{
  "items": [
    {
      "label": {
        "en": "Qwen3-8B Q4_K_M",
        "pl": "Qwen3-8B Q4_K_M"
      },
      "value": 6.2890625
    },
    {
      "label": {
        "en": "Qwen3-32B Q4_K_M",
        "pl": "Qwen3-32B Q4_K_M"
      },
      "value": 20.95703125
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q4_K_M",
        "pl": "Qwen2.5-72B-Instruct Q4_K_M"
      },
      "value": 43.869140625
    },
    {
      "label": {
        "en": "GPT-OSS-120B MXFP4",
        "pl": "GPT-OSS-120B MXFP4"
      },
      "value": 59.55078125
    },
    {
      "label": {
        "en": "Qwen2.5-72B-Instruct Q8_0",
        "pl": "Qwen2.5-72B-Instruct Q8_0"
      },
      "value": 74.505859375
    },
    {
      "label": {
        "en": "Qwen3-235B-A22B Q5_K_M",
        "pl": "Qwen3-235B-A22B Q5_K_M"
      },
      "value": 87.69140625
    }
  ],
  "note": {
    "en": "Observed peak across response windows for 2048 and 8192 input tokens. The 235B model also uses system RAM; values are not minimum requirements.",
    "pl": "Szczyt zaobserwowany w oknach odpowiedzi dla wejść 2048 i 8192 tokeny. Model 235B korzysta także z RAM; wartości nie oznaczają minimalnych wymagań."
  },
  "title": {
    "en": "GPU memory during responses",
    "pl": "Pamięć GPU podczas odpowiedzi"
  },
  "unit": "GiB"
}