{
  "generated": "2026-09-08",
  "benchmark": "arc-agi-1",
  "results": [
    {
      "system": "GPT-4o",
      "developer": "OpenAI",
      "value": 4.5,
      "date": "2024-12-20",
      "source": {
        "url": "https://arcprize.org/blog/oai-o3-pub-breakthrough",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "notes": "ARC Prize testing; the blog states GPT-4o reached 5% in 2024 and the leaderboard lists 4.5%."
      }
    },
    {
      "system": "o3-preview (high efficiency)",
      "developer": "OpenAI",
      "value": 75.7,
      "date": "2024-12-20",
      "source": {
        "url": "https://arcprize.org/blog/oai-o3-pub-breakthrough",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "pass_k": 2,
        "cost_usd_per_task": 26,
        "notes": "ARC Prize verified; the 172x-compute configuration scored 87.5% at about $4,560 per task. Model was trained on 75% of the public training set."
      }
    },
    {
      "system": "o3 (High)",
      "developer": "OpenAI",
      "value": 60.8,
      "date": "2025-04-16",
      "source": {
        "url": "https://arcprize.org/leaderboard",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "pass_k": 2,
        "cost_usd_per_task": 0.5,
        "notes": "Publicly released o3, ARC Prize leaderboard; lower than the December 2024 preview."
      }
    },
    {
      "system": "GPT-5.2 (XHigh)",
      "developer": "OpenAI",
      "value": 86.2,
      "date": "2025-12-11",
      "source": {
        "url": "https://arcprize.org/leaderboard",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "pass_k": 2,
        "cost_usd_per_task": 0.96
      }
    },
    {
      "system": "Gemini 3 Deep Think (2/26)",
      "developer": "Google DeepMind",
      "value": 96.0,
      "date": "2026-02-12",
      "source": {
        "url": "https://arcprize.org/leaderboard",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "pass_k": 2,
        "cost_usd_per_task": 7.17
      }
    },
    {
      "system": "GPT-6 Astra (XHigh)",
      "developer": "OpenAI",
      "value": 98.5,
      "date": "2026-09-02",
      "source": {
        "url": "https://arcprize.org/leaderboard",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "Semi-Private",
        "pass_k": 2,
        "cost_usd_per_task": 0.35,
        "notes": "Ties Claude Fable 5 (Max/XHigh, 98.5%, June 2026) at far lower cost; matches the 98% human panel."
      }
    }
  ]
}