.33340152931243,
            "trapezoid_pieces": 6,
            "watt_hours": 0.1449705957396793
          },
          "gpu_sample_count": 3,
          "host_sample_count": 4,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6721740800,
              "mean": 6704672768,
              "median": 6714998784.0,
              "minimum": 6666952704,
              "sample_count": 4
            },
            "gpu_board_watts": {
              "maximum": 226.38,
              "mean": 158.82666666666665,
              "median": 176.54,
              "minimum": 73.56,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 6752829440.0,
              "mean": 6746537984.0,
              "median": 6752829440.0,
              "minimum": 6733955072.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2310.0,
              "median": 2310.0,
              "minimum": 2295.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96215072768,
              "mean": 96190546944,
              "median": 96184281088.0,
              "minimum": 96178552832,
              "sample_count": 4
            },
            "process_pss_bytes": {
              "maximum": 819189760,
              "mean": 807820288,
              "median": 816168960.0,
              "minimum": 779753472,
              "sample_count": 4
            },
            "process_rss_bytes": {
              "maximum": 822177792,
              "mean": 810808320,
              "median": 819156992.0,
              "minimum": 782741504,
              "sample_count": 4
            }
          },
          "process_sample_count": 4,
          "response_union_seconds": 3.3171223630197346,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 1.0194021269562654,
            "observed_intervals": 1,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 5723.3732039281185,
            "reason": null,
            "response_union_seconds": 19.17764565697871,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 298.4398244862446,
            "trapezoid_pieces": 22,
            "watt_hours": 1.5898258899800328
          },
          "gpu_sample_count": 20,
          "host_sample_count": 19,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6720995328,
              "mean": 6712309221.052631,
              "median": 6713368576,
              "minimum": 6703079424,
              "sample_count": 19
            },
            "gpu_board_watts": {
              "maximum": 301.0,
              "mean": 299.713,
              "median": 299.95,
              "minimum": 297.7,
              "sample_count": 20
            },
            "gpu_memory_used_bytes": {
              "maximum": 6752829440.0,
              "mean": 6752829440.0,
              "median": 6752829440.0,
              "minimum": 6752829440.0,
              "sample_count": 20
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1845.0,
              "mean": 1793.7,
              "median": 1800.0,
              "minimum": 1710.0,
              "sample_count": 20
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 95.9,
              "median": 97.0,
              "minimum": 85.0,
              "sample_count": 20
            },
            "host_mem_available_bytes": {
              "maximum": 96199720960,
              "mean": 96188027095.57895,
              "median": 96187265024,
              "minimum": 96176087040,
              "sample_count": 19
            },
            "process_pss_bytes": {
              "maximum": 819189760,
              "mean": 815475550.3157895,
              "median": 816355328,
              "minimum": 811747328,
              "sample_count": 19
            },
            "process_rss_bytes": {
              "maximum": 822177792,
              "mean": 818463582.3157895,
              "median": 819343360,
              "minimum": 814735360,
              "sample_count": 19
            }
          },
          "process_sample_count": 19,
          "response_union_seconds": 19.17764565697871,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 16384
            },
            "observed_endpoint_seconds": 17.353597005014308,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 888.6925389093352,
            "reason": null,
            "response_union_seconds": 3.8429829700035043,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 231.25070963000516,
            "trapezoid_pieces": 5,
            "watt_hours": 0.24685903858592645
          },
          "gpu_sample_count": 3,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6711808000,
              "mean": 6708021930.666667,
              "median": 6709710848,
              "minimum": 6702546944,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 300.38,
              "mean": 216.23666666666668,
              "median": 180.59,
              "minimum": 167.74,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 6752829440.0,
              "mean": 6752829440.0,
              "median": 6752829440.0,
              "minimum": 6752829440.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1905.0,
              "mean": 1807.3333333333333,
              "median": 1777.0,
              "minimum": 1740.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 98.0,
              "median": 97.0,
              "minimum": 97.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96199966720,
              "mean": 96187517610.66667,
              "median": 96182734848,
              "minimum": 96179851264,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 815646720,
              "mean": 813635584,
              "median": 813856768,
              "minimum": 811403264,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 818634752,
              "mean": 816623616,
              "median": 816844800,
              "minimum": 814391296,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.8429829700035043,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 1.046531479980331,
            "observed_intervals": 1,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "2d1a82644c4ba9ba504ddf8fd2c7fe3c253e15cc9811363c52186bd4b7510a07",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 30,
        "host_samples": 30,
        "maximum_gpu_gap_seconds": 1.0048957320395857,
        "maximum_host_gap_seconds": 1.0701882099965587,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 30,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 30,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 268.15063832100714,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 20773.514431872452,
          "reason": null,
          "response_union_seconds": 69.25191831210395,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 299.9702381997644,
          "trapezoid_pieces": 78,
          "watt_hours": 5.770420675520126
        },
        "gpu_sample_count": 68,
        "host_sample_count": 68,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 17240788992,
            "mean": 17231835497.411766,
            "median": 17233764352.0,
            "minimum": 17221709824,
            "sample_count": 68
          },
          "gpu_board_watts": {
            "maximum": 301.79,
            "mean": 299.9689705882353,
            "median": 300.0,
            "minimum": 297.03,
            "sample_count": 68
          },
          "gpu_memory_used_bytes": {
            "maximum": 22502440960.0,
            "mean": 22502440960.0,
            "median": 22502440960.0,
            "minimum": 22502440960.0,
            "sample_count": 68
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1687.0,
            "mean": 1608.3088235294117,
            "median": 1605.0,
            "minimum": 1575.0,
            "sample_count": 68
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 99.02941176470588,
            "median": 99.0,
            "minimum": 84.0,
            "sample_count": 68
          },
          "host_mem_available_bytes": {
            "maximum": 96136097792,
            "mean": 96108860235.29411,
            "median": 96106113024.0,
            "minimum": 96084488192,
            "sample_count": 68
          },
          "process_pss_bytes": {
            "maximum": 925034496,
            "mean": 919076412.2352941,
            "median": 921339904.0,
            "minimum": 911089664,
            "sample_count": 68
          },
          "process_rss_bytes": {
            "maximum": 928030720,
            "mean": 922093507.7647059,
            "median": 924336128.0,
            "minimum": 915148800,
            "sample_count": 68
          }
        },
        "process_sample_count": 68,
        "response_union_seconds": 69.25191831210395,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 45056
          },
          "observed_endpoint_seconds": 59.35470441589132,
          "observed_intervals": 10,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 46.274918602713925,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1198924303190445,
            "interval_count": 5,
            "interval_seconds": 27.675233017944265,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.83655777020942,
            "process_cpu_seconds": 27.63
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 8298.985135398982,
              "reason": null,
              "response_union_seconds": 27.660772588045802,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 300.02723564509427,
              "trapezoid_pieces": 32,
              "watt_hours": 2.3052736487219394
            },
            "gpu_sample_count": 27,
            "host_sample_count": 27,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 17240031232,
                "mean": 17223894964.148148,
                "median": 17223282688,
                "minimum": 17221709824,
                "sample_count": 27
              },
              "gpu_board_watts": {
                "maximum": 301.79,
                "mean": 300.0274074074074,
                "median": 299.99,
                "minimum": 299.24,
                "sample_count": 27
              },
              "gpu_memory_used_bytes": {
                "maximum": 22502440960.0,
                "mean": 22502440960.0,
                "median": 22502440960.0,
                "minimum": 22502440960.0,
                "sample_count": 27
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1687.0,
                "mean": 1607.5925925925926,
                "median": 1605.0,
                "minimum": 1575.0,
                "sample_count": 27
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 99.0,
                "median": 99.0,
                "minimum": 95.0,
                "sample_count": 27
              },
              "host_mem_available_bytes": {
                "maximum": 96136097792,
                "mean": 96113169445.92592,
                "median": 96115363840,
                "minimum": 96087658496,
                "sample_count": 27
              },
              "process_pss_bytes": {
                "maximum": 913487872,
                "mean": 912757797.925926,
                "median": 912799744,
                "minimum": 911089664,
                "sample_count": 27
              },
              "process_rss_bytes": {
                "maximum": 916484096,
                "mean": 915806587.2592592,
                "median": 915795968,
                "minimum": 915148800,
                "sample_count": 27
              }
            },
            "process_sample_count": 27,
            "response_union_seconds": 27.660772588045802,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 20480
              },
              "observed_endpoint_seconds": 22.504645432985853,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 27.68297484901268,
          "response_seconds": {
            "maximum": 5.5716617869911715,
            "mean": 5.53215451760916,
            "median": 5.527461774996482,
            "minimum": 5.496413041022606,
            "sample_count": 5
          },
          "response_window_union_seconds": 27.660772588045802,
          "server_decode_tokens_per_second": {
            "maximum": 53.86729127375231,
            "mean": 53.507414673402344,
            "median": 53.55129058610313,
            "minimum": 53.12845334946771,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 0.7718379290308803,
            "mean": 0.7662811154150404,
            "median": 0.7656330339959823,
            "minimum": 0.7624987220042385,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 30.775781184108897,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1204527047148796,
            "interval_count": 5,
            "interval_seconds": 41.610548944969196,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.85448655087615,
            "process_cpu_seconds": 41.55
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 12474.529296473467,
              "reason": null,
              "response_union_seconds": 41.59114572405815,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 299.93233125236196,
              "trapezoid_pieces": 46,
              "watt_hours": 3.4651470267981854
            },
            "gpu_sample_count": 41,
            "host_sample_count": 41,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 17240788992,
                "mean": 17237064629.07317,
                "median": 17237659648,
                "minimum": 17232809984,
                "sample_count": 41
              },
              "gpu_board_watts": {
                "maximum": 300.28,
                "mean": 299.93048780487806,
                "median": 300.0,
                "minimum": 297.03,
                "sample_count": 41
              },
              "gpu_memory_used_bytes": {
                "maximum": 22502440960.0,
                "mean": 22502440960.0,
                "median": 22502440960.0,
                "minimum": 22502440960.0,
                "sample_count": 41
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1657.0,
                "mean": 1608.780487804878,
                "median": 1605.0,
                "minimum": 1590.0,
                "sample_count": 41
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 99.04878048780488,
                "median": 99.0,
                "minimum": 84.0,
                "sample_count": 41
              },
              "host_mem_available_bytes": {
                "maximum": 96128155648,
                "mean": 96106022462.43903,
                "median": 96104517632,
                "minimum": 96084488192,
                "sample_count": 41
              },
              "process_pss_bytes": {
                "maximum": 925034496,
                "mean": 923237450.9268292,
                "median": 923604992,
                "minimum": 920840192,
                "sample_count": 41
              },
              "process_rss_bytes": {
                "maximum": 928030720,
                "mean": 926233674.9268292,
                "median": 926601216,
                "minimum": 923836416,
                "sample_count": 41
              }
            },
            "process_sample_count": 41,
            "response_union_seconds": 41.59114572405815,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 24576
              },
              "observed_endpoint_seconds": 36.850058982905466,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 41.62048171600327,
          "response_seconds": {
            "maximum": 8.362828276003711,
            "mean": 8.31822914481163,
            "median": 8.326255902997218,
            "minimum": 8.262266772042494,
            "sample_count": 5
          },
          "response_window_union_seconds": 41.59114572405815,
          "server_decode_tokens_per_second": {
            "maximum": 51.867399264866066,
            "mean": 51.518764345880015,
            "median": 51.44895386118346,
            "minimum": 51.2462896681451,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 3.3867225820431486,
            "mean": 3.3683619036222807,
            "median": 3.3697669280227274,
            "minimum": 3.345733996015042,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "baseline-32b",
      "model": "qwen3-32b-q4",
      "module": "baseline",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 417.3
          },
          {
            "device": "CUDA0",
            "reported_mib": 18423.65
          }
        ],
        "gpu_layers": 65,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 65
      },
      "record_sha256": "31d7b3fec56be9142f89e611f4085991a36d0edab0bd028a37a8232b8093b11b",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16632758272,
              "mean": 16624463872,
              "median": 16624463872.0,
              "minimum": 16616169472,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.33,
              "mean": 16.21,
              "median": 16.21,
              "minimum": 16.09,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96517685248,
              "mean": 96516294656,
              "median": 96516294656.0,
              "minimum": 96514904064,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1007562470040284,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16893157376,
              "mean": 16893157376,
              "median": 16893157376,
              "minimum": 16893157376,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 94.6,
              "mean": 94.6,
              "median": 94.6,
              "minimum": 94.6,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2302.0,
              "mean": 2302.0,
              "median": 2302.0,
              "minimum": 2302.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96570462208,
              "mean": 96570462208,
              "median": 96570462208,
              "minimum": 96570462208,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1624215979827568,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 234.14076826269599,
            "reason": null,
            "response_union_seconds": 3.614590512996074,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 64.77656802917369,
            "trapezoid_pieces": 4,
            "watt_hours": 0.06503910229519333
          },
          "gpu_sample_count": 3,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17189969920,
              "mean": 17152371370.666666,
              "median": 17133707264,
              "minimum": 17133436928,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 78.92,
              "mean": 70.08,
              "median": 78.4,
              "minimum": 52.92,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 19906166784.0,
              "mean": 19906166784.0,
              "median": 19906166784.0,
              "minimum": 19906166784.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2300.0,
              "median": 2310.0,
              "minimum": 2280.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 76.0,
              "mean": 71.0,
              "median": 76.0,
              "minimum": 61.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96221175808,
              "mean": 96199781034.66667,
              "median": 96212836352,
              "minimum": 96165330944,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 20099049472,
              "mean": 13691381760,
              "median": 20098906112,
              "minimum": 876189696,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 20102045696,
              "mean": 13694377984,
              "median": 20101902336,
              "minimum": 879185920,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.614590512996074,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 13,
              "storage_read_bytes": 0,
              "storage_write_bytes": 8192
            },
            "observed_endpoint_seconds": 2.344856615003664,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 548.1502251793638,
            "reason": null,
            "response_union_seconds": 3.31754442199599,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 165.22769719224166,
            "trapezoid_pieces": 8,
            "watt_hours": 0.15226395143871216
          },
          "gpu_sample_count": 5,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17240248320,
              "mean": 17218326528,
              "median": 17224609792,
              "minimum": 17190121472,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 300.01,
              "mean": 183.016,
              "median": 115.73,
              "minimum": 87.89,
              "sample_count": 5
            },
            "gpu_memory_used_bytes": {
              "maximum": 22502440960.0,
              "mean": 22497407795.2,
              "median": 22502440960.0,
              "minimum": 22477275136.0,
              "sample_count": 5
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2340.0,
              "mean": 2029.4,
              "median": 2310.0,
              "minimum": 1575.0,
              "sample_count": 5
            },
            "gpu_utilization_percent": {
              "maximum": 99.0,
              "mean": 39.6,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 5
            },
            "host_mem_available_bytes": {
              "maximum": 96170061824,
              "mean": 96126780757.33333,
              "median": 96106954752,
              "minimum": 96103325696,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 925034496,
              "mean": 904964096,
              "median": 913561600,
              "minimum": 876296192,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 928030720,
              "mean": 907960320,
              "median": 916557824,
              "minimum": 879292416,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.31754442199599,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 20792.77551286387,
            "reason": null,
            "response_union_seconds": 69.31620680802735,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 299.9698983882642,
            "trapezoid_pieces": 70,
            "watt_hours": 5.77577097579552
          },
          "gpu_sample_count": 68,
          "host_sample_count": 68,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17240788992,
              "mean": 17231835497.411766,
              "median": 17233764352.0,
              "minimum": 17221709824,
              "sample_count": 68
            },
            "gpu_board_watts": {
              "maximum": 301.79,
              "mean": 299.9689705882353,
              "median": 300.0,
              "minimum": 297.03,
              "sample_count": 68
            },
            "gpu_memory_used_bytes": {
              "maximum": 22502440960.0,
              "mean": 22502440960.0,
              "median": 22502440960.0,
              "minimum": 22502440960.0,
              "sample_count": 68
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1687.0,
              "mean": 1608.3088235294117,
              "median": 1605.0,
              "minimum": 1575.0,
              "sample_count": 68
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 99.02941176470588,
              "median": 99.0,
              "minimum": 84.0,
              "sample_count": 68
            },
            "host_mem_available_bytes": {
              "maximum": 96136097792,
              "mean": 96108860235.29411,
              "median": 96106113024.0,
              "minimum": 96084488192,
              "sample_count": 68
            },
            "process_pss_bytes": {
              "maximum": 925034496,
              "mean": 919076412.2352941,
              "median": 921339904.0,
              "minimum": 911089664,
              "sample_count": 68
            },
            "process_rss_bytes": {
              "maximum": 928030720,
              "mean": 922093507.7647059,
              "median": 924336128.0,
              "minimum": 915148800,
              "sample_count": 68
            }
          },
          "process_sample_count": 68,
          "response_union_seconds": 69.31620680802735,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 65536
            },
            "observed_endpoint_seconds": 67.56908865296282,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 3891.3219742636566,
            "reason": null,
            "response_union_seconds": 13.74247567699058,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 283.1601863977834,
            "trapezoid_pieces": 16,
            "watt_hours": 1.0809227706287934
          },
          "gpu_sample_count": 14,
          "host_sample_count": 13,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17231908864,
              "mean": 17226236534.153847,
              "median": 17225515008,
              "minimum": 17220960256,
              "sample_count": 13
            },
            "gpu_board_watts": {
              "maximum": 306.92,
              "mean": 284.82928571428573,
              "median": 299.99,
              "minimum": 106.16,
              "sample_count": 14
            },
            "gpu_memory_used_bytes": {
              "maximum": 22502440960.0,
              "mean": 22502440960.0,
              "median": 22502440960.0,
              "minimum": 22502440960.0,
              "sample_count": 14
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2197.0,
              "mean": 1657.7142857142858,
              "median": 1623.5,
              "minimum": 1590.0,
              "sample_count": 14
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 99.28571428571429,
              "median": 99.0,
              "minimum": 99.0,
              "sample_count": 14
            },
            "host_mem_available_bytes": {
              "maximum": 96137310208,
              "mean": 96114068558.76923,
              "median": 96116252672,
              "minimum": 96089333760,
              "sample_count": 13
            },
            "process_pss_bytes": {
              "maximum": 920127488,
              "mean": 915585339.0769231,
              "median": 914270208,
              "minimum": 911812608,
              "sample_count": 13
            },
            "process_rss_bytes": {
              "maximum": 923123712,
              "mean": 918581563.0769231,
              "median": 917266432,
              "minimum": 914808832,
              "sample_count": 13
            }
          },
          "process_sample_count": 13,
          "response_union_seconds": 13.74247567699058,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 8,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 1372160,
              "storage_write_bytes": 12288
            },
            "observed_endpoint_seconds": 11.253299642936327,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "59f09a9b924c4cb4ae1370f85109347542e72457c07f0a7a8ff351b31fedab64",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 93,
        "host_samples": 90,
        "maximum_gpu_gap_seconds": 1.0044987400178798,
        "maximum_host_gap_seconds": 1.322291104006581,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 93,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 93,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 185.7723383929697,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 43316.62505863769,
          "reason": null,
          "response_union_seconds": 144.44773938611615,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 299.8774867833006,
          "trapezoid_pieces": 155,
          "watt_hours": 12.032395849621581
        },
        "gpu_sample_count": 145,
        "host_sample_count": 142,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 783441920,
            "mean": 744136242.4788733,
            "median": 753983488.0,
            "minimum": 708935680,
            "sample_count": 142
          },
          "gpu_board_watts": {
            "maximum": 302.93,
            "mean": 300.00172413793103,
            "median": 300.0,
            "minimum": 297.31,
            "sample_count": 145
          },
          "gpu_memory_used_bytes": {
            "maximum": 47104131072.0,
            "mean": 47104131072.0,
            "median": 47104131072.0,
            "minimum": 47104131072.0,
            "sample_count": 145
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1635.0,
            "mean": 1481.2551724137932,
            "median": 1485.0,
            "minimum": 1312.0,
            "sample_count": 145
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 99.84827586206896,
            "median": 100.0,
            "minimum": 83.0,
            "sample_count": 145
          },
          "host_mem_available_bytes": {
            "maximum": 96093003776,
            "mean": 96040983393.35211,
            "median": 96037416960.0,
            "minimum": 96006701056,
            "sample_count": 142
          },
          "process_pss_bytes": {
            "maximum": 1257277440,
            "mean": 1236598834.4788733,
            "median": 1246618624.0,
            "minimum": 1205324800,
            "sample_count": 142
          },
          "process_rss_bytes": {
            "maximum": 1260240896,
            "mean": 1239593392.6760564,
            "median": 1249583104.0,
            "minimum": 1209503744,
            "sample_count": 142
          }
        },
        "process_sample_count": 142,
        "response_union_seconds": 144.44773938611615,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 102400
          },
          "observed_endpoint_seconds": 135.37246441596653,
          "observed_intervals": 10,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 22.20832513756411,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1211448738400773,
            "interval_count": 5,
            "interval_seconds": 57.6511207500007,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.87663596288247,
            "process_cpu_seconds": 57.58
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 17282.92564468072,
              "reason": null,
              "response_union_seconds": 57.63604378409218,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 299.86315003547986,
              "trapezoid_pieces": 63,
              "watt_hours": 4.8008126790779775
            },
            "gpu_sample_count": 58,
            "host_sample_count": 57,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 748466176,
                "mean": 716765866.6666666,
                "median": 711774208,
                "minimum": 708935680,
                "sample_count": 57
              },
              "gpu_board_watts": {
                "maximum": 300.95,
                "mean": 300.01241379310346,
                "median": 299.99,
                "minimum": 299.66,
                "sample_count": 58
              },
              "gpu_memory_used_bytes": {
                "maximum": 47104131072.0,
                "mean": 47104131072.0,
                "median": 47104131072.0,
                "minimum": 47104131072.0,
                "sample_count": 58
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1627.0,
                "mean": 1446.4827586206898,
                "median": 1451.0,
                "minimum": 1312.0,
                "sample_count": 58
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 100.0,
                "median": 100.0,
                "minimum": 100.0,
                "sample_count": 58
              },
              "host_mem_available_bytes": {
                "maximum": 96093003776,
                "mean": 96060373171.64912,
                "median": 96058638336,
                "minimum": 96031748096,
                "sample_count": 57
              },
              "process_pss_bytes": {
                "maximum": 1225331712,
                "mean": 1211821738.6666667,
                "median": 1207718912,
                "minimum": 1205324800,
                "sample_count": 57
              },
              "process_rss_bytes": {
                "maximum": 1228296192,
                "mean": 1214836376.7017543,
                "median": 1210683392,
                "minimum": 1209503744,
                "sample_count": 57
              }
            },
            "process_sample_count": 57,
            "response_union_seconds": 57.63604378409218,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 40960
              },
              "observed_endpoint_seconds": 53.343125255953055,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 57.6594875259907,
          "response_seconds": {
            "maximum": 11.647003977035638,
            "mean": 11.527208756818435,
            "median": 11.566605183004867,
            "minimum": 11.37437688402133,
            "sample_count": 5
          },
          "response_window_union_seconds": 57.63604378409218,
          "server_decode_tokens_per_second": {
            "maximum": 26.211279772416194,
            "mean": 25.85845974942399,
            "median": 25.778169650885744,
            "minimum": 25.605412361610277,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 1.6881801380077377,
            "mean": 1.665065498405602,
            "median": 1.6596780359977856,
            "minimum": 1.645603878016118,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 14.7445570683008,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1216834193859877,
            "interval_count": 5,
            "interval_seconds": 86.83215546992142,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.8938694203516,
            "process_cpu_seconds": 86.74000000000001
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 26033.69941395698,
              "reason": null,
              "response_union_seconds": 86.81169560202397,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 299.88700524068577,
              "trapezoid_pieces": 92,
              "watt_hours": 7.2315831705436056
            },
            "gpu_sample_count": 87,
            "host_sample_count": 85,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 783441920,
                "mean": 762490494.4941176,
                "median": 764178432,
                "minimum": 749699072,
                "sample_count": 85
              },
              "gpu_board_watts": {
                "maximum": 302.93,
                "mean": 299.9945977011494,
                "median": 300.0,
                "minimum": 297.31,
                "sample_count": 87
              },
              "gpu_memory_used_bytes": {
                "maximum": 47104131072.0,
                "mean": 47104131072.0,
                "median": 47104131072.0,
                "minimum": 47104131072.0,
                "sample_count": 87
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1635.0,
                "mean": 1504.4367816091954,
                "median": 1485.0,
                "minimum": 1462.0,
                "sample_count": 87
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 99.74712643678161,
                "median": 100.0,
                "minimum": 83.0,
                "sample_count": 87
              },
              "host_mem_available_bytes": {
                "maximum": 96051191808,
                "mean": 96027980836.14117,
                "median": 96027217920,
                "minimum": 96006701056,
                "sample_count": 85
              },
              "process_pss_bytes": {
                "maximum": 1257277440,
                "mean": 1253214063.4352942,
                "median": 1255912448,
                "minimum": 1243612160,
                "sample_count": 85
              },
              "process_rss_bytes": {
                "maximum": 1260240896,
                "mean": 1256195156.3294117,
                "median": 1258876928,
                "minimum": 1246576640,
                "sample_count": 85
              }
            },
            "process_sample_count": 85,
            "response_union_seconds": 86.81169560202397,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 61440
              },
              "observed_endpoint_seconds": 82.02933916001348,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 86.84236659400631,
          "response_seconds": {
            "maximum": 17.491869097051676,
            "mean": 17.362339120404794,
            "median": 17.35889967199182,
            "minimum": 17.22252958599711,
            "sample_count": 5
          },
          "response_window_union_seconds": 86.81169560202397,
          "server_decode_tokens_per_second": {
            "maximum": 25.05988575640317,
            "mean": 24.836448251688132,
            "median": 24.82518929206835,
            "minimum": 24.64976311577646,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 7.146799207024742,
            "mean": 7.094700342812575,
            "median": 7.08692858600989,
            "minimum": 7.046769339009188,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "offload-72b-full",
      "model": "qwen25-72b-q4",
      "module": "dense-offload",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 668.25
          },
          {
            "device": "CUDA0",
            "reported_mib": 41297.69
          }
        ],
        "gpu_layers": 81,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 81
      },
      "record_sha256": "f8b20b521b99946cae8150a28b03c57e74bb9a19fc93a5809019ef9e4d3899c0",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 358690816,
              "mean": 349544448,
              "median": 349544448.0,
              "minimum": 340398080,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 15.22,
              "mean": 14.965,
              "median": 14.965,
              "minimum": 14.71,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96567312384,
              "mean": 96558721024,
              "median": 96558721024.0,
              "minimum": 96550129664,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1008167749969289,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 352747520,
              "mean": 352747520,
              "median": 352747520,
              "minimum": 352747520,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 103.77,
              "mean": 103.77,
              "median": 103.77,
              "minimum": 103.77,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2340.0,
              "mean": 2340.0,
              "median": 2340.0,
              "minimum": 2340.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96511545344,
              "mean": 96511545344,
              "median": 96511545344,
              "minimum": 96511545344,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1564114179927856,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 478.07168217720226,
            "reason": null,
            "response_union_seconds": 7.210961889009923,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 66.29790720511522,
            "trapezoid_pieces": 8,
            "watt_hours": 0.1327976894936673
          },
          "gpu_sample_count": 7,
          "host_sample_count": 4,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 656150528,
              "mean": 638523392,
              "median": 638926848.0,
              "minimum": 620089344,
              "sample_count": 4
            },
            "gpu_board_watts": {
              "maximum": 74.94,
              "mean": 69.43428571428572,
              "median": 74.31,
              "minimum": 42.68,
              "sample_count": 7
            },
            "gpu_memory_used_bytes": {
              "maximum": 43891294208.0,
              "mean": 37704995401.14286,
              "median": 43891294208.0,
              "minimum": 587202560.0,
              "sample_count": 7
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2295.0,
              "mean": 2291.714285714286,
              "median": 2295.0,
              "minimum": 2272.0,
              "sample_count": 7
            },
            "gpu_utilization_percent": {
              "maximum": 83.0,
              "mean": 58.857142857142854,
              "median": 82.0,
              "minimum": 0.0,
              "sample_count": 7
            },
            "host_mem_available_bytes": {
              "maximum": 96185810944,
              "mean": 96155684864,
              "median": 96151988224.0,
              "minimum": 96132952064,
              "sample_count": 4
            },
            "process_pss_bytes": {
              "maximum": 44348953600,
              "mean": 42197760000,
              "median": 44348720128.0,
              "minimum": 35744646144,
              "sample_count": 4
            },
            "process_rss_bytes": {
              "maximum": 44351918080,
              "mean": 42200724480,
              "median": 44351684608.0,
              "minimum": 35747610624,
              "sample_count": 4
            }
          },
          "process_sample_count": 4,
          "response_union_seconds": 7.210961889009923,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 5,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 5.209772021975368,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 551.952217391011,
            "reason": null,
            "response_union_seconds": 3.3186683430103585,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 166.3173780391493,
            "trapezoid_pieces": 6,
            "watt_hours": 0.15332006038639195
          },
          "gpu_sample_count": 3,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 769622016,
              "mean": 725244586.6666666,
              "median": 737775616,
              "minimum": 668336128,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 205.83,
              "mean": 158.66,
              "median": 175.53,
              "minimum": 94.62,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 47104131072.0,
              "mean": 47093645312.0,
              "median": 47104131072.0,
              "minimum": 47072673792.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2340.0,
              "mean": 2314.6666666666665,
              "median": 2332.0,
              "minimum": 2272.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96088260608,
              "mean": 96051089408,
              "median": 96037539840,
              "minimum": 96027467776,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 1257276416,
              "mean": 1216586752,
              "median": 1231762432,
              "minimum": 1160721408,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 1260240896,
              "mean": 1219551232,
              "median": 1234726912,
              "minimum": 1163685888,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.3186683430103585,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 43336.282681437755,
            "reason": null,
            "response_union_seconds": 144.51526583201485,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 299.87339006670777,
            "trapezoid_pieces": 147,
            "watt_hours": 12.037856300399376
          },
          "gpu_sample_count": 145,
          "host_sample_count": 142,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 783441920,
              "mean": 744136242.4788733,
              "median": 753983488.0,
              "minimum": 708935680,
              "sample_count": 142
            },
            "gpu_board_watts": {
              "maximum": 302.93,
              "mean": 300.00172413793103,
              "median": 300.0,
              "minimum": 297.31,
              "sample_count": 145
            },
            "gpu_memory_used_bytes": {
              "maximum": 47104131072.0,
              "mean": 47104131072.0,
              "median": 47104131072.0,
              "minimum": 47104131072.0,
              "sample_count": 145
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1635.0,
              "mean": 1481.2551724137932,
              "median": 1485.0,
              "minimum": 1312.0,
              "sample_count": 145
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 99.84827586206896,
              "median": 100.0,
              "minimum": 83.0,
              "sample_count": 145
            },
            "host_mem_available_bytes": {
              "maximum": 96093003776,
              "mean": 96040983393.35211,
              "median": 96037416960.0,
              "minimum": 96006701056,
              "sample_count": 142
            },
            "process_pss_bytes": {
              "maximum": 1257277440,
              "mean": 1236598834.4788733,
              "median": 1246618624.0,
              "minimum": 1205324800,
              "sample_count": 142
            },
            "process_rss_bytes": {
              "maximum": 1260240896,
              "mean": 1239593392.6760564,
              "median": 1249583104.0,
              "minimum": 1209503744,
              "sample_count": 142
            }
          },
          "process_sample_count": 142,
          "response_union_seconds": 144.51526583201485,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 122880
            },
            "observed_endpoint_seconds": 143.58244017901598,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 8285.867276863784,
            "reason": null,
            "response_union_seconds": 28.43168306199368,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 291.4307696381153,
            "trapezoid_pieces": 30,
            "watt_hours": 2.301629799128829
          },
          "gpu_sample_count": 28,
          "host_sample_count": 27,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 748781568,
              "mean": 728878952.2962962,
              "median": 738770944,
              "minimum": 700723200,
              "sample_count": 27
            },
            "gpu_board_watts": {
              "maximum": 300.16,
              "mean": 292.4089285714286,
              "median": 300.0,
              "minimum": 186.04,
              "sample_count": 28
            },
            "gpu_memory_used_bytes": {
              "maximum": 47104131072.0,
              "mean": 47104131072.0,
              "median": 47104131072.0,
              "minimum": 47104131072.0,
              "sample_count": 28
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1642.0,
              "mean": 1531.3214285714287,
              "median": 1511.0,
              "minimum": 1492.0,
              "sample_count": 28
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 100.0,
              "median": 100.0,
              "minimum": 100.0,
              "sample_count": 28
            },
            "host_mem_available_bytes": {
              "maximum": 96063094784,
              "mean": 96039131060.14815,
              "median": 96037965824,
              "minimum": 96024268800,
              "sample_count": 27
            },
            "process_pss_bytes": {
              "maximum": 1242903552,
              "mean": 1224316662.5185184,
              "median": 1232471040,
              "minimum": 1196975104,
              "sample_count": 27
            },
            "process_rss_bytes": {
              "maximum": 1245868032,
              "mean": 1227281142.5185184,
              "median": 1235435520,
              "minimum": 1199939584,
              "sample_count": 27
            }
          },
          "process_sample_count": 27,
          "response_union_seconds": 28.43168306199368,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 12288
            },
            "observed_endpoint_seconds": 25.652496571012307,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "4b17dd2a9c39d7cab209a102bfdbd49f8324b54625bc01cbe7d9aeaf6b87dd5a",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 186,
        "host_samples": 179,
        "maximum_gpu_gap_seconds": 1.0065537570044398,
        "maximum_host_gap_seconds": 1.7706437300075777,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 186,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 186,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 64.85315055400133,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 5556.354650300507,
          "reason": null,
          "response_union_seconds": 18.690550074737985,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 297.28149402143265,
          "trapezoid_pieces": 29,
          "watt_hours": 1.5434318473056965
        },
        "gpu_sample_count": 19,
        "host_sample_count": 18,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 84638580736,
            "mean": 84628920547.55556,
            "median": 84632049664.0,
            "minimum": 84617678848,
            "sample_count": 18
          },
          "gpu_board_watts": {
            "maximum": 303.42,
            "mean": 298.6694736842105,
            "median": 298.73,
            "minimum": 291.54,
            "sample_count": 19
          },
          "gpu_memory_used_bytes": {
            "maximum": 63942164480.0,
            "mean": 63942164480.0,
            "median": 63942164480.0,
            "minimum": 63942164480.0,
            "sample_count": 19
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2212.0,
            "mean": 2094.2631578947367,
            "median": 2175.0,
            "minimum": 1837.0,
            "sample_count": 19
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 96.84210526315789,
            "median": 96.0,
            "minimum": 94.0,
            "sample_count": 19
          },
          "host_mem_available_bytes": {
            "maximum": 95839178752,
            "mean": 95816025201.77777,
            "median": 95815591936.0,
            "minimum": 95802257408,
            "sample_count": 18
          },
          "process_pss_bytes": {
            "maximum": 1353333760,
            "mean": 1348224227.5555556,
            "median": 1350933504.0,
            "minimum": 1341737984,
            "sample_count": 18
          },
          "process_rss_bytes": {
            "maximum": 1356292096,
            "mean": 1351182563.5555556,
            "median": 1353891840.0,
            "minimum": 1344696320,
            "sample_count": 18
          }
        },
        "process_sample_count": 18,
        "response_union_seconds": 18.690550074737985,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 4096
          },
          "observed_endpoint_seconds": 8.252795126929414,
          "observed_intervals": 7,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 178.45760431710494,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1128260975148874,
            "interval_count": 5,
            "interval_seconds": 7.1880019310628995,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.6104351204764,
            "process_cpu_seconds": 7.16
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 2130.4346352334737,
              "reason": null,
              "response_union_seconds": 7.172571910836268,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 297.0252040296493,
              "trapezoid_pieces": 12,
              "watt_hours": 0.591787398675965
            },
            "gpu_sample_count": 7,
            "host_sample_count": 7,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 84621459456,
                "mean": 84619137024,
                "median": 84619370496,
                "minimum": 84617678848,
                "sample_count": 7
              },
              "gpu_board_watts": {
                "maximum": 300.9,
                "mean": 299.23285714285714,
                "median": 300.14,
                "minimum": 296.54,
                "sample_count": 7
              },
              "gpu_memory_used_bytes": {
                "maximum": 63942164480.0,
                "mean": 63942164480.0,
                "median": 63942164480.0,
                "minimum": 63942164480.0,
                "sample_count": 7
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2205.0,
                "mean": 2170.5714285714284,
                "median": 2175.0,
                "minimum": 2077.0,
                "sample_count": 7
              },
              "gpu_utilization_percent": {
                "maximum": 96.0,
                "mean": 95.71428571428571,
                "median": 96.0,
                "minimum": 94.0,
                "sample_count": 7
              },
              "host_mem_available_bytes": {
                "maximum": 95839178752,
                "mean": 95822938697.14285,
                "median": 95823192064,
                "minimum": 95802257408,
                "sample_count": 7
              },
              "process_pss_bytes": {
                "maximum": 1342917632,
                "mean": 1342255250.2857144,
                "median": 1342327808,
                "minimum": 1341737984,
                "sample_count": 7
              },
              "process_rss_bytes": {
                "maximum": 1345875968,
                "mean": 1345213586.2857144,
                "median": 1345286144,
                "minimum": 1344696320,
                "sample_count": 7
              }
            },
            "process_sample_count": 7,
            "response_union_seconds": 7.172571910836268,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 4096
              },
              "observed_endpoint_seconds": 2.0736262949649245,
              "observed_intervals": 2,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 7.20023208897328,
          "response_seconds": {
            "maximum": 1.4361235249671154,
            "mean": 1.4345143821672537,
            "median": 1.4348168549477123,
            "minimum": 1.4328763849916868,
            "sample_count": 5
          },
          "response_window_union_seconds": 7.172571910836268,
          "server_decode_tokens_per_second": {
            "maximum": 221.8905948843341,
            "mean": 221.75877497939115,
            "median": 221.6888167120332,
            "minimum": 221.6466170641819,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 0.2855251739965752,
            "mean": 0.28449933978263287,
            "median": 0.28445326996734366,
            "minimum": 0.28353577299276367,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 111.13061526819216,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1147665128873214,
            "interval_count": 5,
            "interval_seconds": 11.53778296103701,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.67252841239429,
            "process_cpu_seconds": 11.5
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 3425.920015067035,
              "reason": null,
              "response_union_seconds": 11.517978163901716,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 297.44109307344826,
              "trapezoid_pieces": 17,
              "watt_hours": 0.951644448629732
            },
            "gpu_sample_count": 12,
            "host_sample_count": 11,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 84638580736,
                "mean": 84635146426.18182,
                "median": 84635049984,
                "minimum": 84631072768,
                "sample_count": 11
              },
              "gpu_board_watts": {
                "maximum": 303.42,
                "mean": 298.3408333333333,
                "median": 297.11,
                "minimum": 291.54,
                "sample_count": 12
              },
              "gpu_memory_used_bytes": {
                "maximum": 63942164480.0,
                "mean": 63942164480.0,
                "median": 63942164480.0,
                "minimum": 63942164480.0,
                "sample_count": 12
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2212.0,
                "mean": 2049.75,
                "median": 2114.5,
                "minimum": 1837.0,
                "sample_count": 12
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 97.5,
                "median": 96.0,
                "minimum": 95.0,
                "sample_count": 12
              },
              "host_mem_available_bytes": {
                "maximum": 95829737472,
                "mean": 95811625704.72728,
                "median": 95808782336,
                "minimum": 95802593280,
                "sample_count": 11
              },
              "process_pss_bytes": {
                "maximum": 1353333760,
                "mean": 1352022667.6363637,
                "median": 1351961600,
                "minimum": 1350581248,
                "sample_count": 11
              },
              "process_rss_bytes": {
                "maximum": 1356292096,
                "mean": 1354981003.6363637,
                "median": 1354919936,
                "minimum": 1353539584,
                "sample_count": 11
              }
            },
            "process_sample_count": 11,
            "response_union_seconds": 11.517978163901716,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 0
              },
              "observed_endpoint_seconds": 6.179168831964489,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 11.548840367991943,
          "response_seconds": {
            "maximum": 2.306483986962121,
            "mean": 2.3035956327803433,
            "median": 2.3047796469763853,
            "minimum": 2.2995725619839504,
            "sample_count": 5
          },
          "response_window_union_seconds": 11.517978163901716,
          "server_decode_tokens_per_second": {
            "maximum": 215.594317441073,
            "mean": 215.5020711089472,
            "median": 215.50977357951592,
            "minimum": 215.41492293214347,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 1.122534021968022,
            "mean": 1.1201264659990557,
            "median": 1.1210324589628726,
            "minimum": 1.1166345790261403,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "baseline-gpt120",
      "model": "gpt-oss-120b-mxfp4",
      "module": "baseline",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 586.82
          },
          {
            "device": "CUDA0",
            "reported_mib": 59851.68
          }
        ],
        "gpu_layers": 37,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 37
      },
      "record_sha256": "a3433171db7e279c742e5d12ad982b7442dafada0598259e27b35e14652f685c",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 41687240704,
              "mean": 41678069760,
              "median": 41678069760.0,
              "minimum": 41668898816,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 15.04,
              "mean": 14.915,
              "median": 14.915,
              "minimum": 14.79,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96659554304,
              "mean": 96650287104,
              "median": 96650287104.0,
              "minimum": 96641019904,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1007708329707384,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84053004288,
              "mean": 84053004288,
              "median": 84053004288,
              "minimum": 84053004288,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 86.59,
              "mean": 86.59,
              "median": 86.59,
              "minimum": 86.59,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2332.0,
              "mean": 2332.0,
              "median": 2332.0,
              "minimum": 2332.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96486580224,
              "mean": 96486580224,
              "median": 96486580224,
              "minimum": 96486580224,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1648289290023968,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 2002.2728389132026,
            "reason": null,
            "response_union_seconds": 36.58676577400183,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 54.72669684118394,
            "trapezoid_pieces": 37,
            "watt_hours": 0.5561868996981119
          },
          "gpu_sample_count": 36,
          "host_sample_count": 30,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84388233216,
              "mean": 60147514436.26667,
              "median": 56929634304.0,
              "minimum": 41429835776,
              "sample_count": 30
            },
            "gpu_board_watts": {
              "maximum": 76.39,
              "mean": 55.100833333333334,
              "median": 66.89,
              "minimum": 15.26,
              "sample_count": 36
            },
            "gpu_memory_used_bytes": {
              "maximum": 63346573312.0,
              "mean": 16277045248.0,
              "median": 587202560.0,
              "minimum": 587202560.0,
              "sample_count": 36
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2302.0,
              "mean": 1730.6666666666667,
              "median": 2272.0,
              "minimum": 180.0,
              "sample_count": 36
            },
            "gpu_utilization_percent": {
              "maximum": 81.0,
              "mean": 16.61111111111111,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 36
            },
            "host_mem_available_bytes": {
              "maximum": 96304549888,
              "mean": 96224268151.46666,
              "median": 96237541376.0,
              "minimum": 96072572928,
              "sample_count": 30
            },
            "process_pss_bytes": {
              "maximum": 63779191808,
              "mean": 34062087236.266666,
              "median": 33919894528.0,
              "minimum": 721088512,
              "sample_count": 30
            },
            "process_rss_bytes": {
              "maximum": 63782150144,
              "mean": 34065046186.666668,
              "median": 33922852864.0,
              "minimum": 724049920,
              "sample_count": 30
            }
          },
          "process_sample_count": 30,
          "response_union_seconds": 36.58676577400183,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 5,
              "read_syscall_bytes": 5,
              "storage_read_bytes": 63054757888,
              "storage_write_bytes": 4096
            },
            "observed_endpoint_seconds": 35.40200611401815,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 514.5725684548398,
            "reason": null,
            "response_union_seconds": 3.3283357390318997,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 154.6035643040361,
            "trapezoid_pieces": 6,
            "watt_hours": 0.14293682457078885
          },
          "gpu_sample_count": 3,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84639125504,
              "mean": 84545115477.33333,
              "median": 84621885440,
              "minimum": 84374335488,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 202.18,
              "mean": 157.62,
              "median": 196.68,
              "minimum": 74.0,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 63942164480.0,
              "mean": 63922591061.333336,
              "median": 63942164480.0,
              "minimum": 63883444224.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2332.0,
              "mean": 2314.6666666666665,
              "median": 2317.0,
              "minimum": 2295.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96038072320,
              "mean": 95896902314.66667,
              "median": 95834923008,
              "minimum": 95817711616,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 1353333760,
              "mean": 1268859221.3333333,
              "median": 1343036416,
              "minimum": 1110207488,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 1356292096,
              "mean": 1271817557.3333333,
              "median": 1345994752,
              "minimum": 1113165824,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.3283357390318997,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 5577.166580078889,
            "reason": null,
            "response_union_seconds": 18.76270647603087,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 297.24744600165496,
            "trapezoid_pieces": 21,
            "watt_hours": 1.5492129389108025
          },
          "gpu_sample_count": 19,
          "host_sample_count": 18,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84638580736,
              "mean": 84628920547.55556,
              "median": 84632049664.0,
              "minimum": 84617678848,
              "sample_count": 18
            },
            "gpu_board_watts": {
              "maximum": 303.42,
              "mean": 298.6694736842105,
              "median": 298.73,
              "minimum": 291.54,
              "sample_count": 19
            },
            "gpu_memory_used_bytes": {
              "maximum": 63942164480.0,
              "mean": 63942164480.0,
              "median": 63942164480.0,
              "minimum": 63942164480.0,
              "sample_count": 19
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2212.0,
              "mean": 2094.2631578947367,
              "median": 2175.0,
              "minimum": 1837.0,
              "sample_count": 19
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 96.84210526315789,
              "median": 96.0,
              "minimum": 94.0,
              "sample_count": 19
            },
            "host_mem_available_bytes": {
              "maximum": 95839178752,
              "mean": 95816025201.77777,
              "median": 95815591936.0,
              "minimum": 95802257408,
              "sample_count": 18
            },
            "process_pss_bytes": {
              "maximum": 1353333760,
              "mean": 1348224227.5555556,
              "median": 1350933504.0,
              "minimum": 1341737984,
              "sample_count": 18
            },
            "process_rss_bytes": {
              "maximum": 1356292096,
              "mean": 1351182563.5555556,
              "median": 1353891840.0,
              "minimum": 1344696320,
              "sample_count": 18
            }
          },
          "process_sample_count": 18,
          "response_union_seconds": 18.76270647603087,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 28672
            },
            "observed_endpoint_seconds": 16.407505154958926,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 876.6143908214965,
            "reason": null,
            "response_union_seconds": 3.870858257985674,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 226.4651228220563,
            "trapezoid_pieces": 6,
            "watt_hours": 0.2435039974504157
          },
          "gpu_sample_count": 4,
          "host_sample_count": 4,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84629381120,
              "mean": 84622595072,
              "median": 84622559232.0,
              "minimum": 84615880704,
              "sample_count": 4
            },
            "gpu_board_watts": {
              "maximum": 296.46,
              "mean": 207.48,
              "median": 220.79,
              "minimum": 91.88,
              "sample_count": 4
            },
            "gpu_memory_used_bytes": {
              "maximum": 63942164480.0,
              "mean": 63942164480.0,
              "median": 63942164480.0,
              "minimum": 63942164480.0,
              "sample_count": 4
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2287.0,
              "mean": 2141.0,
              "median": 2193.5,
              "minimum": 1890.0,
              "sample_count": 4
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 95.5,
              "median": 96.0,
              "minimum": 90.0,
              "sample_count": 4
            },
            "host_mem_available_bytes": {
              "maximum": 95830974464,
              "mean": 95809520640,
              "median": 95810375680.0,
              "minimum": 95786356736,
              "sample_count": 4
            },
            "process_pss_bytes": {
              "maximum": 1349901312,
              "mean": 1344703488,
              "median": 1343734784.0,
              "minimum": 1341443072,
              "sample_count": 4
            },
            "process_rss_bytes": {
              "maximum": 1352859648,
              "mean": 1347661824,
              "median": 1346693120.0,
              "minimum": 1344401408,
              "sample_count": 4
            }
          },
          "process_sample_count": 4,
          "response_union_seconds": 3.870858257985674,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 8,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 1376256,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 2.0698660400230438,
            "observed_intervals": 1,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 65,
        "host_samples": 58,
        "maximum_gpu_gap_seconds": 1.0076217760215513,
        "maximum_host_gap_seconds": 1.9519810149795376,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 65,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 65,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 355.6453820169554,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 55857.08840928863,
          "reason": null,
          "response_union_seconds": 186.21945289394353,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 299.95302607349294,
          "trapezoid_pieces": 196,
          "watt_hours": 15.515857891469064
        },
        "gpu_sample_count": 186,
        "host_sample_count": 180,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 86966120448,
            "mean": 86940043855.64444,
            "median": 86937204736.0,
            "minimum": 86925795328,
            "sample_count": 180
          },
          "gpu_board_watts": {
            "maximum": 300.63,
            "mean": 299.99005376344087,
            "median": 299.99,
            "minimum": 299.2,
            "sample_count": 186
          },
          "gpu_memory_used_bytes": {
            "maximum": 80000057344.0,
            "mean": 80000057344.0,
            "median": 80000057344.0,
            "minimum": 80000057344.0,
            "sample_count": 186
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1740.0,
            "mean": 1627.5645161290322,
            "median": 1657.0,
            "minimum": 1440.0,
            "sample_count": 186
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 99.9247311827957,
            "median": 100.0,
            "minimum": 89.0,
            "sample_count": 186
          },
          "host_mem_available_bytes": {
            "maximum": 96114024448,
            "mean": 96054613151.2889,
            "median": 96044468224.0,
            "minimum": 96015052800,
            "sample_count": 180
          },
          "process_pss_bytes": {
            "maximum": 1858872320,
            "mean": 1839274490.3111112,
            "median": 1852769280.0,
            "minimum": 1811305472,
            "sample_count": 180
          },
          "process_rss_bytes": {
            "maximum": 1861840896,
            "mean": 1842266703.6444445,
            "median": 1855737856.0,
            "minimum": 1814274048,
            "sample_count": 180
          }
        },
        "process_sample_count": 180,
        "response_union_seconds": 186.21945289394353,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 106496
          },
          "observed_endpoint_seconds": 174.8141460539773,
          "observed_intervals": 10,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 16.40063817021913,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.1211453356351977,
            "interval_count": 5,
            "interval_seconds": 78.06629419594537,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.87665074032633,
            "process_cpu_seconds": 77.97
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 23405.402069666456,
              "reason": null,
              "response_union_seconds": 78.04574350797338,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 299.8933832602324,
              "trapezoid_pieces": 83,
              "watt_hours": 6.5015005749073485
            },
            "gpu_sample_count": 78,
            "host_sample_count": 76,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86966120448,
                "mean": 86947034165.89473,
                "median": 86946766848.0,
                "minimum": 86936510464,
                "sample_count": 76
              },
              "gpu_board_watts": {
                "maximum": 300.33,
                "mean": 299.98320512820516,
                "median": 299.99,
                "minimum": 299.2,
                "sample_count": 78
              },
              "gpu_memory_used_bytes": {
                "maximum": 80000057344.0,
                "mean": 80000057344.0,
                "median": 80000057344.0,
                "minimum": 80000057344.0,
                "sample_count": 78
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1740.0,
                "mean": 1680.9358974358975,
                "median": 1698.5,
                "minimum": 1492.0,
                "sample_count": 78
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 99.83333333333333,
                "median": 100.0,
                "minimum": 89.0,
                "sample_count": 78
              },
              "host_mem_available_bytes": {
                "maximum": 96114024448,
                "mean": 96081347530.10527,
                "median": 96080076800.0,
                "minimum": 96059961344,
                "sample_count": 76
              },
              "process_pss_bytes": {
                "maximum": 1831052288,
                "mean": 1815476493.4736843,
                "median": 1813160960.0,
                "minimum": 1811305472,
                "sample_count": 76
              },
              "process_rss_bytes": {
                "maximum": 1834020864,
                "mean": 1818445069.4736843,
                "median": 1816129536.0,
                "minimum": 1814274048,
                "sample_count": 76
              }
            },
            "process_sample_count": 76,
            "response_union_seconds": 78.04574350797338,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 40960
              },
              "observed_endpoint_seconds": 73.0416899130214,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 78.07481005799491,
          "response_seconds": {
            "maximum": 15.836193532042671,
            "mean": 15.609148701594677,
            "median": 15.605332456005272,
            "minimum": 15.37207987497095,
            "sample_count": 5
          },
          "response_window_union_seconds": 78.04574350797338,
          "server_decode_tokens_per_second": {
            "maximum": 18.551195297162867,
            "mean": 18.2302957298202,
            "median": 18.20923135192617,
            "minimum": 17.94791051833425,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 1.6283667890238576,
            "mean": 1.6195974480011501,
            "median": 1.626210309972521,
            "minimum": 1.6014450550428592,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 11.832819705136346,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 3.121144745345183,
            "interval_count": 5,
            "interval_seconds": 108.19347628898686,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 99.87663185104586,
            "process_cpu_seconds": 108.06
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 32451.686339622163,
              "reason": null,
              "response_union_seconds": 108.17370938597014,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 299.9960574878009,
              "trapezoid_pieces": 113,
              "watt_hours": 9.014357316561712
            },
            "gpu_sample_count": 108,
            "host_sample_count": 104,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86949203968,
                "mean": 86934935552,
                "median": 86933780480.0,
                "minimum": 86925795328,
                "sample_count": 104
              },
              "gpu_board_watts": {
                "maximum": 300.63,
                "mean": 299.995,
                "median": 299.99,
                "minimum": 299.54,
                "sample_count": 108
              },
              "gpu_memory_used_bytes": {
                "maximum": 80000057344.0,
                "mean": 80000057344.0,
                "median": 80000057344.0,
                "minimum": 80000057344.0,
                "sample_count": 108
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1687.0,
                "mean": 1589.0185185185185,
                "median": 1635.0,
                "minimum": 1440.0,
                "sample_count": 108
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 99.99074074074075,
                "median": 100.0,
                "minimum": 99.0,
                "sample_count": 108
              },
              "host_mem_available_bytes": {
                "maximum": 96051195904,
                "mean": 96035076489.84616,
                "median": 96036397056.0,
                "minimum": 96015052800,
                "sample_count": 104
              },
              "process_pss_bytes": {
                "maximum": 1858872320,
                "mean": 1856665334.1538463,
                "median": 1857508352.0,
                "minimum": 1851196416,
                "sample_count": 104
              },
              "process_rss_bytes": {
                "maximum": 1861840896,
                "mean": 1859674820.9230769,
                "median": 1860476928.0,
                "minimum": 1854164992,
                "sample_count": 104
              }
            },
            "process_sample_count": 104,
            "response_union_seconds": 108.17370938597014,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 65536
              },
              "observed_endpoint_seconds": 101.77245614095591,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 108.20392717898358,
          "response_seconds": {
            "maximum": 21.827379069989547,
            "mean": 21.634741877194028,
            "median": 21.627266055962536,
            "minimum": 21.41630671202438,
            "sample_count": 5
          },
          "response_window_union_seconds": 108.17370938597014,
          "server_decode_tokens_per_second": {
            "maximum": 17.577270369850584,
            "mean": 17.411169160770672,
            "median": 17.422599078733956,
            "minimum": 17.266858141452268,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 7.059074364020489,
            "mean": 6.988261981203687,
            "median": 6.990940073970705,
            "minimum": 6.908797004027292,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "baseline-72b-q8",
      "model": "qwen25-72b-q8",
      "module": "baseline",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 1262.25
          },
          {
            "device": "CUDA0",
            "reported_mib": 72670.41
          }
        ],
        "gpu_layers": 81,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 81
      },
      "record_sha256": "d91f7a8a15841ca67a5d66247e9c695bdf3f8871bc97aaa5a1dee9f492d725a5",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84113866752,
              "mean": 84105189376,
              "median": 84105189376.0,
              "minimum": 84096512000,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.24,
              "mean": 16.225,
              "median": 16.225,
              "minimum": 16.21,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96514502656,
              "mean": 96499474432,
              "median": 96499474432.0,
              "minimum": 96484446208,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.100672836008016,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86535368704,
              "mean": 86535368704,
              "median": 86535368704,
              "minimum": 86535368704,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 114.18,
              "mean": 114.18,
              "median": 114.18,
              "minimum": 114.18,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2340.0,
              "mean": 2340.0,
              "median": 2340.0,
              "minimum": 2340.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96538132480,
              "mean": 96538132480,
              "median": 96538132480,
              "minimum": 96538132480,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1574024299625307,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 2275.973838036839,
            "reason": null,
            "response_union_seconds": 38.89415991096757,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 58.51711010719242,
            "trapezoid_pieces": 39,
            "watt_hours": 0.6322149550102331
          },
          "gpu_sample_count": 38,
          "host_sample_count": 30,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86970269696,
              "mean": 86913179511.46666,
              "median": 86960246784.0,
              "minimum": 85595869184,
              "sample_count": 30
            },
            "gpu_board_watts": {
              "maximum": 79.98,
              "mean": 58.390263157894736,
              "median": 71.515,
              "minimum": 16.13,
              "sample_count": 38
            },
            "gpu_memory_used_bytes": {
              "maximum": 76789317632.0,
              "mean": 20640390736.842106,
              "median": 587202560.0,
              "minimum": 587202560.0,
              "sample_count": 38
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 1737.578947368421,
              "median": 2272.0,
              "minimum": 180.0,
              "sample_count": 38
            },
            "gpu_utilization_percent": {
              "maximum": 81.0,
              "mean": 18.81578947368421,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 38
            },
            "host_mem_available_bytes": {
              "maximum": 96234205184,
              "mean": 96109763242.66667,
              "median": 96112695296.0,
              "minimum": 96030871552,
              "sample_count": 30
            },
            "process_pss_bytes": {
              "maximum": 77867508736,
              "mean": 42180037461.333336,
              "median": 39471920128.0,
              "minimum": 1633039360,
              "sample_count": 30
            },
            "process_rss_bytes": {
              "maximum": 77870477312,
              "mean": 42183054131.2,
              "median": 39474888704.0,
              "minimum": 1636007936,
              "sample_count": 30
            }
          },
          "process_sample_count": 30,
          "response_union_seconds": 38.89415991096757,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 5,
              "storage_read_bytes": 76230348800,
              "storage_write_bytes": 4096
            },
            "observed_endpoint_seconds": 37.21782839397201,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 569.13687963413,
            "reason": null,
            "response_union_seconds": 3.3318883720203303,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 170.81511025803874,
            "trapezoid_pieces": 8,
            "watt_hours": 0.1580935776761472
          },
          "gpu_sample_count": 5,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86946590720,
              "mean": 86934203050.66667,
              "median": 86936223744,
              "minimum": 86919794688,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 299.97,
              "mean": 158.414,
              "median": 106.19,
              "minimum": 79.8,
              "sample_count": 5
            },
            "gpu_memory_used_bytes": {
              "maximum": 80000057344.0,
              "mean": 79988313292.8,
              "median": 80000057344.0,
              "minimum": 79970697216.0,
              "sample_count": 5
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2340.0,
              "mean": 2173.2,
              "median": 2310.0,
              "minimum": 1627.0,
              "sample_count": 5
            },
            "gpu_utilization_percent": {
              "maximum": 98.0,
              "mean": 19.6,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 5
            },
            "host_mem_available_bytes": {
              "maximum": 96110792704,
              "mean": 96073691136,
              "median": 96070483968,
              "minimum": 96039796736,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 1858872320,
              "mean": 1824046762.6666667,
              "median": 1831121920,
              "minimum": 1782146048,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 1861840896,
              "mean": 1827015338.6666667,
              "median": 1834090496,
              "minimum": 1785114624,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.3318883720203303,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 55878.75495729765,
            "reason": null,
            "response_union_seconds": 186.29246559401508,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 299.95177088414056,
            "trapezoid_pieces": 188,
            "watt_hours": 15.521876377027125
          },
          "gpu_sample_count": 186,
          "host_sample_count": 181,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86966120448,
              "mean": 86939990570.43094,
              "median": 86937067520,
              "minimum": 86925795328,
              "sample_count": 181
            },
            "gpu_board_watts": {
              "maximum": 300.63,
              "mean": 299.99005376344087,
              "median": 299.99,
              "minimum": 299.2,
              "sample_count": 186
            },
            "gpu_memory_used_bytes": {
              "maximum": 80000057344.0,
              "mean": 80000057344.0,
              "median": 80000057344.0,
              "minimum": 80000057344.0,
              "sample_count": 186
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1740.0,
              "mean": 1627.5645161290322,
              "median": 1657.0,
              "minimum": 1440.0,
              "sample_count": 186
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 99.9247311827957,
              "median": 100.0,
              "minimum": 89.0,
              "sample_count": 186
            },
            "host_mem_available_bytes": {
              "maximum": 96114024448,
              "mean": 96054556117.56906,
              "median": 96044326912,
              "minimum": 96015052800,
              "sample_count": 181
            },
            "process_pss_bytes": {
              "maximum": 1858872320,
              "mean": 1839371450.6961327,
              "median": 1852769280,
              "minimum": 1811305472,
              "sample_count": 181
            },
            "process_rss_bytes": {
              "maximum": 1861840896,
              "mean": 1842363533.436464,
              "median": 1855737856,
              "minimum": 1814274048,
              "sample_count": 181
            }
          },
          "process_sample_count": 181,
          "response_union_seconds": 186.29246559401508,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 131072
            },
            "observed_endpoint_seconds": 184.07744828792056,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 10662.626895456593,
            "reason": null,
            "response_union_seconds": 36.43345087999478,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 292.6603612316979,
            "trapezoid_pieces": 38,
            "watt_hours": 2.961840804293498
          },
          "gpu_sample_count": 36,
          "host_sample_count": 36,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86964477952,
              "mean": 86945317546.66667,
              "median": 86945054720.0,
              "minimum": 86937935872,
              "sample_count": 36
            },
            "gpu_board_watts": {
              "maximum": 300.51,
              "mean": 295.0111111111111,
              "median": 299.995,
              "minimum": 188.62,
              "sample_count": 36
            },
            "gpu_memory_used_bytes": {
              "maximum": 80000057344.0,
              "mean": 80000057344.0,
              "median": 80000057344.0,
              "minimum": 80000057344.0,
              "sample_count": 36
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1762.0,
              "mean": 1664.611111111111,
              "median": 1680.0,
              "minimum": 1485.0,
              "sample_count": 36
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 100.0,
              "median": 100.0,
              "minimum": 100.0,
              "sample_count": 36
            },
            "host_mem_available_bytes": {
              "maximum": 96108740608,
              "mean": 96070393514.66667,
              "median": 96053764096.0,
              "minimum": 96042815488,
              "sample_count": 36
            },
            "process_pss_bytes": {
              "maximum": 1850520576,
              "mean": 1829862172.4444444,
              "median": 1835518976.0,
              "minimum": 1802523648,
              "sample_count": 36
            },
            "process_rss_bytes": {
              "maximum": 1853489152,
              "mean": 1832830748.4444444,
              "median": 1838487552.0,
              "minimum": 1805492224,
              "sample_count": 36
            }
          },
          "process_sample_count": 36,
          "response_union_seconds": 36.43345087999478,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 20480
            },
            "observed_endpoint_seconds": 34.95324771897867,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 268,
        "host_samples": 253,
        "maximum_gpu_gap_seconds": 1.0175007510115393,
        "maximum_host_gap_seconds": 2.0619814730016515,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 268,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 268,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 1352.721352025983,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 87432.48105413717,
          "reason": null,
          "response_union_seconds": 882.8749595810659,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 99.03155606046961,
          "trapezoid_pieces": 893,
          "watt_hours": 24.28680029281588
        },
        "gpu_sample_count": 883,
        "host_sample_count": 434,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 86956191744,
            "mean": 86912404852.79263,
            "median": 86905057280.0,
            "minimum": 86897274880,
            "sample_count": 434
          },
          "gpu_board_watts": {
            "maximum": 137.35,
            "mean": 99.02436013590034,
            "median": 95.14,
            "minimum": 88.97,
            "sample_count": 883
          },
          "gpu_memory_used_bytes": {
            "maximum": 94157930496.0,
            "mean": 94157089735.17554,
            "median": 94157930496.0,
            "minimum": 94155833344.0,
            "sample_count": 883
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2317.0,
            "mean": 2284.4246885617213,
            "median": 2280.0,
            "minimum": 1867.0,
            "sample_count": 883
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 34.90939977349943,
            "median": 10.0,
            "minimum": 2.0,
            "sample_count": 883
          },
          "host_mem_available_bytes": {
            "maximum": 95637037056,
            "mean": 95531510581.08755,
            "median": 95521652736.0,
            "minimum": 95466434560,
            "sample_count": 434
          },
          "process_pss_bytes": {
            "maximum": 55339509760,
            "mean": 55163167262.67281,
            "median": 55336781824.0,
            "minimum": 54836813824,
            "sample_count": 434
          },
          "process_rss_bytes": {
            "maximum": 55342489600,
            "mean": 55166208018.87558,
            "median": 55339761664.0,
            "minimum": 54839808000,
            "sample_count": 434
          }
        },
        "process_sample_count": 434,
        "response_union_seconds": 882.8749595810659,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 360448
          },
          "observed_endpoint_seconds": 859.7034017620608,
          "observed_intervals": 10,
          "requested_intervals": 10
        }
      },
      "groups": [
        {
          "aggregate_output_tokens_per_second": 3.6192757941330425,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 2048,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 38.159420998783055,
            "interval_count": 5,
            "interval_seconds": 353.6806808580877,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 1221.1014719610578,
            "process_cpu_seconds": 4318.8
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 33952.31959522913,
              "reason": null,
              "response_union_seconds": 353.66191271605203,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 96.00219411381389,
              "trapezoid_pieces": 359,
              "watt_hours": 9.431199887563647
            },
            "gpu_sample_count": 354,
            "host_sample_count": 167,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86956191744,
                "mean": 86928198306.49101,
                "median": 86925025280,
                "minimum": 86913802240,
                "sample_count": 167
              },
              "gpu_board_watts": {
                "maximum": 136.38,
                "mean": 95.9926836158192,
                "median": 94.245,
                "minimum": 88.97,
                "sample_count": 354
              },
              "gpu_memory_used_bytes": {
                "maximum": 94155833344.0,
                "mean": 94155833344.0,
                "median": 94155833344.0,
                "minimum": 94155833344.0,
                "sample_count": 354
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2317.0,
                "mean": 2284.135593220339,
                "median": 2280.0,
                "minimum": 1957.0,
                "sample_count": 354
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 23.166666666666668,
                "median": 10.0,
                "minimum": 2.0,
                "sample_count": 354
              },
              "host_mem_available_bytes": {
                "maximum": 95637037056,
                "mean": 95563106218.15569,
                "median": 95547498496,
                "minimum": 95503069184,
                "sample_count": 167
              },
              "process_pss_bytes": {
                "maximum": 54901213184,
                "mean": 54883534688.57485,
                "median": 54899740672,
                "minimum": 54836813824,
                "sample_count": 167
              },
              "process_rss_bytes": {
                "maximum": 54904193024,
                "mean": 54886587704.71856,
                "median": 54903541760,
                "minimum": 54839808000,
                "sample_count": 167
              }
            },
            "process_sample_count": 167,
            "response_union_seconds": 353.66191271605203,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 143360
              },
              "observed_endpoint_seconds": 341.29245696699945,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 353.68840590899345,
          "response_seconds": {
            "maximum": 70.85927938303212,
            "mean": 70.7323825432104,
            "median": 70.71976610599086,
            "minimum": 70.65188393200515,
            "sample_count": 5
          },
          "response_window_union_seconds": 353.66191271605203,
          "server_decode_tokens_per_second": {
            "maximum": 4.72037786569281,
            "mean": 4.705007937571641,
            "median": 4.704372092144592,
            "minimum": 4.692452607700804,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 16.539735411002766,
            "mean": 16.512870826828294,
            "median": 16.516706649039406,
            "minimum": 16.472258550056722,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        },
        {
          "aggregate_output_tokens_per_second": 2.4186856457574994,
          "completed_output_tokens": 1280,
          "completed_request_count": 5,
          "context_tokens": 8192,
          "cpu_snapshot_intervals": {
            "host_capacity_percent": 26.53628150643171,
            "interval_count": 5,
            "interval_seconds": 529.2353224620456,
            "meaning": "Weighted CPU counter rate across round-bracketing snapshots, not exact token time.",
            "one_logical_cpu_percent": 849.1610082058147,
            "process_cpu_seconds": 4494.06
          },
          "failed_request_count": 0,
          "measured_rounds": 5,
          "mode": "completion",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 53480.16145890802,
              "reason": null,
              "response_union_seconds": 529.2130468650139,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 101.05601472926116,
              "trapezoid_pieces": 534,
              "watt_hours": 14.855600405252227
            },
            "gpu_sample_count": 529,
            "host_sample_count": 267,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86921846784,
                "mean": 86902526550.29213,
                "median": 86901604352,
                "minimum": 86897274880,
                "sample_count": 267
              },
              "gpu_board_watts": {
                "maximum": 137.35,
                "mean": 101.0531190926276,
                "median": 95.67,
                "minimum": 90.98,
                "sample_count": 529
              },
              "gpu_memory_used_bytes": {
                "maximum": 94157930496.0,
                "mean": 94157930496.0,
                "median": 94157930496.0,
                "minimum": 94157930496.0,
                "sample_count": 529
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2317.0,
                "mean": 2284.618147448015,
                "median": 2280.0,
                "minimum": 1867.0,
                "sample_count": 529
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 42.76748582230624,
                "median": 10.0,
                "minimum": 2.0,
                "sample_count": 529
              },
              "host_mem_available_bytes": {
                "maximum": 95557570560,
                "mean": 95511748515.95506,
                "median": 95512559616,
                "minimum": 95466434560,
                "sample_count": 267
              },
              "process_pss_bytes": {
                "maximum": 55339509760,
                "mean": 55338068535.61049,
                "median": 55338133504,
                "minimum": 55335350272,
                "sample_count": 267
              },
              "process_rss_bytes": {
                "maximum": 55342489600,
                "mean": 55341101623.61049,
                "median": 55341113344,
                "minimum": 55339749376,
                "sample_count": 267
              }
            },
            "process_sample_count": 267,
            "response_union_seconds": 529.2130468650139,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 217088
              },
              "observed_endpoint_seconds": 518.4109447950614,
              "observed_intervals": 5,
              "requested_intervals": 5
            }
          },
          "response_outer_span_seconds": 529.246267170005,
          "response_seconds": {
            "maximum": 105.9342662889976,
            "mean": 105.84260937300277,
            "median": 105.85307992796879,
            "minimum": 105.70028573501622,
            "sample_count": 5
          },
          "response_window_union_seconds": 529.2130468650139,
          "server_decode_tokens_per_second": {
            "maximum": 4.717623238471859,
            "mean": 4.707775473637765,
            "median": 4.707951723075252,
            "minimum": 4.6967906663607675,
            "sample_count": 5
          },
          "synthetic_first_token_seconds": {
            "maximum": 51.79277144599473,
            "mean": 51.67663359480211,
            "median": 51.64751424099086,
            "minimum": 51.63793887099018,
            "sample_count": 5
          },
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "capacity-235b",
      "model": "qwen3-235b-q5",
      "module": "hybrid-capacity",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 27759.73
          },
          {
            "device": "CPU_Mapped",
            "reported_mib": 28323.86
          },
          {
            "device": "CPU_Mapped",
            "reported_mib": 18322.9
          },
          {
            "device": "CUDA0",
            "reported_mib": 86732.62
          }
        ],
        "gpu_layers": 95,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 44,
        "total_layers": 95
      },
      "record_sha256": "3a49e3e92a194f228457843e1a3264cd5a2af9e3e1302b9862ba46d0660bdd2a",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86569291776,
              "mean": 86560978944,
              "median": 86560978944.0,
              "minimum": 86552666112,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.06,
              "mean": 16.005,
              "median": 16.005,
              "minimum": 15.95,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96450252800,
              "mean": 96438243328,
              "median": 96438243328.0,
              "minimum": 96426233856,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.100754403974861,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86905700352,
              "mean": 86905700352,
              "median": 86905700352,
              "minimum": 86905700352,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 79.73,
              "mean": 79.73,
              "median": 79.73,
              "minimum": 79.73,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 94151639040.0,
              "mean": 94151639040.0,
              "median": 94151639040.0,
              "minimum": 94151639040.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2272.0,
              "mean": 2272.0,
              "median": 2272.0,
              "minimum": 2272.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 95525027840,
              "mean": 95525027840,
              "median": 95525027840,
              "minimum": 95525027840,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1650557749671862,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 5680.773122758429,
            "reason": null,
            "response_union_seconds": 116.2166981890332,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 48.8808683371672,
            "trapezoid_pieces": 117,
            "watt_hours": 1.5779925340995635
          },
          "gpu_sample_count": 116,
          "host_sample_count": 89,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86973087744,
              "mean": 86957874831.82022,
              "median": 86961295360,
              "minimum": 86779006976,
              "sample_count": 89
            },
            "gpu_board_watts": {
              "maximum": 77.03,
              "mean": 49.04491379310345,
              "median": 71.38,
              "minimum": 15.7,
              "sample_count": 116
            },
            "gpu_memory_used_bytes": {
              "maximum": 94088724480.0,
              "mean": 39070574521.37931,
              "median": 587202560.0,
              "minimum": 587202560.0,
              "sample_count": 116
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2280.0,
              "mean": 1387.0086206896551,
              "median": 2272.0,
              "minimum": 172.0,
              "sample_count": 116
            },
            "gpu_utilization_percent": {
              "maximum": 18.0,
              "mean": 4.905172413793103,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 116
            },
            "host_mem_available_bytes": {
              "maximum": 96184197120,
              "mean": 95922319406.02248,
              "median": 95875031040,
              "minimum": 95732076544,
              "sample_count": 89
            },
            "process_pss_bytes": {
              "maximum": 86750541824,
              "mean": 66394780545.4382,
              "median": 85652535296,
              "minimum": 339699712,
              "sample_count": 89
            },
            "process_rss_bytes": {
              "maximum": 86753521664,
              "mean": 66397802449.97753,
              "median": 85655515136,
              "minimum": 342679552,
              "sample_count": 89
            }
          },
          "process_sample_count": 89,
          "response_union_seconds": 116.2166981890332,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 56045,
              "read_syscall_bytes": 13,
              "storage_read_bytes": 265319071744,
              "storage_write_bytes": 12288
            },
            "observed_endpoint_seconds": 115.25819665094605,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 279.6970268935408,
            "reason": null,
            "response_union_seconds": 3.371804744063411,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 82.9517270790935,
            "trapezoid_pieces": 6,
            "watt_hours": 0.07769361858153911
          },
          "gpu_sample_count": 3,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86954340352,
              "mean": 86929561600,
              "median": 86929561600.0,
              "minimum": 86904782848,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 88.89,
              "mean": 83.37,
              "median": 84.96,
              "minimum": 76.26,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 94157930496.0,
              "mean": 94135560874.66667,
              "median": 94155833344.0,
              "minimum": 94092918784.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2272.0,
              "mean": 2272.0,
              "median": 2272.0,
              "minimum": 2272.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 95830491136,
              "mean": 95675967488,
              "median": 95675967488.0,
              "minimum": 95521443840,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 55339509760,
              "mean": 32317054976,
              "median": 32317054976.0,
              "minimum": 9294600192,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 55342489600,
              "mean": 32320034816,
              "median": 32320034816.0,
              "minimum": 9297580032,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 3.371804744063411,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 3
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 87439.59809728934,
            "reason": null,
            "response_union_seconds": 882.9507616019691,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 99.03111464408792,
            "trapezoid_pieces": 885,
            "watt_hours": 24.28877724924704
          },
          "gpu_sample_count": 883,
          "host_sample_count": 436,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86956191744,
              "mean": 86912462077.65138,
              "median": 86905081856.0,
              "minimum": 86897274880,
              "sample_count": 436
            },
            "gpu_board_watts": {
              "maximum": 137.35,
              "mean": 99.02436013590034,
              "median": 95.14,
              "minimum": 88.97,
              "sample_count": 883
            },
            "gpu_memory_used_bytes": {
              "maximum": 94157930496.0,
              "mean": 94157089735.17554,
              "median": 94157930496.0,
              "minimum": 94155833344.0,
              "sample_count": 883
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2317.0,
              "mean": 2284.4246885617213,
              "median": 2280.0,
              "minimum": 1867.0,
              "sample_count": 883
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 34.90939977349943,
              "median": 10.0,
              "minimum": 2.0,
              "sample_count": 883
            },
            "host_mem_available_bytes": {
              "maximum": 95637037056,
              "mean": 95531524809.98166,
              "median": 95521929216.0,
              "minimum": 95466434560,
              "sample_count": 436
            },
            "process_pss_bytes": {
              "maximum": 55339509760,
              "mean": 55161964248.073395,
              "median": 55336781824.0,
              "minimum": 54836813824,
              "sample_count": 436
            },
            "process_rss_bytes": {
              "maximum": 55342489600,
              "mean": 55165004724.84404,
              "median": 55339761664.0,
              "minimum": 54839808000,
              "sample_count": 436
            }
          },
          "process_sample_count": 436,
          "response_union_seconds": 882.9507616019691,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 385024
            },
            "observed_endpoint_seconds": 878.285594719986,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 18356.05087851068,
            "reason": null,
            "response_union_seconds": 192.7558839409612,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 95.22952297597782,
            "trapezoid_pieces": 195,
            "watt_hours": 5.098903021808522
          },
          "gpu_sample_count": 193,
          "host_sample_count": 104,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86972977152,
              "mean": 86935542350.76923,
              "median": 86941855744.0,
              "minimum": 86844063744,
              "sample_count": 104
            },
            "gpu_board_watts": {
              "maximum": 133.51,
              "mean": 95.20875647668393,
              "median": 93.83,
              "minimum": 74.26,
              "sample_count": 193
            },
            "gpu_memory_used_bytes": {
              "maximum": 94157930496.0,
              "mean": 94156430977.98964,
              "median": 94155833344.0,
              "minimum": 94155833344.0,
              "sample_count": 193
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2317.0,
              "mean": 2280.352331606218,
              "median": 2280.0,
              "minimum": 1942.0,
              "sample_count": 193
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 31.139896373056995,
              "median": 10.0,
              "minimum": 2.0,
              "sample_count": 193
            },
            "host_mem_available_bytes": {
              "maximum": 95645704192,
              "mean": 95576961142.15384,
              "median": 95548684288.0,
              "minimum": 95511699456,
              "sample_count": 104
            },
            "process_pss_bytes": {
              "maximum": 55336093696,
              "mean": 50895688615.38461,
              "median": 54901958656.0,
              "minimum": 11203971072,
              "sample_count": 104
            },
            "process_rss_bytes": {
              "maximum": 55339073536,
              "mean": 50898682210.46154,
              "median": 54904938496.0,
              "minimum": 11206950912,
              "sample_count": 104
            }
          },
          "process_sample_count": 104,
          "response_union_seconds": 192.7558839409612,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 5517,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 44982132736,
              "storage_write_bytes": 73728
            },
            "observed_endpoint_seconds": 190.65362677600933,
            "observed_intervals": 2,
            "requested_intervals": 2
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [
          2048,
          8192
        ],
        "endpoint": "/completion",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": true,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 5,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 256,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 44,
          "gpu_layers": "all"
        },
        "reasoning": "not_applicable_raw_completion",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 1198,
        "host_samples": 634,
        "maximum_gpu_gap_seconds": 1.4088406169903465,
        "maximum_host_gap_seconds": 2.6100394439999945,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 1198,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 1198,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 44.14226625201991,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 282.2856086120493,
          "reason": null,
          "response_union_seconds": 1.398443644924555,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 201.85697838919953,
          "trapezoid_pieces": 25,
          "watt_hours": 0.07841266905890258
        },
        "gpu_sample_count": 1,
        "host_sample_count": 0,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": null,
          "gpu_board_watts": {
            "maximum": 167.13,
            "mean": 167.13,
            "median": 167.13,
            "minimum": 167.13,
            "sample_count": 1
          },
          "gpu_memory_used_bytes": {
            "maximum": 6746537984.0,
            "mean": 6746537984.0,
            "median": 6746537984.0,
            "minimum": 6746537984.0,
            "sample_count": 1
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2167.0,
            "mean": 2167.0,
            "median": 2167.0,
            "minimum": 2167.0,
            "sample_count": 1
          },
          "gpu_utilization_percent": {
            "maximum": 89.0,
            "mean": 89.0,
            "median": 89.0,
            "minimum": 89.0,
            "sample_count": 1
          },
          "host_mem_available_bytes": null,
          "process_pss_bytes": null,
          "process_rss_bytes": null
        },
        "process_sample_count": 0,
        "response_union_seconds": 1.398443644924555,
        "sampled_counter_deltas": {
          "deltas": null,
          "observed_endpoint_seconds": 0.0,
          "observed_intervals": 0,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 229,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 282.2856086120493,
              "reason": null,
              "response_union_seconds": 1.398443644924555,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 201.85697838919953,
              "trapezoid_pieces": 25,
              "watt_hours": 0.07841266905890258
            },
            "gpu_sample_count": 1,
            "host_sample_count": 0,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": null,
              "gpu_board_watts": {
                "maximum": 167.13,
                "mean": 167.13,
                "median": 167.13,
                "minimum": 167.13,
                "sample_count": 1
              },
              "gpu_memory_used_bytes": {
                "maximum": 6746537984.0,
                "mean": 6746537984.0,
                "median": 6746537984.0,
                "minimum": 6746537984.0,
                "sample_count": 1
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2167.0,
                "mean": 2167.0,
                "median": 2167.0,
                "minimum": 2167.0,
                "sample_count": 1
              },
              "gpu_utilization_percent": {
                "maximum": 89.0,
                "mean": 89.0,
                "median": 89.0,
                "minimum": 89.0,
                "sample_count": 1
              },
              "host_mem_available_bytes": null,
              "process_pss_bytes": null,
              "process_rss_bytes": null
            },
            "process_sample_count": 0,
            "response_union_seconds": 1.398443644924555,
            "sampled_counter_deltas": {
              "deltas": null,
              "observed_endpoint_seconds": 0.0,
              "observed_intervals": 0,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 1.4396691279835068,
          "response_window_union_seconds": 1.398443644924555,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 2,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 3,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 8,
                "strict_format_pass": 12,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 9,
                "strict_format_pass": 12,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 17,
            "strict_format_pass": 24,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 0.05221215198980644,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.019419765972997993,
              "first_visible_content_seconds": 0.019419765972997993,
              "generated_tokens": 8,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 0.04756913398159668,
              "reasoning_events_observed": false,
              "response_seconds": 0.052119900996331125,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.06447923701489344,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018350900034420192,
              "first_visible_content_seconds": 0.018350900034420192,
              "generated_tokens": 11,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 0.059841028007213026,
              "reasoning_events_observed": false,
              "response_seconds": 0.0643794719944708,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.05317579198163003,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.02051437000045553,
              "first_visible_content_seconds": 0.02051437000045553,
              "generated_tokens": 8,
              "input_tokens": 93,
              "language": "pl",
              "last_visible_content_seconds": 0.04849577596178278,
              "reasoning_events_observed": false,
              "response_seconds": 0.05304892698768526,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.05432305403519422,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.017227592004928738,
              "first_visible_content_seconds": 0.017227592004928738,
              "generated_tokens": 9,
              "input_tokens": 49,
              "language": "en",
              "last_visible_content_seconds": 0.04970139602664858,
              "reasoning_events_observed": false,
              "response_seconds": 0.05423555203014985,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.06272390496451408,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.016606379009317607,
              "first_visible_content_seconds": 0.016606379009317607,
              "generated_tokens": 11,
              "input_tokens": 54,
              "language": "en",
              "last_visible_content_seconds": 0.05807481601368636,
              "reasoning_events_observed": false,
              "response_seconds": 0.06263354799011722,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04934750898974016,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.016784630017355084,
              "first_visible_content_seconds": 0.016784630017355084,
              "generated_tokens": 8,
              "input_tokens": 56,
              "language": "en",
              "last_visible_content_seconds": 0.04469241900369525,
              "reasoning_events_observed": false,
              "response_seconds": 0.049255488032940775,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.079617025970947,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.019807635981123894,
              "first_visible_content_seconds": 0.019807635981123894,
              "generated_tokens": 14,
              "input_tokens": 84,
              "language": "pl",
              "last_visible_content_seconds": 0.0749345850199461,
              "reasoning_events_observed": false,
              "response_seconds": 0.0795295630232431,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.09852313797455281,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.019785535987466574,
              "first_visible_content_seconds": 0.019785535987466574,
              "generated_tokens": 18,
              "input_tokens": 81,
              "language": "pl",
              "last_visible_content_seconds": 0.09379354899283499,
              "reasoning_events_observed": false,
              "response_seconds": 0.09843722794903442,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.08868215198162943,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018457377969752997,
              "first_visible_content_seconds": 0.018457377969752997,
              "generated_tokens": 16,
              "input_tokens": 72,
              "language": "pl",
              "last_visible_content_seconds": 0.08392347896005958,
              "reasoning_events_observed": false,
              "response_seconds": 0.08857695595361292,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.08371880900813267,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.017831003991886973,
              "first_visible_content_seconds": 0.017831003991886973,
              "generated_tokens": 15,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 0.07899835699936375,
              "reasoning_events_observed": false,
              "response_seconds": 0.08363295899471268,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.09630749101052061,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.016833781031891704,
              "first_visible_content_seconds": 0.016833781031891704,
              "generated_tokens": 18,
              "input_tokens": 51,
              "language": "en",
              "last_visible_content_seconds": 0.09156348503893241,
              "reasoning_events_observed": false,
              "response_seconds": 0.09622695099096745,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.07497315696673468,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018429475952871144,
              "first_visible_content_seconds": 0.018429475952871144,
              "generated_tokens": 13,
              "input_tokens": 53,
              "language": "en",
              "last_visible_content_seconds": 0.070229101984296,
              "reasoning_events_observed": false,
              "response_seconds": 0.07487369194859639,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04386618302669376,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0200749640353024,
              "first_visible_content_seconds": 0.0200749640353024,
              "generated_tokens": 6,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 0.039185795991215855,
              "reasoning_events_observed": false,
              "response_seconds": 0.04377738799666986,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04819632897851989,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0199568240204826,
              "first_visible_content_seconds": 0.0199568240204826,
              "generated_tokens": 7,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 0.043546427972614765,
              "reasoning_events_observed": false,
              "response_seconds": 0.04810755397193134,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04204844601918012,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018510867026634514,
              "first_visible_content_seconds": 0.018510867026634514,
              "generated_tokens": 6,
              "input_tokens": 76,
              "language": "pl",
              "last_visible_content_seconds": 0.03741702903062105,
              "reasoning_events_observed": false,
              "response_seconds": 0.04196078301174566,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04613240499747917,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018114059988874942,
              "first_visible_content_seconds": 0.018114059988874942,
              "generated_tokens": 7,
              "input_tokens": 65,
              "language": "en",
              "last_visible_content_seconds": 0.041501428990159184,
              "reasoning_events_observed": false,
              "response_seconds": 0.04604476201348007,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.0452238559955731,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.017137985967565328,
              "first_visible_content_seconds": 0.017137985967565328,
              "generated_tokens": 7,
              "input_tokens": 64,
              "language": "en",
              "last_visible_content_seconds": 0.040544210001826286,
              "reasoning_events_observed": false,
              "response_seconds": 0.04511930199805647,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.039076543005649,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.015550255950074643,
              "first_visible_content_seconds": 0.015550255950074643,
              "generated_tokens": 6,
              "input_tokens": 48,
              "language": "en",
              "last_visible_content_seconds": 0.03445477498462424,
              "reasoning_events_observed": false,
              "response_seconds": 0.03898570500314236,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.04645818000426516,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.018379073007963598,
              "first_visible_content_seconds": 0.018379073007963598,
              "generated_tokens": 7,
              "input_tokens": 78,
              "language": "pl",
              "last_visible_content_seconds": 0.04182571201818064,
              "reasoning_events_observed": false,
              "response_seconds": 0.04636744101298973,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.043439209985081106,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.01990467798896134,
              "first_visible_content_seconds": 0.01990467798896134,
              "generated_tokens": 6,
              "input_tokens": 89,
              "language": "pl",
              "last_visible_content_seconds": 0.03881491703214124,
              "reasoning_events_observed": false,
              "response_seconds": 0.04335429199272767,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.05091221700422466,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.02277130394941196,
              "first_visible_content_seconds": 0.02277130394941196,
              "generated_tokens": 7,
              "input_tokens": 108,
              "language": "pl",
              "last_visible_content_seconds": 0.04627511999569833,
              "reasoning_events_observed": false,
              "response_seconds": 0.050832788983825594,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.03985726402606815,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.016315378015860915,
              "first_visible_content_seconds": 0.016315378015860915,
              "generated_tokens": 6,
              "input_tokens": 52,
              "language": "en",
              "last_visible_content_seconds": 0.03522455500205979,
              "reasoning_events_observed": false,
              "response_seconds": 0.03976457205135375,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.045047869032714516,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.016996193036902696,
              "first_visible_content_seconds": 0.016996193036902696,
              "generated_tokens": 7,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 0.040395855030510575,
              "reasoning_events_observed": false,
              "response_seconds": 0.0449410310247913,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.05231968098087236,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.01978666801005602,
              "first_visible_content_seconds": 0.01978666801005602,
              "generated_tokens": 8,
              "input_tokens": 83,
              "language": "en",
              "last_visible_content_seconds": 0.04768890596460551,
              "reasoning_events_observed": false,
              "response_seconds": 0.05223778897197917,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-8b",
      "model": "qwen3-8b-q4",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 333.84
          },
          {
            "device": "CUDA0",
            "reported_mib": 4455.34
          }
        ],
        "gpu_layers": 37,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 37
      },
      "record_sha256": "61ec4830f18e473deb9b9939e4f9c53c40f203825628698eb0bb29e5107f2cc6",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6413602816,
              "mean": 6405283840,
              "median": 6405283840.0,
              "minimum": 6396964864,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.42,
              "mean": 16.255000000000003,
              "median": 16.255000000000003,
              "minimum": 16.09,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96553963520,
              "mean": 96552726528,
              "median": 96552726528.0,
              "minimum": 96551489536,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1003843810176477,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6400180224,
              "mean": 6400180224,
              "median": 6400180224,
              "minimum": 6400180224,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 76.32,
              "mean": 76.32,
              "median": 76.32,
              "minimum": 76.32,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2310.0,
              "median": 2310.0,
              "minimum": 2310.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96510169088,
              "mean": 96510169088,
              "median": 96510169088,
              "minimum": 96510169088,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1560979000059888,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 42.085601086970726,
            "reason": null,
            "response_union_seconds": 1.286838505009655,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 32.70464858110145,
            "trapezoid_pieces": 2,
            "watt_hours": 0.011690444746380757
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6616887296,
              "mean": 6616887296,
              "median": 6616887296,
              "minimum": 6616887296,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 37.93,
              "mean": 37.93,
              "median": 37.93,
              "minimum": 37.93,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 5259657216.0,
              "mean": 5259657216.0,
              "median": 5259657216.0,
              "minimum": 5259657216.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2302.0,
              "mean": 2302.0,
              "median": 2302.0,
              "minimum": 2302.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 79.0,
              "mean": 79.0,
              "median": 79.0,
              "minimum": 79.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96255836160,
              "mean": 96255836160,
              "median": 96255836160,
              "minimum": 96255836160,
              "sample_count": 1
            },
            "process_pss_bytes": {
              "maximum": 5367099392,
              "mean": 5367099392,
              "median": 5367099392,
              "minimum": 5367099392,
              "sample_count": 1
            },
            "process_rss_bytes": {
              "maximum": 5370085376,
              "mean": 5370085376,
              "median": 5370085376,
              "minimum": 5370085376,
              "sample_count": 1
            }
          },
          "process_sample_count": 1,
          "response_union_seconds": 1.286838505009655,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 300.5957116146231,
            "reason": null,
            "response_union_seconds": 2.213750633003656,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 135.78571458465018,
            "trapezoid_pieces": 5,
            "watt_hours": 0.08349880878183975
          },
          "gpu_sample_count": 3,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6727159808,
              "mean": 6709911552,
              "median": 6709911552.0,
              "minimum": 6692663296,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 293.55,
              "mean": 157.6,
              "median": 100.94,
              "minimum": 78.31,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 6746537984.0,
              "mean": 6742343680.0,
              "median": 6746537984.0,
              "minimum": 6733955072.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2267.3333333333335,
              "median": 2310.0,
              "minimum": 2182.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 81.0,
              "mean": 27.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96154648576,
              "mean": 96140855296,
              "median": 96140855296.0,
              "minimum": 96127062016,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 816381952,
              "mean": 799184896,
              "median": 799184896.0,
              "minimum": 781987840,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 819367936,
              "mean": 802170880,
              "median": 802170880.0,
              "minimum": 784973824,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 2.213750633003656,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 291.35640874325844,
            "reason": null,
            "response_union_seconds": 1.4412586120306514,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 202.1541493741736,
            "trapezoid_pieces": 2,
            "watt_hours": 0.08093233576201624
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 6722093056,
              "mean": 6722093056,
              "median": 6722093056,
              "minimum": 6722093056,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 167.13,
              "mean": 167.13,
              "median": 167.13,
              "minimum": 167.13,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 6746537984.0,
              "mean": 6746537984.0,
              "median": 6746537984.0,
              "minimum": 6746537984.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2167.0,
              "mean": 2167.0,
              "median": 2167.0,
              "minimum": 2167.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 89.0,
              "mean": 89.0,
              "median": 89.0,
              "minimum": 89.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96126963712,
              "mean": 96126963712,
              "median": 96126963712,
              "minimum": 96126963712,
              "sample_count": 1
            },
            "process_pss_bytes": {
              "maximum": 813629440,
              "mean": 813629440,
              "median": 813629440,
              "minimum": 813629440,
              "sample_count": 1
            },
            "process_rss_bytes": {
              "maximum": 816615424,
              "mean": 816615424,
              "median": 816615424,
              "minimum": 816615424,
              "sample_count": 1
            }
          },
          "process_sample_count": 1,
          "response_union_seconds": 1.4412586120306514,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 6.770582108400927,
            "reason": null,
            "response_union_seconds": 0.05492231895914301,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 123.27560519499544,
            "trapezoid_pieces": 1,
            "watt_hours": 0.0018807172523335907
          },
          "gpu_sample_count": 0,
          "host_sample_count": 0,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": null,
            "gpu_board_watts": null,
            "gpu_memory_used_bytes": null,
            "gpu_sm_clock_mhz": null,
            "gpu_utilization_percent": null,
            "host_mem_available_bytes": null,
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 0.05492231895914301,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "2d1a82644c4ba9ba504ddf8fd2c7fe3c253e15cc9811363c52186bd4b7510a07",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "disabled",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 8,
        "host_samples": 7,
        "maximum_gpu_gap_seconds": 1.004459120973479,
        "maximum_host_gap_seconds": 1.1064585979911499,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 8,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 8,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 193.62585014500655,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 1769.5657668781237,
          "reason": null,
          "response_union_seconds": 6.278294139250647,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 281.85454960052766,
          "trapezoid_pieces": 31,
          "watt_hours": 0.49154604635503435
        },
        "gpu_sample_count": 7,
        "host_sample_count": 5,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 17241067520,
            "mean": 17237679308.8,
            "median": 17237442560,
            "minimum": 17235619840,
            "sample_count": 5
          },
          "gpu_board_watts": {
            "maximum": 300.8,
            "mean": 276.22,
            "median": 300.01,
            "minimum": 133.73,
            "sample_count": 7
          },
          "gpu_memory_used_bytes": {
            "maximum": 22491955200.0,
            "mean": 22491056420.57143,
            "median": 22491955200.0,
            "minimum": 22485663744.0,
            "sample_count": 7
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2040.0,
            "mean": 1752.7142857142858,
            "median": 1710.0,
            "minimum": 1665.0,
            "sample_count": 7
          },
          "gpu_utilization_percent": {
            "maximum": 99.0,
            "mean": 96.0,
            "median": 98.0,
            "minimum": 89.0,
            "sample_count": 7
          },
          "host_mem_available_bytes": {
            "maximum": 96108433408,
            "mean": 96096174899.2,
            "median": 96094273536,
            "minimum": 96083144704,
            "sample_count": 5
          },
          "process_pss_bytes": {
            "maximum": 912734208,
            "mean": 909463142.4,
            "median": 909023232,
            "minimum": 907511808,
            "sample_count": 5
          },
          "process_rss_bytes": {
            "maximum": 915722240,
            "mean": 912451174.4,
            "median": 912011264,
            "minimum": 910499840,
            "sample_count": 5
          }
        },
        "process_sample_count": 5,
        "response_union_seconds": 6.278294139250647,
        "sampled_counter_deltas": {
          "deltas": null,
          "observed_endpoint_seconds": 0.0,
          "observed_intervals": 0,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 321,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 1769.5657668781237,
              "reason": null,
              "response_union_seconds": 6.278294139250647,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 281.85454960052766,
              "trapezoid_pieces": 31,
              "watt_hours": 0.49154604635503435
            },
            "gpu_sample_count": 7,
            "host_sample_count": 5,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 17241067520,
                "mean": 17237679308.8,
                "median": 17237442560,
                "minimum": 17235619840,
                "sample_count": 5
              },
              "gpu_board_watts": {
                "maximum": 300.8,
                "mean": 276.22,
                "median": 300.01,
                "minimum": 133.73,
                "sample_count": 7
              },
              "gpu_memory_used_bytes": {
                "maximum": 22491955200.0,
                "mean": 22491056420.57143,
                "median": 22491955200.0,
                "minimum": 22485663744.0,
                "sample_count": 7
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2040.0,
                "mean": 1752.7142857142858,
                "median": 1710.0,
                "minimum": 1665.0,
                "sample_count": 7
              },
              "gpu_utilization_percent": {
                "maximum": 99.0,
                "mean": 96.0,
                "median": 98.0,
                "minimum": 89.0,
                "sample_count": 7
              },
              "host_mem_available_bytes": {
                "maximum": 96108433408,
                "mean": 96096174899.2,
                "median": 96094273536,
                "minimum": 96083144704,
                "sample_count": 5
              },
              "process_pss_bytes": {
                "maximum": 912734208,
                "mean": 909463142.4,
                "median": 909023232,
                "minimum": 907511808,
                "sample_count": 5
              },
              "process_rss_bytes": {
                "maximum": 915722240,
                "mean": 912451174.4,
                "median": 912011264,
                "minimum": 910499840,
                "sample_count": 5
              }
            },
            "process_sample_count": 5,
            "response_union_seconds": 6.278294139250647,
            "sampled_counter_deltas": {
              "deltas": null,
              "observed_endpoint_seconds": 0.0,
              "observed_intervals": 0,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 6.326247751014307,
          "response_window_union_seconds": 6.278294139250647,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 1,
                "strict_format_pass": 2,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 3,
                "strict_format_pass": 3,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 1,
                "strict_format_pass": 1,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 2,
                "strict_format_pass": 2,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 4,
                "strict_format_pass": 4,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 3,
                "strict_format_pass": 4,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 7,
            "strict_format_pass": 8,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 0.18326423701364547,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.046741483034566045,
              "first_visible_content_seconds": 0.046741483034566045,
              "generated_tokens": 9,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 0.16585418704198673,
              "reasoning_events_observed": false,
              "response_seconds": 0.18315864104079083,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.289703368966002,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.048347275995183736,
              "first_visible_content_seconds": 0.048347275995183736,
              "generated_tokens": 15,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 0.2722942619584501,
              "reasoning_events_observed": false,
              "response_seconds": 0.28961531596723944,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.24483341298764572,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.05442306201439351,
              "first_visible_content_seconds": 0.05442306201439351,
              "generated_tokens": 12,
              "input_tokens": 93,
              "language": "pl",
              "last_visible_content_seconds": 0.2274235740187578,
              "reasoning_events_observed": false,
              "response_seconds": 0.2447356610209681,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.16171978804050013,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.041378010995686054,
              "first_visible_content_seconds": 0.041378010995686054,
              "generated_tokens": 8,
              "input_tokens": 49,
              "language": "en",
              "last_visible_content_seconds": 0.1442265540244989,
              "reasoning_events_observed": false,
              "response_seconds": 0.16162134503247216,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2838104229886085,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.041927040030714124,
              "first_visible_content_seconds": 0.041927040030714124,
              "generated_tokens": 15,
              "input_tokens": 54,
              "language": "en",
              "last_visible_content_seconds": 0.26636532903648913,
              "reasoning_events_observed": false,
              "response_seconds": 0.28372969303745776,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.23256938200211152,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.04222554503940046,
              "first_visible_content_seconds": 0.04222554503940046,
              "generated_tokens": 12,
              "input_tokens": 56,
              "language": "en",
              "last_visible_content_seconds": 0.21514906303491443,
              "reasoning_events_observed": false,
              "response_seconds": 0.2324729730025865,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4692047459539026,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.052966326009482145,
              "first_visible_content_seconds": 0.052966326009482145,
              "generated_tokens": 25,
              "input_tokens": 84,
              "language": "pl",
              "last_visible_content_seconds": 0.4516615499742329,
              "reasoning_events_observed": false,
              "response_seconds": 0.4690936499973759,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4865613059955649,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.052676697028800845,
              "first_visible_content_seconds": 0.052676697028800845,
              "generated_tokens": 26,
              "input_tokens": 81,
              "language": "pl",
              "last_visible_content_seconds": 0.4690213550347835,
              "reasoning_events_observed": false,
              "response_seconds": 0.4864655780256726,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.44700151396682486,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.047937755996827036,
              "first_visible_content_seconds": 0.047937755996827036,
              "generated_tokens": 24,
              "input_tokens": 72,
              "language": "pl",
              "last_visible_content_seconds": 0.4294581270078197,
              "reasoning_events_observed": false,
              "response_seconds": 0.4469077499816194,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4244083869853057,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.04277959401952103,
              "first_visible_content_seconds": 0.04277959401952103,
              "generated_tokens": 23,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 0.4068820509710349,
              "reasoning_events_observed": false,
              "response_seconds": 0.42432823800481856,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3364512249827385,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.041769918985664845,
              "first_visible_content_seconds": 0.041769918985664845,
              "generated_tokens": 18,
              "input_tokens": 51,
              "language": "en",
              "last_visible_content_seconds": 0.31892825599061325,
              "reasoning_events_observed": false,
              "response_seconds": 0.336373800993897,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3888603450031951,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.041705158015247434,
              "first_visible_content_seconds": 0.041705158015247434,
              "generated_tokens": 21,
              "input_tokens": 53,
              "language": "en",
              "last_visible_content_seconds": 0.3713369250181131,
              "reasoning_events_observed": false,
              "response_seconds": 0.3887764190440066,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.22667276801075786,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.05335977702634409,
              "first_visible_content_seconds": 0.05335977702634409,
              "generated_tokens": 11,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 0.20918457402149215,
              "reasoning_events_observed": false,
              "response_seconds": 0.22656903602182865,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.22677494899835438,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.053143836033996195,
              "first_visible_content_seconds": 0.053143836033996195,
              "generated_tokens": 11,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 0.20928829704644158,
              "reasoning_events_observed": false,
              "response_seconds": 0.22668170602992177,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.1342115300358273,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.048276664048898965,
              "first_visible_content_seconds": 0.048276664048898965,
              "generated_tokens": 6,
              "input_tokens": 76,
              "language": "pl",
              "last_visible_content_seconds": 0.11689828999806195,
              "reasoning_events_observed": false,
              "response_seconds": 0.13412892702035606,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.1513113440014422,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.048214048030786216,
              "first_visible_content_seconds": 0.048214048030786216,
              "generated_tokens": 7,
              "input_tokens": 65,
              "language": "en",
              "last_visible_content_seconds": 0.13389209704473615,
              "reasoning_events_observed": false,
              "response_seconds": 0.1512268070364371,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.14504845102783293,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.042269286001101136,
              "first_visible_content_seconds": 0.042269286001101136,
              "generated_tokens": 7,
              "input_tokens": 64,
              "language": "en",
              "last_visible_content_seconds": 0.1275256919907406,
              "reasoning_events_observed": false,
              "response_seconds": 0.14493916800711304,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.19332529499661177,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.03851492196554318,
              "first_visible_content_seconds": 0.03851492196554318,
              "generated_tokens": 10,
              "input_tokens": 48,
              "language": "en",
              "last_visible_content_seconds": 0.17569463601103052,
              "reasoning_events_observed": false,
              "response_seconds": 0.19323649996658787,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.15213412197772413,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.048757428012322634,
              "first_visible_content_seconds": 0.048757428012322634,
              "generated_tokens": 7,
              "input_tokens": 78,
              "language": "pl",
              "last_visible_content_seconds": 0.1347074909717776,
              "reasoning_events_observed": false,
              "response_seconds": 0.15204396401531994,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.14033973403275013,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.05410353804472834,
              "first_visible_content_seconds": 0.05410353804472834,
              "generated_tokens": 6,
              "input_tokens": 89,
              "language": "pl",
              "last_visible_content_seconds": 0.1229341720463708,
              "reasoning_events_observed": false,
              "response_seconds": 0.1402405590051785,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.23558119597146288,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.06252375902840868,
              "first_visible_content_seconds": 0.06252375902840868,
              "generated_tokens": 11,
              "input_tokens": 108,
              "language": "pl",
              "last_visible_content_seconds": 0.21811016299761832,
              "reasoning_events_observed": false,
              "response_seconds": 0.23548024898627773,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.21436406497377902,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.04161612200550735,
              "first_visible_content_seconds": 0.04161612200550735,
              "generated_tokens": 11,
              "input_tokens": 52,
              "language": "en",
              "last_visible_content_seconds": 0.19683912198524922,
              "reasoning_events_observed": false,
              "response_seconds": 0.21427161298925057,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.267502931994386,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.042441305005922914,
              "first_visible_content_seconds": 0.042441305005922914,
              "generated_tokens": 14,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 0.2500439510331489,
              "reasoning_events_observed": false,
              "response_seconds": 0.2674041580175981,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.24488870601635426,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.05344039702322334,
              "first_visible_content_seconds": 0.05344039702322334,
              "generated_tokens": 12,
              "input_tokens": 83,
              "language": "en",
              "last_visible_content_seconds": 0.22734448697883636,
              "reasoning_events_observed": false,
              "response_seconds": 0.24479238700587302,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-32b",
      "model": "qwen3-32b-q4",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 417.3
          },
          {
            "device": "CUDA0",
            "reported_mib": 18423.65
          }
        ],
        "gpu_layers": 65,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 65
      },
      "record_sha256": "b462b2b1fc8c26ff1d246dd6d67b75f4a5261c8180a70c606971cb7e2cc5e4b6",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16917901312,
              "mean": 16909461504,
              "median": 16909461504.0,
              "minimum": 16901021696,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.41,
              "mean": 16.384999999999998,
              "median": 16.384999999999998,
              "minimum": 16.36,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96522252288,
              "mean": 96519888896,
              "median": 96519888896.0,
              "minimum": 96517525504,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.100700625043828,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16904814592,
              "mean": 16904814592,
              "median": 16904814592,
              "minimum": 16904814592,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 79.02,
              "mean": 79.02,
              "median": 79.02,
              "minimum": 79.02,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2295.0,
              "mean": 2295.0,
              "median": 2295.0,
              "minimum": 2295.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96481214464,
              "mean": 96481214464,
              "median": 96481214464,
              "minimum": 96481214464,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1636213480378501,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 222.12262861428903,
            "reason": null,
            "response_union_seconds": 3.5050441779894754,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 63.372276449223115,
            "trapezoid_pieces": 4,
            "watt_hours": 0.06170073017063584
          },
          "gpu_sample_count": 3,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17206538240,
              "mean": 17169182720,
              "median": 17150586880,
              "minimum": 17150423040,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 79.28,
              "mean": 69.12333333333333,
              "median": 79.19,
              "minimum": 48.9,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 19906166784.0,
              "mean": 19906166784.0,
              "median": 19906166784.0,
              "minimum": 19906166784.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2300.0,
              "median": 2310.0,
              "minimum": 2280.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 80.0,
              "mean": 80.0,
              "median": 80.0,
              "minimum": 80.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96216006656,
              "mean": 96179727018.66667,
              "median": 96195715072,
              "minimum": 96127459328,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 20099635200,
              "mean": 13691974314.666666,
              "median": 20099520512,
              "minimum": 876767232,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 20102623232,
              "mean": 13694962346.666666,
              "median": 20102508544,
              "minimum": 879755264,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 3.5050441779894754,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 13,
              "storage_read_bytes": 0,
              "storage_write_bytes": 8192
            },
            "observed_endpoint_seconds": 2.2978108749957755,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 316.43031778887746,
            "reason": null,
            "response_union_seconds": 2.218777259986382,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 142.61472906515166,
            "trapezoid_pieces": 4,
            "watt_hours": 0.08789731049691041
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17241284608,
              "mean": 17223997440,
              "median": 17223997440.0,
              "minimum": 17206710272,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 137.52,
              "mean": 113.09,
              "median": 113.09,
              "minimum": 88.66,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 22491955200.0,
              "mean": 22484615168.0,
              "median": 22484615168.0,
              "minimum": 22477275136.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2302.5,
              "median": 2302.5,
              "minimum": 2295.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96120508416,
              "mean": 96110882816,
              "median": 96110882816.0,
              "minimum": 96101257216,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 912889856,
              "mean": 894881792,
              "median": 894881792.0,
              "minimum": 876873728,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 915877888,
              "mean": 897869824,
              "median": 897869824.0,
              "minimum": 879861760,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 2.218777259986382,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 1783.7209205120125,
            "reason": null,
            "response_union_seconds": 6.328095472999848,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 281.8732631520231,
            "trapezoid_pieces": 8,
            "watt_hours": 0.495478033475559
          },
          "gpu_sample_count": 7,
          "host_sample_count": 6,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17241067520,
              "mean": 17237492394.666668,
              "median": 17237000192.0,
              "minimum": 17235619840,
              "sample_count": 6
            },
            "gpu_board_watts": {
              "maximum": 300.8,
              "mean": 276.22,
              "median": 300.01,
              "minimum": 133.73,
              "sample_count": 7
            },
            "gpu_memory_used_bytes": {
              "maximum": 22491955200.0,
              "mean": 22491056420.57143,
              "median": 22491955200.0,
              "minimum": 22485663744.0,
              "sample_count": 7
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2040.0,
              "mean": 1752.7142857142858,
              "median": 1710.0,
              "minimum": 1665.0,
              "sample_count": 7
            },
            "gpu_utilization_percent": {
              "maximum": 99.0,
              "mean": 96.0,
              "median": 98.0,
              "minimum": 89.0,
              "sample_count": 7
            },
            "host_mem_available_bytes": {
              "maximum": 96108433408,
              "mean": 96095860053.33333,
              "median": 96094279680.0,
              "minimum": 96083144704,
              "sample_count": 6
            },
            "process_pss_bytes": {
              "maximum": 912734208,
              "mean": 909296981.3333334,
              "median": 908744704.0,
              "minimum": 907511808,
              "sample_count": 6
            },
            "process_rss_bytes": {
              "maximum": 915722240,
              "mean": 912285013.3333334,
              "median": 911732736.0,
              "minimum": 910499840,
              "sample_count": 6
            }
          },
          "process_sample_count": 6,
          "response_union_seconds": 6.328095472999848,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 45056
            },
            "observed_endpoint_seconds": 5.124694843951147,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 15.51256799570122,
            "reason": null,
            "response_union_seconds": 0.1260066389804706,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 123.10913235377596,
            "trapezoid_pieces": 1,
            "watt_hours": 0.004309046665472561
          },
          "gpu_sample_count": 0,
          "host_sample_count": 0,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": null,
            "gpu_board_watts": null,
            "gpu_memory_used_bytes": null,
            "gpu_sm_clock_mhz": null,
            "gpu_utilization_percent": null,
            "host_mem_available_bytes": null,
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 0.1260066389804706,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "59f09a9b924c4cb4ae1370f85109347542e72457c07f0a7a8ff351b31fedab64",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "disabled",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 15,
        "host_samples": 14,
        "maximum_gpu_gap_seconds": 1.0049945760401897,
        "maximum_host_gap_seconds": 1.2731039230129682,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 15,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 15,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 244.21277845196892,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 3356.7737874238665,
          "reason": null,
          "response_union_seconds": 11.435461274930276,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 293.5407419710171,
          "trapezoid_pieces": 35,
          "watt_hours": 0.9324371631732963
        },
        "gpu_sample_count": 11,
        "host_sample_count": 11,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 17320001536,
            "mean": 17315465029.81818,
            "median": 17315028992,
            "minimum": 17312980992,
            "sample_count": 11
          },
          "gpu_board_watts": {
            "maximum": 301.31,
            "mean": 295.28909090909093,
            "median": 300.03,
            "minimum": 246.92,
            "sample_count": 11
          },
          "gpu_memory_used_bytes": {
            "maximum": 47089451008.0,
            "mean": 47087163205.818184,
            "median": 47087353856.0,
            "minimum": 47083159552.0,
            "sample_count": 11
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1650.0,
            "mean": 1538.6363636363637,
            "median": 1567.0,
            "minimum": 1432.0,
            "sample_count": 11
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 98.36363636363636,
            "median": 100.0,
            "minimum": 92.0,
            "sample_count": 11
          },
          "host_mem_available_bytes": {
            "maximum": 96078073856,
            "mean": 96053993472,
            "median": 96049790976,
            "minimum": 96032055296,
            "sample_count": 11
          },
          "process_pss_bytes": {
            "maximum": 1201642496,
            "mean": 1198231552,
            "median": 1198070784,
            "minimum": 1195694080,
            "sample_count": 11
          },
          "process_rss_bytes": {
            "maximum": 1204600832,
            "mean": 1201318818.909091,
            "median": 1201029120,
            "minimum": 1199714304,
            "sample_count": 11
          }
        },
        "process_sample_count": 11,
        "response_union_seconds": 11.435461274930276,
        "sampled_counter_deltas": {
          "deltas": null,
          "observed_endpoint_seconds": 0.0,
          "observed_intervals": 0,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 268,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 3356.7737874238665,
              "reason": null,
              "response_union_seconds": 11.435461274930276,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 293.5407419710171,
              "trapezoid_pieces": 35,
              "watt_hours": 0.9324371631732963
            },
            "gpu_sample_count": 11,
            "host_sample_count": 11,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 17320001536,
                "mean": 17315465029.81818,
                "median": 17315028992,
                "minimum": 17312980992,
                "sample_count": 11
              },
              "gpu_board_watts": {
                "maximum": 301.31,
                "mean": 295.28909090909093,
                "median": 300.03,
                "minimum": 246.92,
                "sample_count": 11
              },
              "gpu_memory_used_bytes": {
                "maximum": 47089451008.0,
                "mean": 47087163205.818184,
                "median": 47087353856.0,
                "minimum": 47083159552.0,
                "sample_count": 11
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1650.0,
                "mean": 1538.6363636363637,
                "median": 1567.0,
                "minimum": 1432.0,
                "sample_count": 11
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 98.36363636363636,
                "median": 100.0,
                "minimum": 92.0,
                "sample_count": 11
              },
              "host_mem_available_bytes": {
                "maximum": 96078073856,
                "mean": 96053993472,
                "median": 96049790976,
                "minimum": 96032055296,
                "sample_count": 11
              },
              "process_pss_bytes": {
                "maximum": 1201642496,
                "mean": 1198231552,
                "median": 1198070784,
                "minimum": 1195694080,
                "sample_count": 11
              },
              "process_rss_bytes": {
                "maximum": 1204600832,
                "mean": 1201318818.909091,
                "median": 1201029120,
                "minimum": 1199714304,
                "sample_count": 11
              }
            },
            "process_sample_count": 11,
            "response_union_seconds": 11.435461274930276,
            "sampled_counter_deltas": {
              "deltas": null,
              "observed_endpoint_seconds": 0.0,
              "observed_intervals": 0,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 11.485896888014395,
          "response_window_union_seconds": 11.435461274930276,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 2,
                "strict_format_pass": 3,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 9,
                "strict_format_pass": 9,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 11,
                "strict_format_pass": 12,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 20,
            "strict_format_pass": 21,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 0.356745533994399,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10017254197737202,
              "first_visible_content_seconds": 0.10017254197737202,
              "generated_tokens": 8,
              "input_tokens": 86,
              "language": "pl",
              "last_visible_content_seconds": 0.32001081301132217,
              "reasoning_events_observed": false,
              "response_seconds": 0.3566570099792443,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4671982599538751,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10062494198791683,
              "first_visible_content_seconds": 0.10062494198791683,
              "generated_tokens": 11,
              "input_tokens": 86,
              "language": "pl",
              "last_visible_content_seconds": 0.43050856195623055,
              "reasoning_events_observed": false,
              "response_seconds": 0.46707107400288805,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.37288702599471435,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11485582502791658,
              "first_visible_content_seconds": 0.11485582502791658,
              "generated_tokens": 8,
              "input_tokens": 110,
              "language": "pl",
              "last_visible_content_seconds": 0.33612401102436706,
              "reasoning_events_observed": false,
              "response_seconds": 0.3727985810255632,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.34504318801919,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.08835643401835114,
              "first_visible_content_seconds": 0.08835643401835114,
              "generated_tokens": 8,
              "input_tokens": 66,
              "language": "en",
              "last_visible_content_seconds": 0.30829771602293476,
              "reasoning_events_observed": false,
              "response_seconds": 0.3449411480105482,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.45719915902009234,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09012828598497435,
              "first_visible_content_seconds": 0.09012828598497435,
              "generated_tokens": 11,
              "input_tokens": 71,
              "language": "en",
              "last_visible_content_seconds": 0.42045471898745745,
              "reasoning_events_observed": false,
              "response_seconds": 0.4570932820206508,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3471196349710226,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0897748590214178,
              "first_visible_content_seconds": 0.0897748590214178,
              "generated_tokens": 8,
              "input_tokens": 73,
              "language": "en",
              "last_visible_content_seconds": 0.3103491269866936,
              "reasoning_events_observed": false,
              "response_seconds": 0.3470321120112203,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.8486476499820128,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11331227701157331,
              "first_visible_content_seconds": 0.11331227701157331,
              "generated_tokens": 21,
              "input_tokens": 101,
              "language": "pl",
              "last_visible_content_seconds": 0.8117766549694352,
              "reasoning_events_observed": false,
              "response_seconds": 0.8485542259877548,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.8852201999980025,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11255090299528092,
              "first_visible_content_seconds": 0.11255090299528092,
              "generated_tokens": 22,
              "input_tokens": 98,
              "language": "pl",
              "last_visible_content_seconds": 0.8483489549835213,
              "reasoning_events_observed": false,
              "response_seconds": 0.8851136619923636,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.6540029040188529,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10272402100963518,
              "first_visible_content_seconds": 0.10272402100963518,
              "generated_tokens": 16,
              "input_tokens": 89,
              "language": "pl",
              "last_visible_content_seconds": 0.6171705310116522,
              "reasoning_events_observed": false,
              "response_seconds": 0.6539083890384063,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.8992188220145181,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09146723302546889,
              "first_visible_content_seconds": 0.09146723302546889,
              "generated_tokens": 23,
              "input_tokens": 78,
              "language": "en",
              "last_visible_content_seconds": 0.8623383800149895,
              "reasoning_events_observed": false,
              "response_seconds": 0.899088810023386,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.0103754760348238,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09159971901681274,
              "first_visible_content_seconds": 0.09159971901681274,
              "generated_tokens": 26,
              "input_tokens": 68,
              "language": "en",
              "last_visible_content_seconds": 0.973402061033994,
              "reasoning_events_observed": false,
              "response_seconds": 1.010288384044543,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.9362276829779148,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0899380529881455,
              "first_visible_content_seconds": 0.0899380529881455,
              "generated_tokens": 24,
              "input_tokens": 70,
              "language": "en",
              "last_visible_content_seconds": 0.8993266620091163,
              "reasoning_events_observed": false,
              "response_seconds": 0.936124931962695,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3364307919982821,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11384812399046496,
              "first_visible_content_seconds": 0.11384812399046496,
              "generated_tokens": 7,
              "input_tokens": 104,
              "language": "pl",
              "last_visible_content_seconds": 0.2992995039676316,
              "reasoning_events_observed": false,
              "response_seconds": 0.33630286395782605,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3360842969850637,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11398242198629305,
              "first_visible_content_seconds": 0.11398242198629305,
              "generated_tokens": 7,
              "input_tokens": 104,
              "language": "pl",
              "last_visible_content_seconds": 0.29910175496479496,
              "reasoning_events_observed": false,
              "response_seconds": 0.3359908139682375,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.28712529997574165,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10215065599186346,
              "first_visible_content_seconds": 0.10215065599186346,
              "generated_tokens": 6,
              "input_tokens": 93,
              "language": "pl",
              "last_visible_content_seconds": 0.2502421429962851,
              "reasoning_events_observed": false,
              "response_seconds": 0.28704002196900547,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.323599467985332,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10122723097447306,
              "first_visible_content_seconds": 0.10122723097447306,
              "generated_tokens": 7,
              "input_tokens": 82,
              "language": "en",
              "last_visible_content_seconds": 0.2865496819722466,
              "reasoning_events_observed": false,
              "response_seconds": 0.3235049929935485,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.32322676602052525,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10131020500557497,
              "first_visible_content_seconds": 0.10131020500557497,
              "generated_tokens": 7,
              "input_tokens": 81,
              "language": "en",
              "last_visible_content_seconds": 0.2861986099742353,
              "reasoning_events_observed": false,
              "response_seconds": 0.3231347450055182,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2745811709901318,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.08982771699083969,
              "first_visible_content_seconds": 0.08982771699083969,
              "generated_tokens": 6,
              "input_tokens": 65,
              "language": "en",
              "last_visible_content_seconds": 0.2376511769834906,
              "reasoning_events_observed": false,
              "response_seconds": 0.27448595402529463,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.32714697695337236,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10512047196971253,
              "first_visible_content_seconds": 0.10512047196971253,
              "generated_tokens": 7,
              "input_tokens": 95,
              "language": "pl",
              "last_visible_content_seconds": 0.2902187259751372,
              "reasoning_events_observed": false,
              "response_seconds": 0.3270393369602971,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.30081320501631126,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11499183799605817,
              "first_visible_content_seconds": 0.11499183799605817,
              "generated_tokens": 6,
              "input_tokens": 106,
              "language": "pl",
              "last_visible_content_seconds": 0.2638319249963388,
              "reasoning_events_observed": false,
              "response_seconds": 0.3007252609822899,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.34900118998484686,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.12620568595593795,
              "first_visible_content_seconds": 0.12620568595593795,
              "generated_tokens": 7,
              "input_tokens": 125,
              "language": "pl",
              "last_visible_content_seconds": 0.3117869569687173,
              "reasoning_events_observed": false,
              "response_seconds": 0.3489012139616534,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3118731380091049,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09012759401230142,
              "first_visible_content_seconds": 0.09012759401230142,
              "generated_tokens": 7,
              "input_tokens": 69,
              "language": "en",
              "last_visible_content_seconds": 0.2746786019997671,
              "reasoning_events_observed": false,
              "response_seconds": 0.31177029601531103,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.31387646798975766,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09208421799121425,
              "first_visible_content_seconds": 0.09208421799121425,
              "generated_tokens": 7,
              "input_tokens": 78,
              "language": "en",
              "last_visible_content_seconds": 0.2767789129866287,
              "reasoning_events_observed": false,
              "response_seconds": 0.3137474689865485,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3742343579651788,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11490003700600937,
              "first_visible_content_seconds": 0.11490003700600937,
              "generated_tokens": 8,
              "input_tokens": 100,
              "language": "en",
              "last_visible_content_seconds": 0.33724011498270556,
              "reasoning_events_observed": false,
              "response_seconds": 0.3741466960054822,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-72b-q4",
      "model": "qwen25-72b-q4",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 668.25
          },
          {
            "device": "CUDA0",
            "reported_mib": 41297.69
          }
        ],
        "gpu_layers": 81,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 81
      },
      "record_sha256": "d91a774e4a718e7f952298de4d84e2f57ba49fd61a5ac3333acd7b85eba581a9",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16963649536,
              "mean": 16954982400,
              "median": 16954982400.0,
              "minimum": 16946315264,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.24,
              "mean": 16.225,
              "median": 16.225,
              "minimum": 16.21,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96544727040,
              "mean": 96534085632,
              "median": 96534085632.0,
              "minimum": 96523444224,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1003493649768643,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 16957943808,
              "mean": 16957943808,
              "median": 16957943808,
              "minimum": 16957943808,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 83.89,
              "mean": 83.89,
              "median": 83.89,
              "minimum": 83.89,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2325.0,
              "median": 2325.0,
              "minimum": 2325.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96473059328,
              "mean": 96473059328,
              "median": 96473059328,
              "minimum": 96473059328,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.16017364297295,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 495.8568870733002,
            "reason": null,
            "response_union_seconds": 7.020026998012327,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 70.63461254688889,
            "trapezoid_pieces": 8,
            "watt_hours": 0.13773802418702782
          },
          "gpu_sample_count": 7,
          "host_sample_count": 5,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17242787840,
              "mean": 17226300620.8,
              "median": 17242234880,
              "minimum": 17171402752,
              "sample_count": 5
            },
            "gpu_board_watts": {
              "maximum": 78.93,
              "mean": 74.31,
              "median": 78.46,
              "minimum": 50.88,
              "sample_count": 7
            },
            "gpu_memory_used_bytes": {
              "maximum": 47066382336.0,
              "mean": 38158579419.42857,
              "median": 43891294208.0,
              "minimum": 587202560.0,
              "sample_count": 7
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2300.0,
              "median": 2302.0,
              "minimum": 2272.0,
              "sample_count": 7
            },
            "gpu_utilization_percent": {
              "maximum": 76.0,
              "mean": 54.285714285714285,
              "median": 75.0,
              "minimum": 0.0,
              "sample_count": 7
            },
            "host_mem_available_bytes": {
              "maximum": 96168050688,
              "mean": 96159906201.6,
              "median": 96164835328,
              "minimum": 96146849792,
              "sample_count": 5
            },
            "process_pss_bytes": {
              "maximum": 44346969088,
              "mean": 34967483801.6,
              "median": 44346682368,
              "minimum": 2059234304,
              "sample_count": 5
            },
            "process_rss_bytes": {
              "maximum": 44349927424,
              "mean": 34970442137.6,
              "median": 44349640704,
              "minimum": 2062192640,
              "sample_count": 5
            }
          },
          "process_sample_count": 5,
          "response_union_seconds": 7.020026998012327,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 5,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 5.75842740497319,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 330.0755432089877,
            "reason": null,
            "response_union_seconds": 2.230047432007268,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 148.0127904328413,
            "trapezoid_pieces": 5,
            "watt_hours": 0.09168765089138547
          },
          "gpu_sample_count": 3,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17320370176,
              "mean": 17298167808,
              "median": 17298167808.0,
              "minimum": 17275965440,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 298.46,
              "mean": 170.81666666666666,
              "median": 117.85,
              "minimum": 96.14,
              "sample_count": 3
            },
            "gpu_memory_used_bytes": {
              "maximum": 47089451008.0,
              "mean": 47083858602.666664,
              "median": 47089451008.0,
              "minimum": 47072673792.0,
              "sample_count": 3
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2069.6666666666665,
              "median": 2272.0,
              "minimum": 1612.0,
              "sample_count": 3
            },
            "gpu_utilization_percent": {
              "maximum": 83.0,
              "mean": 27.666666666666668,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 3
            },
            "host_mem_available_bytes": {
              "maximum": 96100491264,
              "mean": 96082976768,
              "median": 96082976768.0,
              "minimum": 96065462272,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 1201933312,
              "mean": 1180298240,
              "median": 1180298240.0,
              "minimum": 1158663168,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 1204891648,
              "mean": 1183256576,
              "median": 1183256576.0,
              "minimum": 1161621504,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 2.230047432007268,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 3372.111928686947,
            "reason": null,
            "response_union_seconds": 11.487630557036027,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 293.5428600305719,
            "trapezoid_pieces": 12,
            "watt_hours": 0.9366977579685964
          },
          "gpu_sample_count": 11,
          "host_sample_count": 11,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 17320001536,
              "mean": 17315465029.81818,
              "median": 17315028992,
              "minimum": 17312980992,
              "sample_count": 11
            },
            "gpu_board_watts": {
              "maximum": 301.31,
              "mean": 295.28909090909093,
              "median": 300.03,
              "minimum": 246.92,
              "sample_count": 11
            },
            "gpu_memory_used_bytes": {
              "maximum": 47089451008.0,
              "mean": 47087163205.818184,
              "median": 47087353856.0,
              "minimum": 47083159552.0,
              "sample_count": 11
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1650.0,
              "mean": 1538.6363636363637,
              "median": 1567.0,
              "minimum": 1432.0,
              "sample_count": 11
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 98.36363636363636,
              "median": 100.0,
              "minimum": 92.0,
              "sample_count": 11
            },
            "host_mem_available_bytes": {
              "maximum": 96078073856,
              "mean": 96053993472,
              "median": 96049790976,
              "minimum": 96032055296,
              "sample_count": 11
            },
            "process_pss_bytes": {
              "maximum": 1201642496,
              "mean": 1198231552,
              "median": 1198070784,
              "minimum": 1195694080,
              "sample_count": 11
            },
            "process_rss_bytes": {
              "maximum": 1204600832,
              "mean": 1201318818.909091,
              "median": 1201029120,
              "minimum": 1199714304,
              "sample_count": 11
            }
          },
          "process_sample_count": 11,
          "response_union_seconds": 11.487630557036027,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 53248
            },
            "observed_endpoint_seconds": 10.264362932997756,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 37.868706308886466,
            "reason": null,
            "response_union_seconds": 0.2522864409838803,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 150.10202752555404,
            "trapezoid_pieces": 1,
            "watt_hours": 0.010519085085801795
          },
          "gpu_sample_count": 0,
          "host_sample_count": 0,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": null,
            "gpu_board_watts": null,
            "gpu_memory_used_bytes": null,
            "gpu_sm_clock_mhz": null,
            "gpu_utilization_percent": null,
            "host_mem_available_bytes": null,
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 0.2522864409838803,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "59f09a9b924c4cb4ae1370f85109347542e72457c07f0a7a8ff351b31fedab64",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "model_default",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 24,
        "host_samples": 21,
        "maximum_gpu_gap_seconds": 1.010699467966333,
        "maximum_host_gap_seconds": 1.5830580749898218,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 24,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 24,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 122.35882354300702,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 1890.8853818060957,
          "reason": null,
          "response_union_seconds": 6.825519807287492,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 277.0317038399375,
          "trapezoid_pieces": 31,
          "watt_hours": 0.5252459393905822
        },
        "gpu_sample_count": 7,
        "host_sample_count": 7,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 84697124864,
            "mean": 84681880137.14285,
            "median": 84676489216,
            "minimum": 84673835008,
            "sample_count": 7
          },
          "gpu_board_watts": {
            "maximum": 296.82,
            "mean": 271.7057142857143,
            "median": 295.27,
            "minimum": 130.98,
            "sample_count": 7
          },
          "gpu_memory_used_bytes": {
            "maximum": 63946358784.0,
            "mean": 63942164480.0,
            "median": 63942164480.0,
            "minimum": 63933775872.0,
            "sample_count": 7
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2220.0,
            "mean": 2176.0,
            "median": 2175.0,
            "minimum": 2160.0,
            "sample_count": 7
          },
          "gpu_utilization_percent": {
            "maximum": 96.0,
            "mean": 95.42857142857143,
            "median": 96.0,
            "minimum": 94.0,
            "sample_count": 7
          },
          "host_mem_available_bytes": {
            "maximum": 95883984896,
            "mean": 95873145709.71428,
            "median": 95873601536,
            "minimum": 95859146752,
            "sample_count": 7
          },
          "process_pss_bytes": {
            "maximum": 1369285632,
            "mean": 1357484178.2857144,
            "median": 1353266176,
            "minimum": 1351747584,
            "sample_count": 7
          },
          "process_rss_bytes": {
            "maximum": 1372254208,
            "mean": 1360452461.7142856,
            "median": 1356234752,
            "minimum": 1354715136,
            "sample_count": 7
          }
        },
        "process_sample_count": 7,
        "response_union_seconds": 6.825519807287492,
        "sampled_counter_deltas": {
          "deltas": null,
          "observed_endpoint_seconds": 0.0,
          "observed_intervals": 0,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 1115,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 1890.8853818060957,
              "reason": null,
              "response_union_seconds": 6.825519807287492,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 277.0317038399375,
              "trapezoid_pieces": 31,
              "watt_hours": 0.5252459393905822
            },
            "gpu_sample_count": 7,
            "host_sample_count": 7,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 84697124864,
                "mean": 84681880137.14285,
                "median": 84676489216,
                "minimum": 84673835008,
                "sample_count": 7
              },
              "gpu_board_watts": {
                "maximum": 296.82,
                "mean": 271.7057142857143,
                "median": 295.27,
                "minimum": 130.98,
                "sample_count": 7
              },
              "gpu_memory_used_bytes": {
                "maximum": 63946358784.0,
                "mean": 63942164480.0,
                "median": 63942164480.0,
                "minimum": 63933775872.0,
                "sample_count": 7
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2220.0,
                "mean": 2176.0,
                "median": 2175.0,
                "minimum": 2160.0,
                "sample_count": 7
              },
              "gpu_utilization_percent": {
                "maximum": 96.0,
                "mean": 95.42857142857143,
                "median": 96.0,
                "minimum": 94.0,
                "sample_count": 7
              },
              "host_mem_available_bytes": {
                "maximum": 95883984896,
                "mean": 95873145709.71428,
                "median": 95873601536,
                "minimum": 95859146752,
                "sample_count": 7
              },
              "process_pss_bytes": {
                "maximum": 1369285632,
                "mean": 1357484178.2857144,
                "median": 1353266176,
                "minimum": 1351747584,
                "sample_count": 7
              },
              "process_rss_bytes": {
                "maximum": 1372254208,
                "mean": 1360452461.7142856,
                "median": 1356234752,
                "minimum": 1354715136,
                "sample_count": 7
              }
            },
            "process_sample_count": 7,
            "response_union_seconds": 6.825519807287492,
            "sampled_counter_deltas": {
              "deltas": null,
              "observed_endpoint_seconds": 0.0,
              "observed_intervals": 0,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 6.887551121006254,
          "response_window_union_seconds": 6.825519807287492,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 5,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 11,
                "strict_format_pass": 12,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 12,
                "strict_format_pass": 12,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 23,
            "strict_format_pass": 24,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 0.27955167298205197,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09705409401794896,
              "first_visible_content_seconds": 0.2577958600013517,
              "generated_tokens": 46,
              "input_tokens": 118,
              "language": "pl",
              "last_visible_content_seconds": 0.27508944901637733,
              "reasoning_events_observed": true,
              "response_seconds": 0.2794649610295892,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4165507630095817,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09704593900823966,
              "first_visible_content_seconds": 0.3862377070472576,
              "generated_tokens": 78,
              "input_tokens": 120,
              "language": "pl",
              "last_visible_content_seconds": 0.41210553201381117,
              "reasoning_events_observed": true,
              "response_seconds": 0.4164607660495676,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.5300433269585483,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11108733399305493,
              "first_visible_content_seconds": 0.508374436001759,
              "generated_tokens": 101,
              "input_tokens": 144,
              "language": "pl",
              "last_visible_content_seconds": 0.5256182629964314,
              "reasoning_events_observed": true,
              "response_seconds": 0.529949311981909,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2759973059874028,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09455439797602594,
              "first_visible_content_seconds": 0.2542955539538525,
              "generated_tokens": 46,
              "input_tokens": 102,
              "language": "en",
              "last_visible_content_seconds": 0.2715504619991407,
              "reasoning_events_observed": true,
              "response_seconds": 0.27591482299612835,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.39382727100746706,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0955572510138154,
              "first_visible_content_seconds": 0.36350169201614335,
              "generated_tokens": 73,
              "input_tokens": 105,
              "language": "en",
              "last_visible_content_seconds": 0.38937099900795147,
              "reasoning_events_observed": true,
              "response_seconds": 0.3937421730370261,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.5334408920025453,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09711036900989711,
              "first_visible_content_seconds": 0.5117335090180859,
              "generated_tokens": 105,
              "input_tokens": 111,
              "language": "en",
              "last_visible_content_seconds": 0.5289953900501132,
              "reasoning_events_observed": true,
              "response_seconds": 0.533345805015415,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3152286679833196,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10764644999289885,
              "first_visible_content_seconds": 0.2546262390096672,
              "generated_tokens": 52,
              "input_tokens": 138,
              "language": "pl",
              "last_visible_content_seconds": 0.3107349060010165,
              "reasoning_events_observed": true,
              "response_seconds": 0.31513324996922165,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3926023049862124,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10755501996027306,
              "first_visible_content_seconds": 0.34502433199668303,
              "generated_tokens": 70,
              "input_tokens": 134,
              "language": "pl",
              "last_visible_content_seconds": 0.3881439589895308,
              "reasoning_events_observed": true,
              "response_seconds": 0.39250517496839166,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2182195580098778,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10574875300517306,
              "first_visible_content_seconds": 0.17494896001880988,
              "generated_tokens": 30,
              "input_tokens": 132,
              "language": "pl",
              "last_visible_content_seconds": 0.21378613798879087,
              "reasoning_events_observed": true,
              "response_seconds": 0.21813491004286334,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.31083601398859173,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09474109497386962,
              "first_visible_content_seconds": 0.263139721006155,
              "generated_tokens": 54,
              "input_tokens": 110,
              "language": "en",
              "last_visible_content_seconds": 0.30632333701942116,
              "reasoning_events_observed": true,
              "response_seconds": 0.31073152902536094,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2401642320328392,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09318748803343624,
              "first_visible_content_seconds": 0.19256407802458853,
              "generated_tokens": 38,
              "input_tokens": 103,
              "language": "en",
              "last_visible_content_seconds": 0.23571856901980937,
              "reasoning_events_observed": true,
              "response_seconds": 0.24006823304807767,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.23364904301706702,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0953480820171535,
              "first_visible_content_seconds": 0.19038780202390626,
              "generated_tokens": 36,
              "input_tokens": 108,
              "language": "en",
              "last_visible_content_seconds": 0.22922682401258498,
              "reasoning_events_observed": true,
              "response_seconds": 0.23356682999292389,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.28579313203226775,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10856451600557193,
              "first_visible_content_seconds": 0.2640955669921823,
              "generated_tokens": 45,
              "input_tokens": 133,
              "language": "pl",
              "last_visible_content_seconds": 0.28135421202750877,
              "reasoning_events_observed": true,
              "response_seconds": 0.2857054690248333,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.19931219902355224,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10838362004142255,
              "first_visible_content_seconds": 0.17758144304389134,
              "generated_tokens": 25,
              "input_tokens": 133,
              "language": "pl",
              "last_visible_content_seconds": 0.1948704740498215,
              "reasoning_events_observed": true,
              "response_seconds": 0.1992240950348787,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.19309462304227054,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10218470805557445,
              "first_visible_content_seconds": 0.17135369800962508,
              "generated_tokens": 25,
              "input_tokens": 126,
              "language": "pl",
              "last_visible_content_seconds": 0.18863363203126937,
              "reasoning_events_observed": true,
              "response_seconds": 0.1930002270382829,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3030121630290523,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09548865398392081,
              "first_visible_content_seconds": 0.2813191270106472,
              "generated_tokens": 52,
              "input_tokens": 110,
              "language": "en",
              "last_visible_content_seconds": 0.29857497598277405,
              "reasoning_events_observed": true,
              "response_seconds": 0.30292886798270047,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.18670700205257162,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09583107102662325,
              "first_visible_content_seconds": 0.16501961601898074,
              "generated_tokens": 25,
              "input_tokens": 110,
              "language": "en",
              "last_visible_content_seconds": 0.1822839620290324,
              "reasoning_events_observed": true,
              "response_seconds": 0.18662376701831818,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.17943409498548135,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09285126300528646,
              "first_visible_content_seconds": 0.15771433897316456,
              "generated_tokens": 24,
              "input_tokens": 103,
              "language": "en",
              "last_visible_content_seconds": 0.1749823020072654,
              "reasoning_events_observed": true,
              "response_seconds": 0.17932811798527837,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.21239918103674427,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10815410304348916,
              "first_visible_content_seconds": 0.19072690303437412,
              "generated_tokens": 28,
              "input_tokens": 137,
              "language": "pl",
              "last_visible_content_seconds": 0.20794740703422576,
              "reasoning_events_observed": true,
              "response_seconds": 0.21230887301499024,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.22544415399897844,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10852008295478299,
              "first_visible_content_seconds": 0.20372335700085387,
              "generated_tokens": 31,
              "input_tokens": 142,
              "language": "pl",
              "last_visible_content_seconds": 0.22101183695485815,
              "reasoning_events_observed": true,
              "response_seconds": 0.2253553689806722,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.2214502920396626,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11320191400591284,
              "first_visible_content_seconds": 0.1997798079974018,
              "generated_tokens": 29,
              "input_tokens": 153,
              "language": "pl",
              "last_visible_content_seconds": 0.21700421901186928,
              "reasoning_events_observed": true,
              "response_seconds": 0.22136740799760446,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.258762295008637,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09460269700502977,
              "first_visible_content_seconds": 0.23707950802054256,
              "generated_tokens": 42,
              "input_tokens": 107,
              "language": "en",
              "last_visible_content_seconds": 0.25433214200893417,
              "reasoning_events_observed": true,
              "response_seconds": 0.2586870560189709,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.21704611700261012,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0960083199897781,
              "first_visible_content_seconds": 0.18673405202571303,
              "generated_tokens": 32,
              "input_tokens": 116,
              "language": "en",
              "last_visible_content_seconds": 0.21261256700381637,
              "reasoning_events_observed": true,
              "response_seconds": 0.21696273202542216,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.20509582699742168,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.101186954008881,
              "first_visible_content_seconds": 0.17902238003443927,
              "generated_tokens": 28,
              "input_tokens": 129,
              "language": "en",
              "last_visible_content_seconds": 0.20065375202102587,
              "reasoning_events_observed": true,
              "response_seconds": 0.2050100580090657,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-gpt120",
      "model": "gpt-oss-120b-mxfp4",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 586.82
          },
          {
            "device": "CUDA0",
            "reported_mib": 59851.68
          }
        ],
        "gpu_layers": 37,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 37
      },
      "record_sha256": "f0d0de57ed5676c038b2de188e9af7391ee892fd664158ca0e5f2cf8c440ec33",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84082905088,
              "mean": 84074186752,
              "median": 84074186752.0,
              "minimum": 84065468416,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.27,
              "mean": 16.21,
              "median": 16.21,
              "minimum": 16.15,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96528797696,
              "mean": 96522641408,
              "median": 96522641408.0,
              "minimum": 96516485120,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1008607699768618,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84094517248,
              "mean": 84094517248,
              "median": 84094517248,
              "minimum": 84094517248,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 80.87,
              "mean": 80.87,
              "median": 80.87,
              "minimum": 80.87,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2325.0,
              "median": 2325.0,
              "minimum": 2325.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 62.0,
              "mean": 62.0,
              "median": 62.0,
              "minimum": 62.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96561479680,
              "mean": 96561479680,
              "median": 96561479680,
              "minimum": 96561479680,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1581742509733886,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 747.1318841848013,
            "reason": null,
            "response_union_seconds": 10.409931548987515,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 71.77106599298132,
            "trapezoid_pieces": 11,
            "watt_hours": 0.20753663449577814
          },
          "gpu_sample_count": 10,
          "host_sample_count": 6,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84477476864,
              "mean": 84459245568,
              "median": 84476350464.0,
              "minimum": 84401012736,
              "sample_count": 6
            },
            "gpu_board_watts": {
              "maximum": 78.61,
              "mean": 74.135,
              "median": 78.39,
              "minimum": 43.17,
              "sample_count": 10
            },
            "gpu_memory_used_bytes": {
              "maximum": 63346573312.0,
              "mean": 57070636236.8,
              "median": 63346573312.0,
              "minimum": 587202560.0,
              "sample_count": 10
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2303.2,
              "median": 2310.0,
              "minimum": 2272.0,
              "sample_count": 10
            },
            "gpu_utilization_percent": {
              "maximum": 82.0,
              "mean": 64.9,
              "median": 80.5,
              "minimum": 0.0,
              "sample_count": 10
            },
            "host_mem_available_bytes": {
              "maximum": 96115118080,
              "mean": 96061430442.66667,
              "median": 96042070016.0,
              "minimum": 96027320320,
              "sample_count": 6
            },
            "process_pss_bytes": {
              "maximum": 63779092480,
              "mean": 53334008490.666664,
              "median": 63778757632.0,
              "minimum": 1110169600,
              "sample_count": 6
            },
            "process_rss_bytes": {
              "maximum": 63782060032,
              "mean": 53336976725.333336,
              "median": 63781726208.0,
              "minimum": 1113137152,
              "sample_count": 6
            }
          },
          "process_sample_count": 6,
          "response_union_seconds": 10.409931548987515,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 13,
              "storage_read_bytes": 0,
              "storage_write_bytes": 8192
            },
            "observed_endpoint_seconds": 8.570322451996617,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 314.0380837256312,
            "reason": null,
            "response_union_seconds": 2.214553338009864,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 141.80651164981444,
            "trapezoid_pieces": 4,
            "watt_hours": 0.08723280103489756
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84695801856,
              "mean": 84571525120,
              "median": 84571525120.0,
              "minimum": 84447248384,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 247.3,
              "mean": 161.58,
              "median": 161.58,
              "minimum": 75.86,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 63946358784.0,
              "mean": 63914901504.0,
              "median": 63914901504.0,
              "minimum": 63883444224.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2276.0,
              "median": 2276.0,
              "minimum": 2242.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96095813632,
              "mean": 95979352064,
              "median": 95979352064.0,
              "minimum": 95862890496,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 1369285632,
              "mean": 1239780864,
              "median": 1239780864.0,
              "minimum": 1110276096,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 1372254208,
              "mean": 1242748928,
              "median": 1242748928.0,
              "minimum": 1113243648,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 2.214553338009864,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 1908.9087139509015,
            "reason": null,
            "response_union_seconds": 6.889701746986248,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 277.0669593623429,
            "trapezoid_pieces": 8,
            "watt_hours": 0.5302524205419171
          },
          "gpu_sample_count": 7,
          "host_sample_count": 7,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 84697124864,
              "mean": 84681880137.14285,
              "median": 84676489216,
              "minimum": 84673835008,
              "sample_count": 7
            },
            "gpu_board_watts": {
              "maximum": 296.82,
              "mean": 271.7057142857143,
              "median": 295.27,
              "minimum": 130.98,
              "sample_count": 7
            },
            "gpu_memory_used_bytes": {
              "maximum": 63946358784.0,
              "mean": 63942164480.0,
              "median": 63942164480.0,
              "minimum": 63933775872.0,
              "sample_count": 7
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2220.0,
              "mean": 2176.0,
              "median": 2175.0,
              "minimum": 2160.0,
              "sample_count": 7
            },
            "gpu_utilization_percent": {
              "maximum": 96.0,
              "mean": 95.42857142857143,
              "median": 96.0,
              "minimum": 94.0,
              "sample_count": 7
            },
            "host_mem_available_bytes": {
              "maximum": 95883984896,
              "mean": 95873145709.71428,
              "median": 95873601536,
              "minimum": 95859146752,
              "sample_count": 7
            },
            "process_pss_bytes": {
              "maximum": 1369285632,
              "mean": 1357484178.2857144,
              "median": 1353266176,
              "minimum": 1351747584,
              "sample_count": 7
            },
            "process_rss_bytes": {
              "maximum": 1372254208,
              "mean": 1360452461.7142856,
              "median": 1356234752,
              "minimum": 1354715136,
              "sample_count": 7
            }
          },
          "process_sample_count": 7,
          "response_union_seconds": 6.889701746986248,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 69632
            },
            "observed_endpoint_seconds": 6.151284010964446,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 25.06160632528632,
            "reason": null,
            "response_union_seconds": 0.217665508040227,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 115.13816107536275,
            "trapezoid_pieces": 1,
            "watt_hours": 0.006961557312579534
          },
          "gpu_sample_count": 0,
          "host_sample_count": 0,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": null,
            "gpu_board_watts": null,
            "gpu_memory_used_bytes": null,
            "gpu_sm_clock_mhz": null,
            "gpu_utilization_percent": null,
            "host_mem_available_bytes": null,
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 0.217665508040227,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "low",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 22,
        "host_samples": 18,
        "maximum_gpu_gap_seconds": 1.0091279879561625,
        "maximum_host_gap_seconds": 1.8987281449954025,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 22,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 22,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 334.61070701800054,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 4465.851881307789,
          "reason": null,
          "response_union_seconds": 15.147101391979959,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 294.8321111570795,
          "trapezoid_pieces": 40,
          "watt_hours": 1.2405144114743858
        },
        "gpu_sample_count": 16,
        "host_sample_count": 15,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 86901219328,
            "mean": 86897635054.93333,
            "median": 86897152000,
            "minimum": 86896111616,
            "sample_count": 15
          },
          "gpu_board_watts": {
            "maximum": 301.31,
            "mean": 292.749375,
            "median": 299.985,
            "minimum": 185.35,
            "sample_count": 16
          },
          "gpu_memory_used_bytes": {
            "maximum": 79985377280.0,
            "mean": 79983411200.0,
            "median": 79983280128.0,
            "minimum": 79981182976.0,
            "sample_count": 16
          },
          "gpu_sm_clock_mhz": {
            "maximum": 1912.0,
            "mean": 1761.875,
            "median": 1755.0,
            "minimum": 1725.0,
            "sample_count": 16
          },
          "gpu_utilization_percent": {
            "maximum": 100.0,
            "mean": 97.0625,
            "median": 99.0,
            "minimum": 89.0,
            "sample_count": 16
          },
          "host_mem_available_bytes": {
            "maximum": 96044224512,
            "mean": 96037115494.4,
            "median": 96037216256,
            "minimum": 96029982720,
            "sample_count": 15
          },
          "process_pss_bytes": {
            "maximum": 1807780864,
            "mean": 1805991731.2,
            "median": 1805884416,
            "minimum": 1804573696,
            "sample_count": 15
          },
          "process_rss_bytes": {
            "maximum": 1810751488,
            "mean": 1808962355.2,
            "median": 1808855040,
            "minimum": 1807544320,
            "sample_count": 15
          }
        },
        "process_sample_count": 15,
        "response_union_seconds": 15.147101391979959,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 0,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 0,
            "storage_write_bytes": 0
          },
          "observed_endpoint_seconds": 1.0237997599761002,
          "observed_intervals": 1,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 268,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 4465.851881307789,
              "reason": null,
              "response_union_seconds": 15.147101391979959,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 294.8321111570795,
              "trapezoid_pieces": 40,
              "watt_hours": 1.2405144114743858
            },
            "gpu_sample_count": 16,
            "host_sample_count": 15,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86901219328,
                "mean": 86897635054.93333,
                "median": 86897152000,
                "minimum": 86896111616,
                "sample_count": 15
              },
              "gpu_board_watts": {
                "maximum": 301.31,
                "mean": 292.749375,
                "median": 299.985,
                "minimum": 185.35,
                "sample_count": 16
              },
              "gpu_memory_used_bytes": {
                "maximum": 79985377280.0,
                "mean": 79983411200.0,
                "median": 79983280128.0,
                "minimum": 79981182976.0,
                "sample_count": 16
              },
              "gpu_sm_clock_mhz": {
                "maximum": 1912.0,
                "mean": 1761.875,
                "median": 1755.0,
                "minimum": 1725.0,
                "sample_count": 16
              },
              "gpu_utilization_percent": {
                "maximum": 100.0,
                "mean": 97.0625,
                "median": 99.0,
                "minimum": 89.0,
                "sample_count": 16
              },
              "host_mem_available_bytes": {
                "maximum": 96044224512,
                "mean": 96037115494.4,
                "median": 96037216256,
                "minimum": 96029982720,
                "sample_count": 15
              },
              "process_pss_bytes": {
                "maximum": 1807780864,
                "mean": 1805991731.2,
                "median": 1805884416,
                "minimum": 1804573696,
                "sample_count": 15
              },
              "process_rss_bytes": {
                "maximum": 1810751488,
                "mean": 1808962355.2,
                "median": 1808855040,
                "minimum": 1807544320,
                "sample_count": 15
              }
            },
            "process_sample_count": 15,
            "response_union_seconds": 15.147101391979959,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 0,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 0,
                "storage_write_bytes": 0
              },
              "observed_endpoint_seconds": 1.0237997599761002,
              "observed_intervals": 1,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 15.199260192981455,
          "response_window_union_seconds": 15.147101391979959,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 2,
                "strict_format_pass": 3,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 9,
                "strict_format_pass": 9,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 11,
                "strict_format_pass": 12,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 20,
            "strict_format_pass": 21,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 0.4624123440007679,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.1003733419929631,
              "first_visible_content_seconds": 0.1003733419929631,
              "generated_tokens": 8,
              "input_tokens": 86,
              "language": "pl",
              "last_visible_content_seconds": 0.41084933996899053,
              "reasoning_events_observed": false,
              "response_seconds": 0.4623165460070595,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.6178736919537187,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10087333095725626,
              "first_visible_content_seconds": 0.10087333095725626,
              "generated_tokens": 11,
              "input_tokens": 86,
              "language": "pl",
              "last_visible_content_seconds": 0.5661676619783975,
              "reasoning_events_observed": false,
              "response_seconds": 0.6177548510022461,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.47930059902137145,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11358177597867325,
              "first_visible_content_seconds": 0.11358177597867325,
              "generated_tokens": 8,
              "input_tokens": 110,
              "language": "pl",
              "last_visible_content_seconds": 0.4245128500042483,
              "reasoning_events_observed": false,
              "response_seconds": 0.4761067419894971,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.453706119034905,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09103933000005782,
              "first_visible_content_seconds": 0.09103933000005782,
              "generated_tokens": 8,
              "input_tokens": 66,
              "language": "en",
              "last_visible_content_seconds": 0.4019894190132618,
              "reasoning_events_observed": false,
              "response_seconds": 0.4536073360359296,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.6101116340141743,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09264816099312156,
              "first_visible_content_seconds": 0.09264816099312156,
              "generated_tokens": 11,
              "input_tokens": 71,
              "language": "en",
              "last_visible_content_seconds": 0.5583812679979019,
              "reasoning_events_observed": false,
              "response_seconds": 0.6100081210024655,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4552373259793967,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09299267100868747,
              "first_visible_content_seconds": 0.09299267100868747,
              "generated_tokens": 8,
              "input_tokens": 73,
              "language": "en",
              "last_visible_content_seconds": 0.40348986798198894,
              "reasoning_events_observed": false,
              "response_seconds": 0.45513054699404165,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.1484679460409097,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11286671704147011,
              "first_visible_content_seconds": 0.11286671704147011,
              "generated_tokens": 21,
              "input_tokens": 101,
              "language": "pl",
              "last_visible_content_seconds": 1.096593552036211,
              "reasoning_events_observed": false,
              "response_seconds": 1.1483607460395433,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.197766219964251,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11130067595513538,
              "first_visible_content_seconds": 0.11130067595513538,
              "generated_tokens": 22,
              "input_tokens": 98,
              "language": "pl",
              "last_visible_content_seconds": 1.145948311954271,
              "reasoning_events_observed": false,
              "response_seconds": 1.1976569370017387,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.8809563079848886,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10292497597401962,
              "first_visible_content_seconds": 0.10292497597401962,
              "generated_tokens": 16,
              "input_tokens": 89,
              "language": "pl",
              "last_visible_content_seconds": 0.8289367749821395,
              "reasoning_events_observed": false,
              "response_seconds": 0.8808293719775975,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.2341569290147163,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.0936925420537591,
              "first_visible_content_seconds": 0.0936925420537591,
              "generated_tokens": 23,
              "input_tokens": 78,
              "language": "en",
              "last_visible_content_seconds": 1.1822401970275678,
              "reasoning_events_observed": false,
              "response_seconds": 1.2340589670347981,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.3918048930354416,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09384725004201755,
              "first_visible_content_seconds": 0.09384725004201755,
              "generated_tokens": 26,
              "input_tokens": 68,
              "language": "en",
              "last_visible_content_seconds": 1.3398392700473778,
              "reasoning_events_observed": false,
              "response_seconds": 1.391702413035091,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 1.2889309329912066,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09309692500391975,
              "first_visible_content_seconds": 0.09309692500391975,
              "generated_tokens": 24,
              "input_tokens": 70,
              "language": "en",
              "last_visible_content_seconds": 1.2368159629986621,
              "reasoning_events_observed": false,
              "response_seconds": 1.2888320589554496,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4270828170119785,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11355894402367994,
              "first_visible_content_seconds": 0.11355894402367994,
              "generated_tokens": 7,
              "input_tokens": 104,
              "language": "pl",
              "last_visible_content_seconds": 0.3750819690176286,
              "reasoning_events_observed": false,
              "response_seconds": 0.42699231900041923,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4275304999900982,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11382147198310122,
              "first_visible_content_seconds": 0.11382147198310122,
              "generated_tokens": 7,
              "input_tokens": 104,
              "language": "pl",
              "last_visible_content_seconds": 0.37546139396727085,
              "reasoning_events_observed": false,
              "response_seconds": 0.4274168179836124,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3644571070326492,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10282866598572582,
              "first_visible_content_seconds": 0.10282866598572582,
              "generated_tokens": 6,
              "input_tokens": 93,
              "language": "pl",
              "last_visible_content_seconds": 0.3122624789830297,
              "reasoning_events_observed": false,
              "response_seconds": 0.3643620199873112,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.41477912495611235,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10162348597077653,
              "first_visible_content_seconds": 0.10162348597077653,
              "generated_tokens": 7,
              "input_tokens": 82,
              "language": "en",
              "last_visible_content_seconds": 0.3627258089836687,
              "reasoning_events_observed": false,
              "response_seconds": 0.41469444800168276,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4143401890178211,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10151976300403476,
              "first_visible_content_seconds": 0.10151976300403476,
              "generated_tokens": 7,
              "input_tokens": 81,
              "language": "en",
              "last_visible_content_seconds": 0.3623804870294407,
              "reasoning_events_observed": false,
              "response_seconds": 0.4142360750120133,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.3534971229964867,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09264861198607832,
              "first_visible_content_seconds": 0.09264861198607832,
              "generated_tokens": 6,
              "input_tokens": 65,
              "language": "en",
              "last_visible_content_seconds": 0.3013498419895768,
              "reasoning_events_observed": false,
              "response_seconds": 0.3534003730164841,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4183506650151685,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.10539302398683503,
              "first_visible_content_seconds": 0.10539302398683503,
              "generated_tokens": 7,
              "input_tokens": 95,
              "language": "pl",
              "last_visible_content_seconds": 0.3663132779765874,
              "reasoning_events_observed": false,
              "response_seconds": 0.41825576801784337,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.37726473598740995,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11510742199607193,
              "first_visible_content_seconds": 0.11510742199607193,
              "generated_tokens": 6,
              "input_tokens": 106,
              "language": "pl",
              "last_visible_content_seconds": 0.3250673629809171,
              "reasoning_events_observed": false,
              "response_seconds": 0.37717469898052514,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4391804459737614,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.12564693397143856,
              "first_visible_content_seconds": 0.12564693397143856,
              "generated_tokens": 7,
              "input_tokens": 125,
              "language": "pl",
              "last_visible_content_seconds": 0.3870974849560298,
              "reasoning_events_observed": false,
              "response_seconds": 0.43910245096776634,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.40666094300104305,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09354595001786947,
              "first_visible_content_seconds": 0.09354595001786947,
              "generated_tokens": 7,
              "input_tokens": 69,
              "language": "en",
              "last_visible_content_seconds": 0.35457959602354094,
              "reasoning_events_observed": false,
              "response_seconds": 0.40656067698728293,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4079053469467908,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.09445987897925079,
              "first_visible_content_seconds": 0.09445987897925079,
              "generated_tokens": 7,
              "input_tokens": 78,
              "language": "en",
              "last_visible_content_seconds": 0.35581853898474947,
              "reasoning_events_observed": false,
              "response_seconds": 0.40781516896095127,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 0.4808241510181688,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 0.11479741497896612,
              "first_visible_content_seconds": 0.11479741497896612,
              "generated_tokens": 8,
              "input_tokens": 100,
              "language": "en",
              "last_visible_content_seconds": 0.4286229700082913,
              "reasoning_events_observed": false,
              "response_seconds": 0.4807259379886091,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-72b-q8",
      "model": "qwen25-72b-q8",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 1262.25
          },
          {
            "device": "CUDA0",
            "reported_mib": 72670.41
          }
        ],
        "gpu_layers": 81,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 0,
        "total_layers": 81
      },
      "record_sha256": "2e2d3711b58b2eed8dd42766e8441ed2285fd45bf7b5bc37044eb2edc06cb834",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86562975744,
              "mean": 86554726400,
              "median": 86554726400.0,
              "minimum": 86546477056,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.1,
              "mean": 16.065,
              "median": 16.065,
              "minimum": 16.03,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96535875584,
              "mean": 96524048384,
              "median": 96524048384.0,
              "minimum": 96512221184,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1007932139909826,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86551470080,
              "mean": 86551470080,
              "median": 86551470080,
              "minimum": 86551470080,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 90.5,
              "mean": 90.5,
              "median": 90.5,
              "minimum": 90.5,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2332.0,
              "mean": 2332.0,
              "median": 2332.0,
              "minimum": 2332.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 96491347968,
              "mean": 96491347968,
              "median": 96491347968,
              "minimum": 96491347968,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1605551209650002,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 907.0206510687725,
            "reason": null,
            "response_union_seconds": 12.308446155046113,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 73.69091432365083,
            "trapezoid_pieces": 13,
            "watt_hours": 0.2519501808524368
          },
          "gpu_sample_count": 12,
          "host_sample_count": 6,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86911909888,
              "mean": 86897436672,
              "median": 86911610880.0,
              "minimum": 86827036672,
              "sample_count": 6
            },
            "gpu_board_watts": {
              "maximum": 79.04,
              "mean": 75.555,
              "median": 78.85,
              "minimum": 47.35,
              "sample_count": 12
            },
            "gpu_memory_used_bytes": {
              "maximum": 76789317632.0,
              "mean": 70439141376.0,
              "median": 76789317632.0,
              "minimum": 587202560.0,
              "sample_count": 12
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2310.0,
              "mean": 2303.6666666666665,
              "median": 2310.0,
              "minimum": 2272.0,
              "sample_count": 12
            },
            "gpu_utilization_percent": {
              "maximum": 78.0,
              "mean": 64.16666666666667,
              "median": 77.0,
              "minimum": 0.0,
              "sample_count": 12
            },
            "host_mem_available_bytes": {
              "maximum": 96154288128,
              "mean": 96082595157.33333,
              "median": 96070154240.0,
              "minimum": 96058236928,
              "sample_count": 6
            },
            "process_pss_bytes": {
              "maximum": 77867432960,
              "mean": 70829067264,
              "median": 77867031552.0,
              "minimum": 35638844416,
              "sample_count": 6
            },
            "process_rss_bytes": {
              "maximum": 77870403584,
              "mean": 70832037888,
              "median": 77870002176.0,
              "minimum": 35641815040,
              "sample_count": 6
            }
          },
          "process_sample_count": 6,
          "response_union_seconds": 12.308446155046113,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 5,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 10.228211377980188,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 320.98164345760625,
            "reason": null,
            "response_union_seconds": 2.213284064026084,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 145.02505515434075,
            "trapezoid_pieces": 4,
            "watt_hours": 0.08916156762711284
          },
          "gpu_sample_count": 2,
          "host_sample_count": 3,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86901002240,
              "mean": 86884474880,
              "median": 86876241920,
              "minimum": 86876180480,
              "sample_count": 3
            },
            "gpu_board_watts": {
              "maximum": 105.28,
              "mean": 100.3,
              "median": 100.3,
              "minimum": 95.32,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 79985377280.0,
              "mean": 79978037248.0,
              "median": 79978037248.0,
              "minimum": 79970697216.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2332.0,
              "mean": 2321.0,
              "median": 2321.0,
              "minimum": 2310.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96050864128,
              "mean": 96043317930.66667,
              "median": 96040165376,
              "minimum": 96038924288,
              "sample_count": 3
            },
            "process_pss_bytes": {
              "maximum": 1807780864,
              "mean": 1790665045.3333333,
              "median": 1782107136,
              "minimum": 1782107136,
              "sample_count": 3
            },
            "process_rss_bytes": {
              "maximum": 1810751488,
              "mean": 1793635669.3333333,
              "median": 1785077760,
              "minimum": 1785077760,
              "sample_count": 3
            }
          },
          "process_sample_count": 3,
          "response_union_seconds": 2.213284064026084,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 1.0560765829868615,
            "observed_intervals": 1,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 4481.917668225537,
            "reason": null,
            "response_union_seconds": 15.201420327997766,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 294.83545428783407,
            "trapezoid_pieces": 17,
            "watt_hours": 1.2449771300626493
          },
          "gpu_sample_count": 16,
          "host_sample_count": 15,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86901219328,
              "mean": 86897635054.93333,
              "median": 86897152000,
              "minimum": 86896111616,
              "sample_count": 15
            },
            "gpu_board_watts": {
              "maximum": 301.31,
              "mean": 292.749375,
              "median": 299.985,
              "minimum": 185.35,
              "sample_count": 16
            },
            "gpu_memory_used_bytes": {
              "maximum": 79985377280.0,
              "mean": 79983411200.0,
              "median": 79983280128.0,
              "minimum": 79981182976.0,
              "sample_count": 16
            },
            "gpu_sm_clock_mhz": {
              "maximum": 1912.0,
              "mean": 1761.875,
              "median": 1755.0,
              "minimum": 1725.0,
              "sample_count": 16
            },
            "gpu_utilization_percent": {
              "maximum": 100.0,
              "mean": 97.0625,
              "median": 99.0,
              "minimum": 89.0,
              "sample_count": 16
            },
            "host_mem_available_bytes": {
              "maximum": 96044224512,
              "mean": 96037115494.4,
              "median": 96037216256,
              "minimum": 96029982720,
              "sample_count": 15
            },
            "process_pss_bytes": {
              "maximum": 1807780864,
              "mean": 1805991731.2,
              "median": 1805884416,
              "minimum": 1804573696,
              "sample_count": 15
            },
            "process_rss_bytes": {
              "maximum": 1810751488,
              "mean": 1808962355.2,
              "median": 1808855040,
              "minimum": 1807544320,
              "sample_count": 15
            }
          },
          "process_sample_count": 15,
          "response_union_seconds": 15.201420327997766,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 0,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 0,
              "storage_write_bytes": 61440
            },
            "observed_endpoint_seconds": 14.409904278989416,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 53.988835613617184,
            "reason": null,
            "response_union_seconds": 0.34023551997961476,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 158.68077388525433,
            "trapezoid_pieces": 1,
            "watt_hours": 0.014996898781560328
          },
          "gpu_sample_count": 0,
          "host_sample_count": 0,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": null,
            "gpu_board_watts": null,
            "gpu_memory_used_bytes": null,
            "gpu_sm_clock_mhz": null,
            "gpu_utilization_percent": null,
            "host_mem_available_bytes": null,
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 0.34023551997961476,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 0,
          "gpu_layers": "all"
        },
        "reasoning": "model_default",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 33,
        "host_samples": 27,
        "maximum_gpu_gap_seconds": 1.010140597994905,
        "maximum_host_gap_seconds": 2.0882626120001078,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 33,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 33,
            "unavailable": 0
          }
        }
      }
    },
    {
      "case_wall_seconds": 519.5668174049933,
      "completed_response_resources": {
        "gpu_energy": {
          "available": true,
          "interpolation": "linear_at_bracketed_boundaries",
          "joules": 19300.75655985953,
          "reason": null,
          "response_union_seconds": 214.645721802779,
          "scope": "GPU board only; not whole-server AC energy",
          "time_weighted_mean_watts": 89.91912998663662,
          "trapezoid_pieces": 239,
          "watt_hours": 5.361321266627646
        },
        "gpu_sample_count": 215,
        "host_sample_count": 104,
        "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
        "metrics": {
          "cgroup_memory_current_bytes": {
            "maximum": 86972829696,
            "mean": 86943777083.07692,
            "median": 86953607168.0,
            "minimum": 86751805440,
            "sample_count": 104
          },
          "gpu_board_watts": {
            "maximum": 118.27,
            "mean": 89.92758139534884,
            "median": 87.15,
            "minimum": 80.39,
            "sample_count": 215
          },
          "gpu_memory_used_bytes": {
            "maximum": 94176804864.0,
            "mean": 94171683911.44186,
            "median": 94172610560.0,
            "minimum": 94145347584.0,
            "sample_count": 215
          },
          "gpu_sm_clock_mhz": {
            "maximum": 2325.0,
            "mean": 2310.8418604651165,
            "median": 2317.0,
            "minimum": 2130.0,
            "sample_count": 215
          },
          "gpu_utilization_percent": {
            "maximum": 96.0,
            "mean": 58.62325581395349,
            "median": 78.0,
            "minimum": 0.0,
            "sample_count": 215
          },
          "host_mem_available_bytes": {
            "maximum": 95620997120,
            "mean": 95580679483.07692,
            "median": 95579918336.0,
            "minimum": 95542239232,
            "sample_count": 104
          },
          "process_pss_bytes": {
            "maximum": 73515068416,
            "mean": 68495140657.23077,
            "median": 71700847616.0,
            "minimum": 43000757248,
            "sample_count": 104
          },
          "process_rss_bytes": {
            "maximum": 73518055424,
            "mean": 68498155047.38461,
            "median": 71703834624.0,
            "minimum": 43003744256,
            "sample_count": 104
          }
        },
        "process_sample_count": 104,
        "response_union_seconds": 214.645721802779,
        "sampled_counter_deltas": {
          "deltas": {
            "major_faults": 10941,
            "read_syscall_bytes": 0,
            "storage_read_bytes": 26791116800,
            "storage_write_bytes": 32768
          },
          "observed_endpoint_seconds": 166.1631237110705,
          "observed_intervals": 24,
          "requested_intervals": 24
        }
      },
      "groups": [
        {
          "completed_output_tokens": 262,
          "completed_request_count": 24,
          "context_tokens": null,
          "failed_request_count": 0,
          "meaning": "Distinct authored tasks, one attempt each; no pooled practical latency ranking or general quality claim.",
          "measured_rounds": 24,
          "mode": "chat",
          "resources": {
            "gpu_energy": {
              "available": true,
              "interpolation": "linear_at_bracketed_boundaries",
              "joules": 19300.75655985953,
              "reason": null,
              "response_union_seconds": 214.645721802779,
              "scope": "GPU board only; not whole-server AC energy",
              "time_weighted_mean_watts": 89.91912998663662,
              "trapezoid_pieces": 239,
              "watt_hours": 5.361321266627646
            },
            "gpu_sample_count": 215,
            "host_sample_count": 104,
            "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
            "metrics": {
              "cgroup_memory_current_bytes": {
                "maximum": 86972829696,
                "mean": 86943777083.07692,
                "median": 86953607168.0,
                "minimum": 86751805440,
                "sample_count": 104
              },
              "gpu_board_watts": {
                "maximum": 118.27,
                "mean": 89.92758139534884,
                "median": 87.15,
                "minimum": 80.39,
                "sample_count": 215
              },
              "gpu_memory_used_bytes": {
                "maximum": 94176804864.0,
                "mean": 94171683911.44186,
                "median": 94172610560.0,
                "minimum": 94145347584.0,
                "sample_count": 215
              },
              "gpu_sm_clock_mhz": {
                "maximum": 2325.0,
                "mean": 2310.8418604651165,
                "median": 2317.0,
                "minimum": 2130.0,
                "sample_count": 215
              },
              "gpu_utilization_percent": {
                "maximum": 96.0,
                "mean": 58.62325581395349,
                "median": 78.0,
                "minimum": 0.0,
                "sample_count": 215
              },
              "host_mem_available_bytes": {
                "maximum": 95620997120,
                "mean": 95580679483.07692,
                "median": 95579918336.0,
                "minimum": 95542239232,
                "sample_count": 104
              },
              "process_pss_bytes": {
                "maximum": 73515068416,
                "mean": 68495140657.23077,
                "median": 71700847616.0,
                "minimum": 43000757248,
                "sample_count": 104
              },
              "process_rss_bytes": {
                "maximum": 73518055424,
                "mean": 68498155047.38461,
                "median": 71703834624.0,
                "minimum": 43003744256,
                "sample_count": 104
              }
            },
            "process_sample_count": 104,
            "response_union_seconds": 214.645721802779,
            "sampled_counter_deltas": {
              "deltas": {
                "major_faults": 10941,
                "read_syscall_bytes": 0,
                "storage_read_bytes": 26791116800,
                "storage_write_bytes": 32768
              },
              "observed_endpoint_seconds": 166.1631237110705,
              "observed_intervals": 24,
              "requested_intervals": 24
            }
          },
          "response_outer_span_seconds": 214.70433678099653,
          "response_window_union_seconds": 214.645721802779,
          "score": {
            "by_family": {
              "arithmetic": {
                "expected_value_pass": 4,
                "strict_format_pass": 5,
                "task_count": 6
              },
              "classification": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "extraction": {
                "expected_value_pass": 5,
                "strict_format_pass": 6,
                "task_count": 6
              },
              "grounded-qa": {
                "expected_value_pass": 6,
                "strict_format_pass": 6,
                "task_count": 6
              }
            },
            "by_language": {
              "en": {
                "expected_value_pass": 9,
                "strict_format_pass": 11,
                "task_count": 12
              },
              "pl": {
                "expected_value_pass": 12,
                "strict_format_pass": 12,
                "task_count": 12
              }
            },
            "completed_output": 24,
            "completed_transport": 24,
            "expected_value_pass": 21,
            "strict_format_pass": 23,
            "task_count": 24,
            "truncated": 0
          },
          "tasks": [
            {
              "completed_output": true,
              "connection_seconds": 13.583070762979332,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 11.753905638004653,
              "first_visible_content_seconds": 11.753905638004653,
              "generated_tokens": 8,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 13.320375241979491,
              "reasoning_events_observed": false,
              "response_seconds": 13.58291348197963,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 11.222968201967888,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 8.15195504296571,
              "first_visible_content_seconds": 8.15195504296571,
              "generated_tokens": 14,
              "input_tokens": 69,
              "language": "pl",
              "last_visible_content_seconds": 10.913710985973012,
              "reasoning_events_observed": false,
              "response_seconds": 11.222829844977241,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 9.111604891950265,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 7.609585767961107,
              "first_visible_content_seconds": 7.609585767961107,
              "generated_tokens": 8,
              "input_tokens": 93,
              "language": "pl",
              "last_visible_content_seconds": 8.899494110955857,
              "reasoning_events_observed": false,
              "response_seconds": 9.111451656965073,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.83672762895003,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.289020258001983,
              "first_visible_content_seconds": 6.289020258001983,
              "generated_tokens": 8,
              "input_tokens": 49,
              "language": "en",
              "last_visible_content_seconds": 7.617563018982764,
              "reasoning_events_observed": false,
              "response_seconds": 7.836561900970992,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 8.593278787040617,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.61590343300486,
              "first_visible_content_seconds": 6.61590343300486,
              "generated_tokens": 10,
              "input_tokens": 54,
              "language": "en",
              "last_visible_content_seconds": 8.366578131041024,
              "reasoning_events_observed": false,
              "response_seconds": 8.59308045799844,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 8.821014977991581,
              "failure_reason": null,
              "family": "arithmetic",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.420646424987353,
              "first_visible_content_seconds": 6.420646424987353,
              "generated_tokens": 12,
              "input_tokens": 56,
              "language": "en",
              "last_visible_content_seconds": 8.583703558018897,
              "reasoning_events_observed": false,
              "response_seconds": 8.820840994012542,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": false
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "arithmetic-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 11.624865641002543,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 7.236178473976906,
              "first_visible_content_seconds": 7.236178473976906,
              "generated_tokens": 21,
              "input_tokens": 84,
              "language": "pl",
              "last_visible_content_seconds": 11.386044683982618,
              "reasoning_events_observed": false,
              "response_seconds": 11.624679514963645,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 11.437166781979613,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.874179092003033,
              "first_visible_content_seconds": 6.874179092003033,
              "generated_tokens": 22,
              "input_tokens": 81,
              "language": "pl",
              "last_visible_content_seconds": 11.210961457982194,
              "reasoning_events_observed": false,
              "response_seconds": 11.43702499900246,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 9.50213718204759,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.080808499013074,
              "first_visible_content_seconds": 6.080808499013074,
              "generated_tokens": 17,
              "input_tokens": 72,
              "language": "pl",
              "last_visible_content_seconds": 9.297318112046923,
              "reasoning_events_observed": false,
              "response_seconds": 9.50197277701227,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 10.387426305038389,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.486654097039718,
              "first_visible_content_seconds": 6.486654097039718,
              "generated_tokens": 19,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 10.170321426005103,
              "reasoning_events_observed": false,
              "response_seconds": 10.387277828995138,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 9.730554731970187,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.0840944799711,
              "first_visible_content_seconds": 6.0840944799711,
              "generated_tokens": 18,
              "input_tokens": 51,
              "language": "en",
              "last_visible_content_seconds": 9.521599187981337,
              "reasoning_events_observed": false,
              "response_seconds": 9.730390105978586,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 9.24161141901277,
              "failure_reason": null,
              "family": "extraction",
              "finish_reason": "stop",
              "first_generated_event_seconds": 5.852490445016883,
              "first_visible_content_seconds": 5.852490445016883,
              "generated_tokens": 17,
              "input_tokens": 53,
              "language": "en",
              "last_visible_content_seconds": 9.026987073011696,
              "reasoning_events_observed": false,
              "response_seconds": 9.241449408000335,
              "score": {
                "completed": true,
                "expected_value_pass": false,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "extraction-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.9818881780374795,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.718955708027352,
              "first_visible_content_seconds": 6.718955708027352,
              "generated_tokens": 7,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 7.771279089036398,
              "reasoning_events_observed": false,
              "response_seconds": 7.981710618012585,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 8.087027255038265,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.827422423986718,
              "first_visible_content_seconds": 6.827422423986718,
              "generated_tokens": 7,
              "input_tokens": 87,
              "language": "pl",
              "last_visible_content_seconds": 7.865569265035447,
              "reasoning_events_observed": false,
              "response_seconds": 8.08685941301519,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.949002139037475,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.248805720009841,
              "first_visible_content_seconds": 6.248805720009841,
              "generated_tokens": 9,
              "input_tokens": 76,
              "language": "pl",
              "last_visible_content_seconds": 7.737396181037184,
              "reasoning_events_observed": false,
              "response_seconds": 7.9487804470118135,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.826057630998548,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.5673799850046635,
              "first_visible_content_seconds": 6.5673799850046635,
              "generated_tokens": 7,
              "input_tokens": 65,
              "language": "en",
              "last_visible_content_seconds": 7.621764203009661,
              "reasoning_events_observed": false,
              "response_seconds": 7.825891762971878,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.492189083015546,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.23327365500154,
              "first_visible_content_seconds": 6.23327365500154,
              "generated_tokens": 7,
              "input_tokens": 64,
              "language": "en",
              "last_visible_content_seconds": 7.279204571968876,
              "reasoning_events_observed": false,
              "response_seconds": 7.492010992020369,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 6.762584539013915,
              "failure_reason": null,
              "family": "classification",
              "finish_reason": "stop",
              "first_generated_event_seconds": 5.67244211898651,
              "first_visible_content_seconds": 5.67244211898651,
              "generated_tokens": 6,
              "input_tokens": 48,
              "language": "en",
              "last_visible_content_seconds": 6.542209693987388,
              "reasoning_events_observed": false,
              "response_seconds": 6.762443727988284,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "classification-en-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.668065410980489,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.403421438997611,
              "first_visible_content_seconds": 6.403421438997611,
              "generated_tokens": 7,
              "input_tokens": 78,
              "language": "pl",
              "last_visible_content_seconds": 7.461564408033155,
              "reasoning_events_observed": false,
              "response_seconds": 7.667886228009593,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.881595277984161,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.181176976999268,
              "first_visible_content_seconds": 6.181176976999268,
              "generated_tokens": 9,
              "input_tokens": 89,
              "language": "pl",
              "last_visible_content_seconds": 7.651056489965413,
              "reasoning_events_observed": false,
              "response_seconds": 7.881407508975826,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 8.305406797968317,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 7.048749969981145,
              "first_visible_content_seconds": 7.048749969981145,
              "generated_tokens": 7,
              "input_tokens": 108,
              "language": "pl",
              "last_visible_content_seconds": 8.094621035968885,
              "reasoning_events_observed": false,
              "response_seconds": 8.305239226960111,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-pl-3",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.758344300033059,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.494405768986326,
              "first_visible_content_seconds": 6.494405768986326,
              "generated_tokens": 7,
              "input_tokens": 52,
              "language": "en",
              "last_visible_content_seconds": 7.553320523991715,
              "reasoning_events_observed": false,
              "response_seconds": 7.7581756060244516,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-1",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 7.272968183970079,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 6.025387093948666,
              "first_visible_content_seconds": 6.025387093948666,
              "generated_tokens": 7,
              "input_tokens": 61,
              "language": "en",
              "last_visible_content_seconds": 7.066188130993396,
              "reasoning_events_observed": false,
              "response_seconds": 7.272804338950664,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-2",
              "truncated": false
            },
            {
              "completed_output": true,
              "connection_seconds": 8.572214197949506,
              "failure_reason": null,
              "family": "grounded-qa",
              "finish_reason": "stop",
              "first_generated_event_seconds": 7.092147689952981,
              "first_visible_content_seconds": 7.092147689952981,
              "generated_tokens": 8,
              "input_tokens": 83,
              "language": "en",
              "last_visible_content_seconds": 8.351030136982445,
              "reasoning_events_observed": false,
              "response_seconds": 8.572038961981889,
              "score": {
                "completed": true,
                "expected_value_pass": true,
                "strict_format_pass": true
              },
              "server_decode_tokens_per_second": null,
              "status": "completed",
              "task_id": "grounded-en-3",
              "truncated": false
            }
          ],
          "warmup_completed_rounds": 1,
          "warmup_requests_per_round": 1
        }
      ],
      "id": "practical-235b",
      "model": "qwen3-235b-q5",
      "module": "practical",
      "placement": {
        "buffers": [
          {
            "device": "CPU_Mapped",
            "reported_mib": 27759.73
          },
          {
            "device": "CPU_Mapped",
            "reported_mib": 28323.86
          },
          {
            "device": "CPU_Mapped",
            "reported_mib": 18322.9
          },
          {
            "device": "CUDA0",
            "reported_mib": 86732.62
          }
        ],
        "gpu_layers": 95,
        "meaning": "Runtime buffer allocation is not measured resident process RAM.",
        "requested_cpu_expert_layers": 44,
        "total_layers": 95
      },
      "record_sha256": "e4b245570fb95c8646e8413e1491daee97364c0f61d1eb5d5f55d65ad0f7354c",
      "resources_by_phase": {
        "baseline": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86158008320,
              "mean": 86149492736,
              "median": 86149492736.0,
              "minimum": 86140977152,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 16.24,
              "mean": 16.18,
              "median": 16.18,
              "minimum": 16.12,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 2097152.0,
              "mean": 2097152.0,
              "median": 2097152.0,
              "minimum": 2097152.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 180.0,
              "mean": 180.0,
              "median": 180.0,
              "minimum": 180.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 96376918016,
              "mean": 96365586432,
              "median": 96365586432.0,
              "minimum": 96354254848,
              "sample_count": 2
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1007638040464371,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "cleanup": {
          "gpu_energy": {
            "available": false,
            "joules": null,
            "reason": "unbracketed_boundary"
          },
          "gpu_sample_count": 1,
          "host_sample_count": 1,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86669819904,
              "mean": 86669819904,
              "median": 86669819904,
              "minimum": 86669819904,
              "sample_count": 1
            },
            "gpu_board_watts": {
              "maximum": 78.13,
              "mean": 78.13,
              "median": 78.13,
              "minimum": 78.13,
              "sample_count": 1
            },
            "gpu_memory_used_bytes": {
              "maximum": 94176804864.0,
              "mean": 94176804864.0,
              "median": 94176804864.0,
              "minimum": 94176804864.0,
              "sample_count": 1
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2272.0,
              "mean": 2272.0,
              "median": 2272.0,
              "minimum": 2272.0,
              "sample_count": 1
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 1
            },
            "host_mem_available_bytes": {
              "maximum": 95835246592,
              "mean": 95835246592,
              "median": 95835246592,
              "minimum": 95835246592,
              "sample_count": 1
            },
            "process_pss_bytes": null,
            "process_rss_bytes": null
          },
          "process_sample_count": 0,
          "response_union_seconds": 1.1610513600171544,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 1
          }
        },
        "load": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 4127.699400766261,
            "reason": null,
            "response_union_seconds": 79.19833586795721,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 52.11851177843098,
            "trapezoid_pieces": 80,
            "watt_hours": 1.1465831668795168
          },
          "gpu_sample_count": 79,
          "host_sample_count": 58,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86973087744,
              "mean": 86923087448.27586,
              "median": 86962081792.0,
              "minimum": 85281234944,
              "sample_count": 58
            },
            "gpu_board_watts": {
              "maximum": 77.81,
              "mean": 52.37987341772152,
              "median": 71.54,
              "minimum": 15.46,
              "sample_count": 79
            },
            "gpu_memory_used_bytes": {
              "maximum": 94088724480.0,
              "mean": 38674748066.025314,
              "median": 587202560.0,
              "minimum": 587202560.0,
              "sample_count": 79
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2280.0,
              "mean": 1507.860759493671,
              "median": 2272.0,
              "minimum": 180.0,
              "sample_count": 79
            },
            "gpu_utilization_percent": {
              "maximum": 34.0,
              "mean": 7.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 79
            },
            "host_mem_available_bytes": {
              "maximum": 96077516800,
              "mean": 95845226142.89655,
              "median": 95821776896.0,
              "minimum": 95730827264,
              "sample_count": 58
            },
            "process_pss_bytes": {
              "maximum": 86550830080,
              "mean": 72792922747.58621,
              "median": 85804739584.0,
              "minimum": 1283374080,
              "sample_count": 58
            },
            "process_rss_bytes": {
              "maximum": 86553817088,
              "mean": 72795910002.75862,
              "median": 85807726592.0,
              "minimum": 1286361088,
              "sample_count": 58
            }
          },
          "process_sample_count": 58,
          "response_union_seconds": 79.19833586795721,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 57576,
              "read_syscall_bytes": 13,
              "storage_read_bytes": 144988631040,
              "storage_write_bytes": 12288
            },
            "observed_endpoint_seconds": 77.89480808604276,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "ready": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 190.05088581740472,
            "reason": null,
            "response_union_seconds": 2.2796509800245985,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 83.36841362240196,
            "trapezoid_pieces": 4,
            "watt_hours": 0.05279191272705687
          },
          "gpu_sample_count": 2,
          "host_sample_count": 2,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86942056448,
              "mean": 86935918592,
              "median": 86935918592.0,
              "minimum": 86929780736,
              "sample_count": 2
            },
            "gpu_board_watts": {
              "maximum": 87.87,
              "mean": 82.33000000000001,
              "median": 82.33000000000001,
              "minimum": 76.79,
              "sample_count": 2
            },
            "gpu_memory_used_bytes": {
              "maximum": 94176804864.0,
              "mean": 94134861824.0,
              "median": 94134861824.0,
              "minimum": 94092918784.0,
              "sample_count": 2
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2295.0,
              "mean": 2283.5,
              "median": 2283.5,
              "minimum": 2272.0,
              "sample_count": 2
            },
            "gpu_utilization_percent": {
              "maximum": 0.0,
              "mean": 0.0,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 2
            },
            "host_mem_available_bytes": {
              "maximum": 95807406080,
              "mean": 95690938368,
              "median": 95690938368.0,
              "minimum": 95574470656,
              "sample_count": 2
            },
            "process_pss_bytes": {
              "maximum": 73541479424,
              "mean": 42100476928,
              "median": 42100476928.0,
              "minimum": 10659474432,
              "sample_count": 2
            },
            "process_rss_bytes": {
              "maximum": 73544466432,
              "mean": 42103463936,
              "median": 42103463936.0,
              "minimum": 10662461440,
              "sample_count": 2
            }
          },
          "process_sample_count": 2,
          "response_union_seconds": 2.2796509800245985,
          "sampled_counter_deltas": {
            "deltas": null,
            "observed_endpoint_seconds": 0.0,
            "observed_intervals": 0,
            "requested_intervals": 2
          }
        },
        "responses": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 19306.502753033474,
            "reason": null,
            "response_union_seconds": 214.70691833097953,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 89.92026387930224,
            "trapezoid_pieces": 216,
            "watt_hours": 5.362917431398187
          },
          "gpu_sample_count": 215,
          "host_sample_count": 104,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86972829696,
              "mean": 86943777083.07692,
              "median": 86953607168.0,
              "minimum": 86751805440,
              "sample_count": 104
            },
            "gpu_board_watts": {
              "maximum": 118.27,
              "mean": 89.92758139534884,
              "median": 87.15,
              "minimum": 80.39,
              "sample_count": 215
            },
            "gpu_memory_used_bytes": {
              "maximum": 94176804864.0,
              "mean": 94171683911.44186,
              "median": 94172610560.0,
              "minimum": 94145347584.0,
              "sample_count": 215
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2325.0,
              "mean": 2310.8418604651165,
              "median": 2317.0,
              "minimum": 2130.0,
              "sample_count": 215
            },
            "gpu_utilization_percent": {
              "maximum": 96.0,
              "mean": 58.62325581395349,
              "median": 78.0,
              "minimum": 0.0,
              "sample_count": 215
            },
            "host_mem_available_bytes": {
              "maximum": 95620997120,
              "mean": 95580679483.07692,
              "median": 95579918336.0,
              "minimum": 95542239232,
              "sample_count": 104
            },
            "process_pss_bytes": {
              "maximum": 73515068416,
              "mean": 68495140657.23077,
              "median": 71700847616.0,
              "minimum": 43000757248,
              "sample_count": 104
            },
            "process_rss_bytes": {
              "maximum": 73518055424,
              "mean": 68498155047.38461,
              "median": 71703834624.0,
              "minimum": 43003744256,
              "sample_count": 104
            }
          },
          "process_sample_count": 104,
          "response_union_seconds": 214.70691833097953,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 23997,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 29071040512,
              "storage_write_bytes": 147456
            },
            "observed_endpoint_seconds": 212.04975535703124,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        },
        "warmup": {
          "gpu_energy": {
            "available": true,
            "interpolation": "linear_at_bracketed_boundaries",
            "joules": 995.0526985724146,
            "reason": null,
            "response_union_seconds": 13.300178417004645,
            "scope": "GPU board only; not whole-server AC energy",
            "time_weighted_mean_watts": 74.81498874482858,
            "trapezoid_pieces": 14,
            "watt_hours": 0.27640352738122625
          },
          "gpu_sample_count": 13,
          "host_sample_count": 11,
          "meaning": "Samples inside declared interval union. Means are arithmetic sample means; peaks need not coincide. Counter deltas exclude unsampled edges; not minimum required RAM.",
          "metrics": {
            "cgroup_memory_current_bytes": {
              "maximum": 86970953728,
              "mean": 86944415744,
              "median": 86959644672,
              "minimum": 86869540864,
              "sample_count": 11
            },
            "gpu_board_watts": {
              "maximum": 90.66,
              "mean": 74.63384615384615,
              "median": 73.3,
              "minimum": 73.06,
              "sample_count": 13
            },
            "gpu_memory_used_bytes": {
              "maximum": 94141153280.0,
              "mean": 94141153280.0,
              "median": 94141153280.0,
              "minimum": 94141153280.0,
              "sample_count": 13
            },
            "gpu_sm_clock_mhz": {
              "maximum": 2272.0,
              "mean": 2272.0,
              "median": 2272.0,
              "minimum": 2272.0,
              "sample_count": 13
            },
            "gpu_utilization_percent": {
              "maximum": 1.0,
              "mean": 0.07692307692307693,
              "median": 0.0,
              "minimum": 0.0,
              "sample_count": 13
            },
            "host_mem_available_bytes": {
              "maximum": 95628480512,
              "mean": 95611195019.63637,
              "median": 95612923904,
              "minimum": 95555829760,
              "sample_count": 11
            },
            "process_pss_bytes": {
              "maximum": 41014807552,
              "mean": 26555915264,
              "median": 26415000576,
              "minimum": 12266339328,
              "sample_count": 11
            },
            "process_rss_bytes": {
              "maximum": 41017794560,
              "mean": 26558902272,
              "median": 26417987584,
              "minimum": 12269326336,
              "sample_count": 11
            }
          },
          "process_sample_count": 11,
          "response_union_seconds": 13.300178417004645,
          "sampled_counter_deltas": {
            "deltas": {
              "major_faults": 224642,
              "read_syscall_bytes": 0,
              "storage_read_bytes": 31496257536,
              "storage_write_bytes": 0
            },
            "observed_endpoint_seconds": 12.271783648000564,
            "observed_intervals": 1,
            "requested_intervals": 1
          }
        }
      },
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "settings": {
        "batch": 2048,
        "cache_prompt": false,
        "cache_ram_bytes": 0,
        "concurrency": 1,
        "context_shift": false,
        "contexts": [],
        "endpoint": "/v1/chat/completions",
        "fit": "off",
        "flash_attention": true,
        "ignore_eos": false,
        "kv_cache": "f16",
        "lazy_mode": "off",
        "load_mode": "mmap",
        "measured_rounds": 24,
        "mlock": false,
        "network_scope": "loopback",
        "output_tokens": 1024,
        "p95_minimum_requests": 200,
        "placement": {
          "cpu_expert_layers": 44,
          "gpu_layers": "all"
        },
        "reasoning": "disabled",
        "seed": 20260927,
        "server_slots": 1,
        "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187",
        "temperature": 0,
        "threads": 16,
        "tokenization": "natural_chat_template_no_padding",
        "total_context_tokens": 8704,
        "ubatch": 512,
        "warmup": "one_disjoint_transport_request",
        "warmup_rounds": 1
      },
      "status": "completed",
      "telemetry": {
        "gpu_samples": 312,
        "host_samples": 178,
        "maximum_gpu_gap_seconds": 1.1819236329756677,
        "maximum_host_gap_seconds": 2.7705111099639907,
        "meaning": "Sampled observations; no numeric temperatures published and no warranty between samples.",
        "thermal_slowdown_fully_available": true,
        "thermal_slowdown_observed_samples": 0,
        "thermal_slowdown_states": {
          "hardware": {
            "active": 0,
            "not_active": 312,
            "unavailable": 0
          },
          "software": {
            "active": 0,
            "not_active": 312,
            "unavailable": 0
          }
        }
      }
    }
  ],
  "comparisons": [],
  "environment": {
    "kernel": "6.12.0-211.56.1.el10_2.0.1.x86_64",
    "nvidia_driver": "595.91.07",
    "os": {
      "id": "rocky",
      "name": "Rocky Linux",
      "version_id": "10.2"
    }
  },
  "hardware": {
    "cpu": "AMD EPYC 4565P 16-Core Processor",
    "fan_profile": "GPU300WAuto",
    "gpu": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
    "gpu_nominal_vram_gb": 96,
    "gpu_vram_bytes": 102641958912,
    "installed_ram_bytes": 103079215104,
    "kernel_mem_total_bytes": 100271353856,
    "logical_cpu_count": 32,
    "physical_cpu_count": 16
  },
  "historical_results_pooled": false,
  "limitations": [
    "Externally pinned capture integrity is not remote hardware attestation.",
    "Five repeated rounds share a process and are not five independent installations.",
    "Synthetic token latency is not practical first-visible answer latency.",
    "GPU board power is not whole-server AC power; energy is unavailable without bracketing samples.",
    "Sampled memory peaks may not coincide and do not establish minimum required system RAM.",
    "The practical corpus contains 24 authored tasks, not a broad model-quality benchmark.",
    "Engine telemetry includes process loading and internal warmup; exact per-sample resource windows are unavailable.",
    "No historical result is relabeled or pooled into this versioned series."
  ],
  "models": [
    {
      "architecture": "dense",
      "chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "files": [
        {
          "bytes": 5027783488,
          "sha256": "d98cdcbd03e17ce47681435b5150e34c1417f50b5c0019dd560e4882c5745785"
        }
      ],
      "id": "qwen3-8b-q4",
      "name": "Qwen3-8B",
      "quantization": "Q4_K_M",
      "repository": "Qwen/Qwen3-8B-GGUF",
      "revision": "7c41481f57cb95916b40956ab2f0b139b296d974",
      "tokenizer_sha256": "2629971f294a1a43ef545d689a2090a820e0e1a15e5371e1c88d1c21159ef849"
    },
    {
      "architecture": "dense",
      "chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "files": [
        {
          "bytes": 19762149024,
          "sha256": "efd971561896866f0e910cce52761ca77b1b138090c7f15fe284676d57d1f689"
        }
      ],
      "id": "qwen3-32b-q4",
      "name": "Qwen3-32B",
      "quantization": "Q4_K_M",
      "repository": "Qwen/Qwen3-32B-GGUF",
      "revision": "938a7432affaec9157f883a87164e2646ae17555",
      "tokenizer_sha256": "2629971f294a1a43ef545d689a2090a820e0e1a15e5371e1c88d1c21159ef849"
    },
    {
      "architecture": "dense",
      "chat_template_sha256": "d5495a1e5db0611132a97e46a65dbb64a642a499421228b9c8b93229097fa9a4",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "4e9918361c284a93880606d182d64da6a9fe97cdc1f5c5a78c1c8840444246fc",
      "files": [
        {
          "bytes": 3986869120,
          "sha256": "3531a90691b5f4edb4bdf488246d72ce0b5d40cacfab1489718a662cec386e43"
        },
        {
          "bytes": 3877864544,
          "sha256": "a4cfb079461591e2c23ef73953aeba4b2520a9b93afaa841f31adf4e480647ab"
        },
        {
          "bytes": 3963622720,
          "sha256": "c67c3641d018b7ca646e7c37b50152687d4b52ce6ab6bf170b006a756b52778d"
        },
        {
          "bytes": 3941463040,
          "sha256": "92f5149d685ad5c5ab0fc30f5e0725d95a746d9a8e54d8397707b97dc435cd76"
        },
        {
          "bytes": 3963688384,
          "sha256": "f5abb6ecd7f3ec4438a4e89ae704acd43405182a0477776d5e8aa35a71a5ecdc"
        },
        {
          "bytes": 3941463040,
          "sha256": "ead67a4ac2648b0f7b71fea755f910ed208f3f8b881b55c208fc3fdca67359ee"
        },
        {
          "bytes": 3826849152,
          "sha256": "f7b168a5fdf78695736a6441dc846b19251339750ea78bf993057b7bf499671d"
        },
        {
          "bytes": 3983901792,
          "sha256": "4ab253ada85c14097597546ac13cfb1426cb175d1d8da520b7a2904d0c323e32"
        },
        {
          "bytes": 3921183968,
          "sha256": "51661b80af773b488e807a03fc7ea31bb64630c50e3ee77fbb5e37bdd9ff9979"
        },
        {
          "bytes": 3995005888,
          "sha256": "fcc180433606a81e2137ba1f088defe3ad51ec9353f5d129a7209614597e16f9"
        },
        {
          "bytes": 3586683584,
          "sha256": "57b9437649f12062bdbed6c623c3f6adc6924a3d2f4e7aabdd3ab32883f5485d"
        },
        {
          "bytes": 1021870240,
          "sha256": "60ae16428f9edb62b60e6c8657ba9344c6fae6ef44c497845dc6a6fd002e4d5b"
        }
      ],
      "id": "qwen25-72b-q4",
      "name": "Qwen2.5-72B-Instruct",
      "quantization": "Q4_K_M",
      "repository": "Qwen/Qwen2.5-72B-Instruct-GGUF",
      "revision": "7ca3bc388f97b264c4283bc9bf1055e2abc38441",
      "tokenizer_sha256": "67b3d77a05acfb80f96742cec2740cc6ee8fffadef98e655e855baf58a3946b1"
    },
    {
      "architecture": "moe",
      "chat_template_sha256": "a4c9919cbbd4acdd51ccffe22da049264b1b73e59055fa58811a99efbd7c8146",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "b2215de6da8ba369957eece8c5aa18f4af94f6c4d311d85e691373a421d80e89",
      "files": [
        {
          "bytes": 63387346208,
          "sha256": "582bd40f6886200101f4c4ed9f25f3fe80cc14c86e9e2b37746cd8904a0c622d"
        }
      ],
      "id": "gpt-oss-120b-mxfp4",
      "name": "GPT-OSS-120B",
      "quantization": "MXFP4",
      "repository": "ggml-org/gpt-oss-120b-GGUF",
      "revision": "238abdd290bb874b90a5da1b4549881b7d05c091",
      "tokenizer_sha256": "d67292e223a919a15af6ec5257ae8e5e6c6782203636375f7008603d27b7a391"
    },
    {
      "architecture": "dense",
      "chat_template_sha256": "d5495a1e5db0611132a97e46a65dbb64a642a499421228b9c8b93229097fa9a4",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "4e9918361c284a93880606d182d64da6a9fe97cdc1f5c5a78c1c8840444246fc",
      "files": [
        {
          "bytes": 3878876224,
          "sha256": "31a2b3c4e62dc185b54b9c33b1b2898f06fde167e3311cc5380e08692098e781"
        },
        {
          "bytes": 3743845120,
          "sha256": "747e66c38a1f4f8d496533e2108c7592e10220afc65c480d13cbe14d0109d700"
        },
        {
          "bytes": 3743845152,
          "sha256": "4346236ce045e136ad4cbcd7b86a88ae66dfadeed033175965e9f331a76e90d9"
        },
        {
          "bytes": 3743845184,
          "sha256": "152f922d2dd4c18d49f6b3c00bb1cc20629ed042d494527e1c350da41043a446"
        },
        {
          "bytes": 3743845184,
          "sha256": "24d6a097d3d3ce727b1f123957713cdaba0d5d64b9aaf559afb4703ea51ce577"
        },
        {
          "bytes": 3743845184,
          "sha256": "4ef22906aa68f1b7c067c4469677d7490455a9ea1a2fcda67af29bb7d1ffc49e"
        },
        {
          "bytes": 3743845184,
          "sha256": "b9dd79952b0a7d771633f32c253da0b771eb09efe173eeef50c250979c5ddd22"
        },
        {
          "bytes": 3743845184,
          "sha256": "af925e8985c7d5e30d6ee98e275d93cdc2ca5777173d5601ebc9d6fbd4f4d01e"
        },
        {
          "bytes": 3743845184,
          "sha256": "057a909a1bfd954aaf9b4b94b226b98a1a7b2a5b19861a708bf2879bb12c704e"
        },
        {
          "bytes": 3743845184,
          "sha256": "c84a7441c2feab7213ee35b989334079276701a4bc39017e68ac7096966a08b6"
        },
        {
          "bytes": 3743845184,
          "sha256": "48f91fc504df34b89edc7d709925b9b69900a3a3671b1b4ce8d7384ff3117255"
        },
        {
          "bytes": 3743845184,
          "sha256": "84c3b6949803166b84bb8ffda810fbf2f56aec4e32b9e0f1614a43f84d4214a2"
        },
        {
          "bytes": 3743845184,
          "sha256": "00fd616480122ef2fddf3888436ee62d04ec5dc081cad0f8d0db9e7f1b241faf"
        },
        {
          "bytes": 3743845184,
          "sha256": "58a93e4a25a3bd34ae4327f4b40d1fd7c90b20ae34ed29a5fc0804333031d998"
        },
        {
          "bytes": 3743845184,
          "sha256": "f3b643f7b02f52da7a68ab0eee711ba160904ed235bea64da6cca5b46bb5db48"
        },
        {
          "bytes": 3743845184,
          "sha256": "721536952fea2b497e74681da2c65b1bd783ef2f3193ab41057fcaa31a64b495"
        },
        {
          "bytes": 3743845184,
          "sha256": "3b21b0bd05319458227e5865d1a79e62e1d63168bf8060e8d0d6f1e0da51c324"
        },
        {
          "bytes": 3743845184,
          "sha256": "1804cb39eefead3acd09ffafd064d969ec07161aceb62edbbe8975a0472391eb"
        },
        {
          "bytes": 3743845184,
          "sha256": "a6c8e5f402cb84f8d29813f811cb93c3a694dfc8a6398e2081031ced3a49b5ad"
        },
        {
          "bytes": 3743845184,
          "sha256": "047d7b4c5c363d17e86f19b3bd66e38663cb1484555aa897cd4892c1993881d3"
        },
        {
          "bytes": 2518066176,
          "sha256": "b6a7a5be990bed3f18695ded8d8fca3f545176045a37784e0443e2bd6633c6bc"
        }
      ],
      "id": "qwen25-72b-q8",
      "name": "Qwen2.5-72B-Instruct",
      "quantization": "Q8_0",
      "repository": "Qwen/Qwen2.5-72B-Instruct-GGUF",
      "revision": "7ca3bc388f97b264c4283bc9bf1055e2abc38441",
      "tokenizer_sha256": "67b3d77a05acfb80f96742cec2740cc6ee8fffadef98e655e855baf58a3946b1"
    },
    {
      "architecture": "moe",
      "chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "effective_chat_template_rule": "llama.cpp-7fe450e19-chat-source-v1",
      "effective_chat_template_sha256": "57f1fd00f0013a2be96aa79b857391f27e23df5b5f847072b524c897e24d0361",
      "files": [
        {
          "bytes": 29673975136,
          "sha256": "706382f26d0d6f72cd89c9ac1bfeb40d4b4afaee986dad2ea229ea24f9168fab"
        },
        {
          "bytes": 29748770496,
          "sha256": "9ab54360c3809c580439b75fe7e16f3af24adcc701c352e39b06ad557aca0ca3"
        },
        {
          "bytes": 29749049024,
          "sha256": "93c863b35e810b97605dc973c2881e608751a269e773b3772c42e5e25cb01d79"
        },
        {
          "bytes": 29642094272,
          "sha256": "c0322fa3f66ca392a9d7f95bb2543b6f6213ac04c16e613a2ae3a82429452071"
        },
        {
          "bytes": 29856282304,
          "sha256": "1dce96f79ba9c2d86d36bbd856101f7a2e4f3c4cff4ef4a2a970c0ba336edb15"
        },
        {
          "bytes": 18143985760,
          "sha256": "15a2192eadc2cc2da6fc0df6d25cb6d39626f292716526cfee8baba9043fbe62"
        }
      ],
      "id": "qwen3-235b-q5",
      "name": "Qwen3-235B-A22B",
      "quantization": "Q5_K_M",
      "repository": "Qwen/Qwen3-235B-A22B-GGUF",
      "revision": "211e807ecd6d37446a747409108783e39a04e80f",
      "tokenizer_sha256": "2629971f294a1a43ef545d689a2090a820e0e1a15e5371e1c88d1c21159ef849"
    }
  ],
  "protocol": "gsh-llm-core/1.1",
  "runtime": {
    "archives_sha256": [
      "c2ab9e19838513ff69d1af8d999ad717dd3c7ee4714ac04c7ed5ab9077c50e4e",
      "1466daea60aad1144819e151b2bae19d54556cf1da6c129c4f55a5ded2637c25"
    ],
    "binary_sha256": "376626f953efd7bbf16dee112747196d2b1de26717bcd1d37e203340dc5c043c",
    "commit": "7fe450e19305b828c199d602c23a8337aaa1f03b",
    "name": "llama.cpp",
    "release": "b11146"
  },
  "schema": "gsh-llm-core-article/1",
  "sources": [
    {
      "exporter_sha256": "4f4046b91d99ba35175c3c6c04473a0021f54c6baa4892625cb66435f4c7e2ec",
      "finished_at": "2026-09-27T13:53:28.036613+00:00",
      "plan_id": "core-pilot8-20260927t135100-pilot8",
      "plan_sha256": "2952d4f1f06fa8282ea9ea2126cf5954bf3d774d16f9bc55a160ed6d808c3d49",
      "protocol_sha256": "3ce8ce9f2b17e66a655c07fd7e7aa3f1762790ad9be38f87d29e5e1279dec3ac",
      "record_sha256": "6c09a429d3b733b934749454b62d1a4243f5dce7c508680647c7d703a2b5daca",
      "reviewed_export_sha256": "2d1a82644c4ba9ba504ddf8fd2c7fe3c253e15cc9811363c52186bd4b7510a07",
      "runner_sha256": "1bac2436c5aa2eca4f4b42cc9b8f0942b5c66256eb1a41a2169b88c12b044f4e",
      "started_at": "2026-09-27T13:51:57.017889+00:00",
      "task_corpus_sha256": "0642db889565ce25b980e24afa5048eb83e4d15c36754457474a65b2c1f71c21"
    },
    {
      "exporter_sha256": "4f4046b91d99ba35175c3c6c04473a0021f54c6baa4892625cb66435f4c7e2ec",
      "finished_at": "2026-09-27T15:11:44.063808+00:00",
      "plan_id": "core-core-20260927t135400-core",
      "plan_sha256": "d6b25477667ae18cdbc23760f8a5540291e3d88246734ade84d167a32fc9011c",
      "protocol_sha256": "3ce8ce9f2b17e66a655c07fd7e7aa3f1762790ad9be38f87d29e5e1279dec3ac",
      "record_sha256": "fb26d7e444a2a826e9f652b7e048338d0bb8ddb92c1672ae6b919fc2334d492c",
      "reviewed_export_sha256": "59f09a9b924c4cb4ae1370f85109347542e72457c07f0a7a8ff351b31fedab64",
      "runner_sha256": "1bac2436c5aa2eca4f4b42cc9b8f0942b5c66256eb1a41a2169b88c12b044f4e",
      "started_at": "2026-09-27T13:56:01.723397+00:00",
      "task_corpus_sha256": "0642db889565ce25b980e24afa5048eb83e4d15c36754457474a65b2c1f71c21"
    },
    {
      "exporter_sha256": "4f4046b91d99ba35175c3c6c04473a0021f54c6baa4892625cb66435f4c7e2ec",
      "finished_at": "2026-09-27T20:35:44.649240+00:00",
      "plan_id": "core-offload-20260927t180810-offloade234fdb2",
      "plan_sha256": "337114b4eef68828eabe560dad626826411d46979051d71e89bd51d5941dca27",
      "protocol_sha256": "3ce8ce9f2b17e66a655c07fd7e7aa3f1762790ad9be38f87d29e5e1279dec3ac",
      "record_sha256": "711f4d34a08d22e6937d72802601ed59fe398c64f9fb191925cb9eebb2cd6b2d",
      "reviewed_export_sha256": "4b17dd2a9c39d7cab209a102bfdbd49f8324b54625bc01cbe7d9aeaf6b87dd5a",
      "runner_sha256": "1bac2436c5aa2eca4f4b42cc9b8f0942b5c66256eb1a41a2169b88c12b044f4e",
      "started_at": "2026-09-27T18:08:57.430966+00:00",
      "task_corpus_sha256": "0642db889565ce25b980e24afa5048eb83e4d15c36754457474a65b2c1f71c21"
    },
    {
      "exporter_sha256": "4f4046b91d99ba35175c3c6c04473a0021f54c6baa4892625cb66435f4c7e2ec",
      "finished_at": "2026-09-27T16:13:37.029489+00:00",
      "plan_id": "core-capacity-20260927t151737-capacityb1f78281",
      "plan_sha256": "20b8022e204c32d4826a7ef00445ccd3e6fc8b0fe2bbccc222714e0fb4fbe2d3",
      "protocol_sha256": "3ce8ce9f2b17e66a655c07fd7e7aa3f1762790ad9be38f87d29e5e1279dec3ac",
      "record_sha256": "b80b2a18f3d821f8c7929b5d1d418771b14e88b3e73c0059056a1a8c31d33416",
      "reviewed_export_sha256": "f32d7fb2e6a87c7e3188b5b27d217655dd296aa783ea2397dc4822420a6d555c",
      "runner_sha256": "1bac2436c5aa2eca4f4b42cc9b8f0942b5c66256eb1a41a2169b88c12b044f4e",
      "started_at": "2026-09-27T15:24:32.393663+00:00",
      "task_corpus_sha256": "0642db889565ce25b980e24afa5048eb83e4d15c36754457474a65b2c1f71c21"
    }
  ],
  "status": "completed",
  "test_date_timezone": "Europe/Warsaw",
  "tested_at": "2026-09-27"
}
