{
  "source": {
    "name": "Epoch AI",
    "kind": "independent internal evaluations",
    "url": "https://epoch.ai/benchmarks/use-this-data",
    "methodology": "https://epoch.ai/benchmarks/about",
    "download": "https://epoch.ai/data/benchmark_data.zip",
    "license": "CC BY 4.0",
    "localLicense": "/assets/licenses/epoch-benchmarks.txt",
    "evidence": "/data/independent-model-benchmark-evidence-2026-10-19.json"
  },
  "retrievedOn": "2026-10-19",
  "selection": "Six selected current configurations with results in all three internal-run datasets. Latest dated run for each exact configuration; never the highest score. Not an exhaustive model leaderboard.",
  "datasets": [
    {
      "file": "gpqa_diamond.csv",
      "sha256": "46fc3c5b7e94e748031cfcb4d54ee43eeae4e9020aeecc62634a019dbc4744a1",
      "rows": [
        {
          "Model version": "gemini-4-argon_high",
          "mean_score": "0.9640151515151515",
          "Best score (across scorers)": "0.9640151515151515",
          "Release date": "2026-09-30",
          "Organization": "Google DeepMind",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.011918129589602694",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-13T17:15:08.000Z",
          "id": "QEouTvczNsxgkR4i7i8wBF"
        },
        {
          "Model version": "grok-4.7_xhigh",
          "mean_score": "0.92739898989899",
          "Best score (across scorers)": "0.92739898989899",
          "Release date": "2026-09-21",
          "Organization": "xAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.015505824295534454",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T21:49:22.000Z",
          "id": "hxAhYJfXdPNanxatXpG5hC"
        },
        {
          "Model version": "gpt-6.1-sol_max",
          "mean_score": "0.9539141414141414",
          "Best score (across scorers)": "0.9539141414141414",
          "Release date": "2026-09-29",
          "Organization": "OpenAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.013780748450364692",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-29T22:00:34.000Z",
          "id": "LspviLnbeZ82WCngWp6rwR"
        },
        {
          "Model version": "claude-opus-5-5_max",
          "mean_score": "0.9059343434343434",
          "Best score (across scorers)": "0.9059343434343434",
          "Release date": "2026-09-22",
          "Organization": "Anthropic",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.019434397190192135",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T18:27:39.000Z",
          "id": "6gmbahPSxHf9xVhAYqpnhi"
        },
        {
          "Model version": "deepseek-v4.1-flash_max",
          "mean_score": "0.8983585858585859",
          "Best score (across scorers)": "0.8983585858585859",
          "Release date": "2026-09-09",
          "Organization": "DeepSeek",
          "Country": "China",
          "Training compute (FLOP)": "4.32e+24",
          "Training compute notes": "(Pre-training) - 6 * 16e9 active parameters * 45e12 tokens = 4.32e+24 FLOP\n",
          "stderr": "0.017287759372442393",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-09T14:38:01.000Z",
          "id": "ZU5ebpDaDPiDZLgv7cG8B9"
        },
        {
          "Model version": "glm-5.3_max",
          "mean_score": "0.9090909090909091",
          "Best score (across scorers)": "0.9090909090909091",
          "Release date": "2026-08-14",
          "Organization": "Z.ai (Zhipu AI)",
          "Country": "China",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.015851578202142853",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-08-24T19:44:18.000Z",
          "id": "dLDBahYBTGRNjbsXA5pyr3"
        }
      ]
    },
    {
      "file": "otis_mock_aime_2024_2025.csv",
      "sha256": "5b411612f91e155dff4b09d09b548b42f7936482ea8b3c00b510bd46e2f6bc2c",
      "rows": [
        {
          "Model version": "gemini-4-argon_high",
          "mean_score": "1.0",
          "Best score (across scorers)": "1.0",
          "Release date": "2026-09-30",
          "Organization": "Google DeepMind",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.0",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-13T17:14:57.000Z",
          "id": "43f36RTDoJGxXWQDVvTLJj"
        },
        {
          "Model version": "grok-4.7_xhigh",
          "mean_score": "0.9805555555555555",
          "Best score (across scorers)": "0.9805555555555555",
          "Release date": "2026-09-21",
          "Organization": "xAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.014299486306926387",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T21:49:22.000Z",
          "id": "Jhgwm7QpspGcvm7DUVqUDg"
        },
        {
          "Model version": "gpt-6.1-sol_max",
          "mean_score": "1.0",
          "Best score (across scorers)": "1.0",
          "Release date": "2026-09-29",
          "Organization": "OpenAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.0",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-29T22:00:34.000Z",
          "id": "Di7y6sES4PGQbXNkXYhdqM"
        },
        {
          "Model version": "claude-opus-5-5_max",
          "mean_score": "1.0",
          "Best score (across scorers)": "1.0",
          "Release date": "2026-09-22",
          "Organization": "Anthropic",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.0",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T18:27:39.000Z",
          "id": "LSfona7P3jCAnZFXJ2YA7F"
        },
        {
          "Model version": "deepseek-v4.1-flash_max",
          "mean_score": "0.9833333333333333",
          "Best score (across scorers)": "0.9833333333333333",
          "Release date": "2026-09-09",
          "Organization": "DeepSeek",
          "Country": "China",
          "Training compute (FLOP)": "4.32e+24",
          "Training compute notes": "(Pre-training) - 6 * 16e9 active parameters * 45e12 tokens = 4.32e+24 FLOP\n",
          "stderr": "0.006405884289881728",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-08T19:02:32.000Z",
          "id": "7eHNiVQCzLcAyj5LxmeeSU"
        },
        {
          "Model version": "glm-5.3_max",
          "mean_score": "0.9111111111111111",
          "Best score (across scorers)": "0.9111111111111111",
          "Release date": "2026-08-14",
          "Organization": "Z.ai (Zhipu AI)",
          "Country": "China",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.03454754040013868",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-08-24T19:44:18.000Z",
          "id": "goZR5yGxZic9TZ23ezCL3w"
        }
      ]
    },
    {
      "file": "simpleqa_verified.csv",
      "sha256": "d4ef3a7d6ccc9c611a30e98d0561021e903c0f4a224851311a7c420862154dd8",
      "rows": [
        {
          "Model version": "gemini-4-argon_high",
          "mean_score": "0.76",
          "Best score (across scorers)": "0.76",
          "Release date": "2026-09-30",
          "Organization": "Google DeepMind",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.013512312258920845",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-13T17:15:14.000Z",
          "id": "iq4Y9tXpiyLP25yotySxJW"
        },
        {
          "Model version": "grok-4.7_xhigh",
          "mean_score": "0.56",
          "Best score (across scorers)": "0.56",
          "Release date": "2026-09-21",
          "Organization": "xAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.015704987954361718",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T21:49:22.000Z",
          "id": "6izDCw5ygCLZJEKPS8jV5T"
        },
        {
          "Model version": "gpt-6.1-sol_max",
          "mean_score": "0.739",
          "Best score (across scorers)": "0.739",
          "Release date": "2026-09-29",
          "Organization": "OpenAI",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.013895037677965185",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-29T22:00:34.000Z",
          "id": "dfT5VD7cQug8XW6GQHyJUF"
        },
        {
          "Model version": "claude-opus-5-5_max",
          "mean_score": "0.722",
          "Best score (across scorers)": "0.722",
          "Release date": "2026-09-22",
          "Organization": "Anthropic",
          "Country": "United States of America",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.014174516461485273",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-09-22T18:27:39.000Z",
          "id": "CgLfiGcNwMPVzxr8GdNUAK"
        },
        {
          "Model version": "deepseek-v4.1-flash_max",
          "mean_score": "0.429",
          "Best score (across scorers)": "0.429",
          "Release date": "2026-09-09",
          "Organization": "DeepSeek",
          "Country": "China",
          "Training compute (FLOP)": "4.32e+24",
          "Training compute notes": "(Pre-training) - 6 * 16e9 active parameters * 45e12 tokens = 4.32e+24 FLOP\n",
          "stderr": "0.015658997547870177",
          "Log viewer": "",
          "Logs": "",
          "Started at": "2026-10-13T12:59:33.000Z",
          "id": "3JJ4ChNxMo8CGAh3HrKNrw"
        },
        {
          "Model version": "glm-5.3_max",
          "mean_score": "0.41",
          "Best score (across scorers)": "0.41",
          "Release date": "2026-08-14",
          "Organization": "Z.ai (Zhipu AI)",
          "Country": "China",
          "Training compute (FLOP)": "",
          "Training compute notes": "",
          "stderr": "0.015560917136921659",
          "Log viewer": "https://logs.epoch.ai/inspect-viewer/36231d6d/viewer.html?log_file=https%3A%2F%2Flogs.epoch.ai%2Finspect_ai_logs%2FTY8R8FV7zCwByEipkqyg2P.eval",
          "Logs": "https://epoch-benchmarks-production-public.s3.us-east-2.amazonaws.com/inspect_ai_logs/TY8R8FV7zCwByEipkqyg2P.eval",
          "Started at": "2026-08-28T17:00:06.000Z",
          "id": "TY8R8FV7zCwByEipkqyg2P"
        }
      ]
    }
  ]
}