{
  "data": {
    "a": {
      "id": "gpt-5",
      "name": "GPT-5",
      "providerId": "openai",
      "releaseDate": "2025-08-07",
      "contextWindow": 400000,
      "maxOutput": 128000,
      "modalities": [
        "text",
        "vision"
      ],
      "pricing": {
        "inputPerMTokens": 1.25,
        "outputPerMTokens": 10,
        "cachedInputPerMTokens": 0.125,
        "sourceUrl": "https://developers.openai.com/api/docs/models/gpt-5",
        "updatedAt": "2026-09-16"
      },
      "benchmarks": [
        {
          "benchmark": "AIME 2025 (MathArena)",
          "score": 95,
          "unit": "percent",
          "sourceUrl": "https://matharena.ai/?comp=aime--aime_2025",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "FrontierMath Tier 4 v2 (Epoch AI run)",
          "score": 22,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/frontiermath",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "FrontierMath Tiers 1-3 v2 (Epoch AI run)",
          "score": 55.4,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/frontiermath",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "GPQA Diamond (Epoch AI run)",
          "score": 86.2,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/gpqa-diamond",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "OTIS Mock AIME 2024-2025 (Epoch AI run)",
          "score": 91.4,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/otis-mock-aime-2024-2025",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "SimpleQA Verified (Epoch AI run)",
          "score": 50.1,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "SWE-bench Verified (Epoch AI run)",
          "score": 73.6,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/swe-bench-verified",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "AIME 2025 (no tools)",
          "score": 94.6,
          "unit": "percent",
          "sourceUrl": "https://openai.com/index/introducing-gpt-5/",
          "measuredAt": "2025-08-07"
        },
        {
          "benchmark": "SWE-bench Verified",
          "score": 74.9,
          "unit": "percent",
          "sourceUrl": "https://openai.com/index/introducing-gpt-5/",
          "measuredAt": "2025-08-07"
        }
      ]
    },
    "b": {
      "id": "claude-opus-4",
      "name": "Claude Opus 4",
      "providerId": "anthropic",
      "releaseDate": "2025-05-22",
      "contextWindow": 200000,
      "maxOutput": 32000,
      "modalities": [
        "text",
        "vision"
      ],
      "pricing": {
        "inputPerMTokens": 15,
        "outputPerMTokens": 75,
        "cachedInputPerMTokens": 1.5,
        "sourceUrl": "https://platform.claude.com/docs/en/about-claude/pricing",
        "updatedAt": "2026-09-16"
      },
      "benchmarks": [
        {
          "benchmark": "GPQA Diamond (Epoch AI run)",
          "score": 76.3,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/gpqa-diamond",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "OTIS Mock AIME 2024-2025 (Epoch AI run)",
          "score": 64.4,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/otis-mock-aime-2024-2025",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "SWE-bench Verified (Epoch AI run)",
          "score": 70.7,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/swe-bench-verified",
          "measuredAt": "2026-09-17"
        },
        {
          "benchmark": "AIME 2024 (no extended thinking)",
          "score": 33.9,
          "unit": "percent",
          "sourceUrl": "https://www.anthropic.com/news/claude-4",
          "measuredAt": "2025-05-22"
        },
        {
          "benchmark": "SWE-bench Verified",
          "score": 72.5,
          "unit": "percent",
          "sourceUrl": "https://www.anthropic.com/news/claude-4",
          "measuredAt": "2025-05-22"
        },
        {
          "benchmark": "Terminal-bench",
          "score": 43.2,
          "unit": "percent",
          "sourceUrl": "https://www.anthropic.com/news/claude-4",
          "measuredAt": "2025-05-22"
        }
      ]
    },
    "shared": [
      {
        "benchmark": "GPQA Diamond",
        "a": {
          "score": 86.2,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/gpqa-diamond"
        },
        "b": {
          "score": 76.3,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/gpqa-diamond"
        },
        "winner": "gpt-5"
      },
      {
        "benchmark": "OTIS Mock AIME 2024-2025 (Epoch AI run)",
        "a": {
          "score": 91.4,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/otis-mock-aime-2024-2025"
        },
        "b": {
          "score": 64.4,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/otis-mock-aime-2024-2025"
        },
        "winner": "gpt-5"
      },
      {
        "benchmark": "SWE-bench Verified (Epoch AI run)",
        "a": {
          "score": 73.6,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/swe-bench-verified"
        },
        "b": {
          "score": 70.7,
          "unit": "percent",
          "sourceUrl": "https://epoch.ai/benchmarks/swe-bench-verified"
        },
        "winner": "gpt-5"
      },
      {
        "benchmark": "SWE-bench Verified",
        "a": {
          "score": 74.9,
          "unit": "percent",
          "sourceUrl": "https://openai.com/index/introducing-gpt-5/"
        },
        "b": {
          "score": 72.5,
          "unit": "percent",
          "sourceUrl": "https://www.anthropic.com/news/claude-4"
        },
        "winner": "gpt-5"
      }
    ]
  },
  "meta": {
    "source": "https://llmmetric.com",
    "attribution": "Please cite llmmetric.com. Upstream records remain subject to their source terms.",
    "generatedAt": "2026-09-18T05:31:52.869Z"
  }
}