{
  "generated": "2026-09-08",
  "benchmark": "mmlu-pro",
  "results": [
    {
      "system": "GPT-4o",
      "developer": "OpenAI",
      "value": 72.6,
      "date": "2024-06-03",
      "source": {
        "url": "https://arxiv.org/abs/2406.01574",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "shots": 5,
        "notes": "Table of the MMLU-Pro paper, CoT prompting."
      }
    },
    {
      "system": "DeepSeek-V4-Pro (Think Max)",
      "developer": "DeepSeek",
      "value": 87.5,
      "date": "2026-04-22",
      "source": {
        "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro",
        "kind": "developer-report",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Model card comparison table; the same table lists Gemini-3.1-Pro (High) at 91.0 and Claude Opus 4.6 (Max) at 89.1 as run by DeepSeek."
      }
    },
    {
      "system": "Gemini 3.1 Pro (High)",
      "developer": "Google DeepMind",
      "value": 91.0,
      "date": "2026-04-22",
      "source": {
        "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro",
        "kind": "independent-evaluation",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Run by DeepSeek for its V4-Pro model card comparison table; setting not further specified."
      }
    },
    {
      "system": "Qwen3.7 Max",
      "developer": "Alibaba",
      "value": 89.6,
      "date": "2026-05-16",
      "source": {
        "url": "https://benchlm.ai/benchmarks/mmlu-pro",
        "kind": "aggregator",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Developer-reported number republished by aggregator; prompt/effort setting unspecified. Date is the model release date shown by benchlm.ai."
      }
    }
  ]
}