{ "article_key": "blackwell-llm-benchmark", "cases": [ { "case_wall_seconds": 1827.0776107079582, "groups": [ { "context_tokens": 2048, "internal_seconds": { "maximum": 1.658697238, "mean": 1.6562117116, "median": 1.655943854, "minimum": 1.653675521, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "prefill", "process_seconds_including_load_and_warmup": 17.40895069204271, "sample_count": 5, "telemetry": { "gpu_samples": 20, "host_samples": 18, "maximum_gpu_gap_seconds": 1.0054792480077595, "maximum_host_gap_seconds": 1.576730280998163, "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": 20, "unavailable": 0 }, "software": { "active": 0, "not_active": 20, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 1238.4533567755461, "mean": 1236.5588412384782, "median": 1236.7569075805152, "minimum": 1234.7039309412535, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 3446.098740907861, "reason": null, "response_union_seconds": 17.40895069204271, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 197.94982488421923, "trapezoid_pieces": 18, "watt_hours": 0.9572496502521837 }, "gpu_sample_count": 17, "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": 17356328960, "mean": 17327834453.333332, "median": 17333223424, "minimum": 17241808896, "sample_count": 15 }, "gpu_board_watts": { "maximum": 300.2, "mean": 203.9064705882353, "median": 299.81, "minimum": 42.1, "sample_count": 17 }, "gpu_memory_used_bytes": { "maximum": 45006979072.0, "mean": 41999539742.117645, "median": 45000687616.0, "minimum": 587202560.0, "sample_count": 17 }, "gpu_sm_clock_mhz": { "maximum": 2310.0, "mean": 1877.764705882353, "median": 1612.0, "minimum": 1567.0, "sample_count": 17 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 82.11764705882354, "median": 100.0, "minimum": 0.0, "sample_count": 17 }, "host_mem_available_bytes": { "maximum": 96195272704, "mean": 96123561301.33333, "median": 96104046592, "minimum": 96052207616, "sample_count": 15 }, "process_pss_bytes": { "maximum": 44343740416, "mean": 12896031334.4, "median": 1184750592, "minimum": 459343872, "sample_count": 15 }, "process_rss_bytes": { "maximum": 44346585088, "mean": 12898970282.666666, "median": 1187594240, "minimum": 462188544, "sample_count": 15 } }, "process_sample_count": 15, "response_union_seconds": 17.40895069204271, "sampled_counter_deltas": { "deltas": { "major_faults": 1040, "read_syscall_bytes": 8, "storage_read_bytes": 22769664, "storage_write_bytes": 4096 }, "observed_endpoint_seconds": 16.057237872038968, "observed_intervals": 1, "requested_intervals": 1 } } }, { "context_tokens": 2048, "internal_seconds": { "maximum": 9.95410588, "mean": 9.871335827, "median": 9.89882508, "minimum": 9.741621938, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "decode", "process_seconds_including_load_and_warmup": 58.6424638669705, "sample_count": 5, "telemetry": { "gpu_samples": 61, "host_samples": 58, "maximum_gpu_gap_seconds": 1.006989141984377, "maximum_host_gap_seconds": 1.5916822329745628, "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": 61, "unavailable": 0 }, "software": { "active": 0, "not_active": 61, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 26.27899148923018, "mean": 25.93544400792132, "median": 25.861655088464296, "minimum": 25.71803063842837, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 15796.466041639229, "reason": null, "response_union_seconds": 58.6424638669705, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 269.36907148842283, "trapezoid_pieces": 59, "watt_hours": 4.387907233788675 }, "gpu_sample_count": 58, "host_sample_count": 55, "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": 18062663680, "mean": 17965912287.418182, "median": 18044813312, "minimum": 17252597760, "sample_count": 55 }, "gpu_board_watts": { "maximum": 301.08, "mean": 271.4705172413793, "median": 299.99, "minimum": 43.93, "sample_count": 58 }, "gpu_memory_used_bytes": { "maximum": 45090865152.0, "mean": 44219534653.793106, "median": 45090865152.0, "minimum": 587202560.0, "sample_count": 58 }, "gpu_sm_clock_mhz": { "maximum": 2310.0, "mean": 1546.0344827586207, "median": 1447.0, "minimum": 1312.0, "sample_count": 58 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 95.08620689655173, "median": 100.0, "minimum": 0.0, "sample_count": 58 }, "host_mem_available_bytes": { "maximum": 96179335168, "mean": 95469332293.81818, "median": 95393185792, "minimum": 95371411456, "sample_count": 55 }, "process_pss_bytes": { "maximum": 44343626752, "mean": 4826962683.345454, "median": 1847659520, "minimum": 1162419200, "sample_count": 55 }, "process_rss_bytes": { "maximum": 44346470400, "mean": 4829806889.890909, "median": 1850503168, "minimum": 1165262848, "sample_count": 55 } }, "process_sample_count": 55, "response_union_seconds": 58.6424638669705, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 8, "storage_read_bytes": 0, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 57.17124111595331, "observed_intervals": 1, "requested_intervals": 1 } } }, { "context_tokens": 8192, "internal_seconds": { "maximum": 7.046474218, "mean": 6.9842637318, "median": 6.971638056, "minimum": 6.947936969, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "prefill", "process_seconds_including_load_and_warmup": 48.933231414004695, "sample_count": 5, "telemetry": { "gpu_samples": 52, "host_samples": 49, "maximum_gpu_gap_seconds": 1.0068673229543492, "maximum_host_gap_seconds": 1.6071946779848076, "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": 52, "unavailable": 0 }, "software": { "active": 0, "not_active": 52, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 1179.0550254774482, "mean": 1172.9528514340318, "median": 1175.046658216819, "minimum": 1162.5672281711882, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 12941.518359739151, "reason": null, "response_union_seconds": 48.933231414004695, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 264.47299689338087, "trapezoid_pieces": 50, "watt_hours": 3.5948662110386533 }, "gpu_sample_count": 49, "host_sample_count": 46, "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": 17365073920, "mean": 17351436733.217392, "median": 17364377600.0, "minimum": 17048735744, "sample_count": 46 }, "gpu_board_watts": { "maximum": 300.11, "mean": 266.7261224489796, "median": 299.97, "minimum": 40.87, "sample_count": 49 }, "gpu_memory_used_bytes": { "maximum": 47013953536.0, "mean": 44788404474.77551, "median": 47013953536.0, "minimum": 2097152.0, "sample_count": 49 }, "gpu_sm_clock_mhz": { "maximum": 2310.0, "mean": 1683.2857142857142, "median": 1567.0, "minimum": 1545.0, "sample_count": 49 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 92.46938775510205, "median": 100.0, "minimum": 0.0, "sample_count": 49 }, "host_mem_available_bytes": { "maximum": 96464650240, "mean": 96078368055.65218, "median": 96067065856.0, "minimum": 96036200448, "sample_count": 46 }, "process_pss_bytes": { "maximum": 44343596032, "mean": 4888073124.9777775, "median": 1166414848, "minimum": 1074246656, "sample_count": 45 }, "process_rss_bytes": { "maximum": 44346441728, "mean": 4890918820.9777775, "median": 1169260544, "minimum": 1077092352, "sample_count": 45 } }, "process_sample_count": 45, "response_union_seconds": 48.933231414004695, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 8, "storage_read_bytes": 0, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 46.782026912027504, "observed_intervals": 1, "requested_intervals": 1 } } }, { "context_tokens": 8192, "internal_seconds": { "maximum": 10.101934588, "mean": 10.0163859246, "median": 10.05206316, "minimum": 9.895870681, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "decode", "process_seconds_including_load_and_warmup": 66.24555310799042, "sample_count": 5, "telemetry": { "gpu_samples": 69, "host_samples": 65, "maximum_gpu_gap_seconds": 1.004757751012221, "maximum_host_gap_seconds": 1.5790673490264453, "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": 69, "unavailable": 0 }, "software": { "active": 0, "not_active": 69, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 25.869376051115758, "mean": 25.559495016860822, "median": 25.467408623007497, "minimum": 25.341680622650255, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 17767.71185915779, "reason": null, "response_union_seconds": 66.24555310799042, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 268.2098801438595, "trapezoid_pieces": 67, "watt_hours": 4.935475516432719 }, "gpu_sample_count": 66, "host_sample_count": 62, "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": 20093181952, "mean": 19573842052.129032, "median": 20075718656.0, "minimum": 17320812544, "sample_count": 62 }, "gpu_board_watts": { "maximum": 305.51, "mean": 270.04015151515154, "median": 299.99, "minimum": 34.33, "sample_count": 66 }, "gpu_memory_used_bytes": { "maximum": 47104131072.0, "mean": 46155392217.21212, "median": 47104131072.0, "minimum": 587202560.0, "sample_count": 66 }, "gpu_sm_clock_mhz": { "maximum": 2310.0, "mean": 1577.2575757575758, "median": 1500.0, "minimum": 1335.0, "sample_count": 66 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 94.46969696969697, "median": 100.0, "minimum": 0.0, "sample_count": 66 }, "host_mem_available_bytes": { "maximum": 96098467840, "mean": 93860503882.32259, "median": 93367040000.0, "minimum": 93340299264, "sample_count": 62 }, "process_pss_bytes": { "maximum": 44343625728, "mean": 6109299546.83871, "median": 3874158592.0, "minimum": 1121785856, "sample_count": 62 }, "process_rss_bytes": { "maximum": 44346482688, "mean": 6112179365.16129, "median": 3877015552.0, "minimum": 1124642816, "sample_count": 62 } }, "process_sample_count": 62, "response_union_seconds": 66.24555310799042, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 8, "storage_read_bytes": 0, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 64.63138355803676, "observed_intervals": 1, "requested_intervals": 1 } } }, { "context_tokens": 32768, "internal_seconds": { "maximum": 34.475327335, "mean": 33.9969558686, "median": 34.054812408, "minimum": 33.375423471, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "prefill", "process_seconds_including_load_and_warmup": 210.42748850595672, "sample_count": 5, "telemetry": { "gpu_samples": 213, "host_samples": 206, "maximum_gpu_gap_seconds": 1.00626217701938, "maximum_host_gap_seconds": 1.6164581719785929, "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": 213, "unavailable": 0 }, "software": { "active": 0, "not_active": 213, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 981.8002767357307, "mean": 963.9831469880572, "median": 962.2134929835142, "minimum": 950.4768346820978, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 61407.47267151111, "reason": null, "response_union_seconds": 210.42748850595672, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 291.82248530126236, "trapezoid_pieces": 211, "watt_hours": 17.057631297641976 }, "gpu_sample_count": 210, "host_sample_count": 203, "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": 17452949504, "mean": 17415677316.41379, "median": 17421238272, "minimum": 17245638656, "sample_count": 203 }, "gpu_board_watts": { "maximum": 300.37, "mean": 292.44866666666667, "median": 299.99, "minimum": 42.58, "sample_count": 210 }, "gpu_memory_used_bytes": { "maximum": 55067017216.0, "mean": 54541500884.11429, "median": 55067017216.0, "minimum": 587202560.0, "sample_count": 210 }, "gpu_sm_clock_mhz": { "maximum": 2302.0, "mean": 1550.7142857142858, "median": 1522.0, "minimum": 1477.0, "sample_count": 210 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 98.72380952380952, "median": 100.0, "minimum": 0.0, "sample_count": 210 }, "host_mem_available_bytes": { "maximum": 96197402624, "mean": 96030243673.53694, "median": 96028155904, "minimum": 95982125056, "sample_count": 203 }, "process_pss_bytes": { "maximum": 44343539712, "mean": 2042630930.9162562, "median": 1222236160, "minimum": 1097951232, "sample_count": 203 }, "process_rss_bytes": { "maximum": 44346355712, "mean": 2045468470.226601, "median": 1225052160, "minimum": 1100767232, "sample_count": 203 } }, "process_sample_count": 203, "response_union_seconds": 210.42748850595672, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 8, "storage_read_bytes": 0, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 208.5821728379815, "observed_intervals": 1, "requested_intervals": 1 } } }, { "context_tokens": 32768, "internal_seconds": { "maximum": 11.465163352, "mean": 11.367803044, "median": 11.382011832, "minimum": 11.26537137, "sample_count": 5 }, "meaning": "Five internal llama-bench measurements; rates recomputed from integer nanosecond durations after verifying tool JSON rounding. No fabricated HTTP or absolute per-sample timestamps.", "mode": "decode", "process_seconds_including_load_and_warmup": 105.98348441597773, "sample_count": 5, "telemetry": { "gpu_samples": 109, "host_samples": 101, "maximum_gpu_gap_seconds": 1.0101722169783898, "maximum_host_gap_seconds": 1.6088830140070058, "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": 109, "unavailable": 0 }, "software": { "active": 0, "not_active": 109, "unavailable": 0 } } }, "tokens_per_second": { "maximum": 22.724506063043354, "mean": 22.520664140441617, "median": 22.49163010710179, "minimum": 22.32850873034818, "sample_count": 5 }, "whole_process_resources": { "gpu_energy": { "available": true, "interpolation": "linear_at_bracketed_boundaries", "joules": 28329.983386864616, "reason": null, "response_union_seconds": 105.98348441597773, "scope": "GPU board only; not whole-server AC energy", "time_weighted_mean_watts": 267.3056424119009, "trapezoid_pieces": 107, "watt_hours": 7.869439829684616 }, "gpu_sample_count": 106, "host_sample_count": 98, "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": 28180320256, "mean": 24010712544.65306, "median": 28179181568.0, "minimum": 17242284032, "sample_count": 98 }, "gpu_board_watts": { "maximum": 300.98, "mean": 268.24283018867925, "median": 299.99, "minimum": 42.7, "sample_count": 106 }, "gpu_memory_used_bytes": { "maximum": 55157194752.0, "mean": 53589039451.77358, "median": 55157194752.0, "minimum": 2097152.0, "sample_count": 106 }, "gpu_sm_clock_mhz": { "maximum": 2340.0, "mean": 1660.6792452830189, "median": 1552.0, "minimum": 1455.0, "sample_count": 106 }, "gpu_utilization_percent": { "maximum": 100.0, "mean": 92.99056603773585, "median": 100.0, "minimum": 0.0, "sample_count": 106 }, "host_mem_available_bytes": { "maximum": 96205852672, "mean": 89495125514.44897, "median": 85380409344.0, "minimum": 85302349824, "sample_count": 98 }, "process_pss_bytes": { "maximum": 44343589888, "mean": 9520023092.244898, "median": 11959954432.0, "minimum": 1031908352, "sample_count": 98 }, "process_rss_bytes": { "maximum": 44346441728, "mean": 9522918337.306122, "median": 11962806272.0, "minimum": 1034760192, "sample_count": 98 } }, "process_sample_count": 98, "response_union_seconds": 105.98348441597773, "sampled_counter_deltas": { "deltas": { "major_faults": 0, "read_syscall_bytes": 8, "storage_read_bytes": 0, "storage_write_bytes": 0 }, "observed_endpoint_seconds": 104.06500446301652, "observed_intervals": 1, "requested_intervals": 1 } } } ], "id": "engine-72b-q4", "meaning": "Engine telemetry includes loading and built-in warmup. No API latency or isolated internal-sample resource claim.", "model": "qwen25-72b-q4", "module": "engine", "record_sha256": "87e302f28f976b0c4a6bd54bd1f71aeeb2e6e0c8fcbda314aa406ad744cf23fe", "reviewed_export_sha256": "59f09a9b924c4cb4ae1370f85109347542e72457c07f0a7a8ff351b31fedab64", "settings": { "batch": 2048, "cache_prompt": false, "cache_ram_bytes": 0, "concurrency": 1, "context_shift": false, "contexts": [ 2048, 8192, 32768 ], "endpoint": null, "fit": "off", "flash_attention": true, "ignore_eos": true, "kv_cache": "f16", "lazy_mode": "off", "load_mode": "mmap", "measured_rounds": 5, "mlock": false, "modes": [ "prefill", "decode" ], "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": 0, "synthetic_prompt_sha256": "c099b5abc034050fac1133e0214990d1f72a124b64f36e8a80ba3209bc07d187", "temperature": 0, "threads": 16, "tokenization": "model_tokenizer_add_special_false_repeat_then_truncate", "total_context_tokens": 0, "ubatch": 512, "warmup": "one_tool_warmup_per_process", "warmup_rounds": 1 }, "status": "completed" } ], "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": "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" } ], "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": "2026-09-27T15:11:44.063808+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" } ], "status": "completed", "test_date_timezone": "Europe/Warsaw", "tested_at": "2026-09-27" }