{
  "items": [
    {
      "label": {
        "en": "Qwen3-8B Q4_K_M — Model on GPU",
        "pl": "Qwen3-8B Q4_K_M — Model na GPU"
      },
      "value": 0.7629299163818359
    },
    {
      "label": {
        "en": "Qwen3-32B Q4_K_M — Model on GPU",
        "pl": "Qwen3-32B Q4_K_M — Model na GPU"
      },
      "value": 0.8615055084228516
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q4_K_M — Model on GPU",
        "pl": "Qwen2.5-72B-Instruct Q4_K_M — Model na GPU"
      },
      "value": 1.1709308624267578
    },
    {
      "label": {
        "en": "GPT-OSS-120B MXFP4 — Model on GPU",
        "pl": "GPT-OSS-120B MXFP4 — Model na GPU"
      },
      "value": 1.260390281677246
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q8_0 — Model on GPU",
        "pl": "Qwen2.5-72B-Instruct Q8_0 — Model na GPU"
      },
      "value": 1.7312097549438477
    },
    {
      "label": {
        "en": "Qwen3-235B-A22B Q5_K_M — GPU + RAM",
        "pl": "Qwen3-235B-A22B Q5_K_M — GPU + RAM"
      },
      "value": 51.5389347076416
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q4_K_M — GPU + CPU: lower VRAM use",
        "pl": "Qwen2.5-72B-Instruct Q4_K_M — GPU + CPU: mniejsze zajęcie VRAM"
      },
      "value": 22.68654727935791
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q4_K_M — GPU + CPU: further reduced VRAM use",
        "pl": "Qwen2.5-72B-Instruct Q4_K_M — GPU + CPU: jeszcze mniejsze zajęcie VRAM"
      },
      "value": 33.01048564910889
    }
  ],
  "note": {
    "en": "Peak PSS: proportionally attributed process memory, across response windows for 2048 and 8192 input tokens. Excludes loading; not minimum required RAM.",
    "pl": "Szczyt PSS: proporcjonalnie przypisanej pamięci procesu, w oknach odpowiedzi dla wejść 2048 i 8192 tokeny. Bez fazy ładowania; nie minimalna pojemność RAM."
  },
  "title": {
    "en": "Process RAM during responses",
    "pl": "RAM procesu podczas odpowiedzi"
  },
  "unit": "GiB"
}