{
  "retrievedAt": "2026-11-05",
  "origin": "published third-party evaluations; not a test of this service",
  "artificialAnalysis": {
    "url": "https://artificialanalysis.ai/models/releases/comparisons/gpt-6-1-sol-vs-claude-opus-5-5",
    "indexVersion": "5.0.1",
    "models": [
      {
        "id": "gpt-6.1-sol",
        "label": "GPT-6.1 Sol",
        "configuration": "Max"
      },
      {
        "id": "claude-opus-5-5",
        "label": "Claude Opus 5.5",
        "configuration": "Max, Default Fallback"
      }
    ],
    "metrics": [
      {
        "benchmark": "Intelligence Index v5.0.1",
        "unit": "index points",
        "values": [
          55,
          59
        ],
        "sourcePrecision": "preserved as displayed"
      },
      {
        "benchmark": "Terminal-Bench 4.0",
        "unit": "percent",
        "values": [
          56.1,
          59.6
        ],
        "sourcePrecision": "preserved as displayed"
      },
      {
        "benchmark": "FrontierCode 2.0",
        "unit": "percent",
        "values": [
          40.6,
          46.4
        ],
        "sourcePrecision": "preserved as displayed"
      },
      {
        "benchmark": "GDP.pdf",
        "unit": "percent",
        "values": [
          31.0,
          26.2
        ],
        "sourcePrecision": "preserved as displayed"
      }
    ],
    "confidenceIntervals": null,
    "note": "Comparison page does not publish row-level confidence intervals or exact run timestamps. Claude Default Fallback is part of the source configuration.",
    "methodology": "https://artificialanalysis.ai/methodology/intelligence-benchmarking",
    "terminalBench": {
      "version": "4.0",
      "tasks": 66,
      "repeats": 3,
      "harness": "mini-swe-agent",
      "scoring": "pass@1; all verifier tests must pass"
    },
    "frontierCode": {
      "version": "2.0",
      "tasks": 132,
      "repeats": 3,
      "harness": "Stirrup",
      "scoring": "held-out tests and rubric; LLM judge",
      "judge": "GPT-6 Luna (medium)",
      "turnCap": 500
    },
    "indexCategories": {
      "Knowledge Work": 0.25,
      "Coding": 0.25,
      "Science": 0.25,
      "General": 0.25
    },
    "gdpPdf": {
      "tasks": 100,
      "repeats": 5,
      "scoring": "All-pass over rubric checks; LLM judge",
      "methodology": "https://artificialanalysis.ai/methodology/intelligence-benchmarking"
    }
  },
  "agentArena": {
    "url": "https://arena.ai/leaderboard/agent",
    "methodology": "https://arena.ai/blog/agent-arena-methodology",
    "sourceUpdatedAt": "2026-11-04",
    "configuration": "Max for each named model",
    "metric": "Net Improvement",
    "unit": "percentage-point treatment-effect estimate; not raw completion rate",
    "confidenceLevel": 0.95,
    "models": [
      {
        "id": "gpt-6.1-sol",
        "netImprovement": 12.51,
        "ci95HalfWidth": 1.11,
        "sessions": 30285
      },
      {
        "id": "claude-opus-5-5",
        "netImprovement": 20.14,
        "ci95HalfWidth": 2.09,
        "sessions": 14169
      }
    ]
  }
}