{
  "generated": "2026-09-08",
  "benchmark": "arc-agi-3",
  "results": [
    {
      "system": "Opus 4.6 (Max)",
      "developer": "Anthropic",
      "value": 0.5,
      "date": "2026-03-24",
      "source": {
        "url": "https://arxiv.org/abs/2603.24621",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "semi-private",
        "reasoning_effort": "max",
        "notes": "Table 2 of the paper: semi-private leaderboard scores for frontier models at release."
      }
    },
    {
      "system": "Gemini 3.1 Pro Preview",
      "developer": "Google DeepMind",
      "value": 0.4,
      "date": "2026-03-24",
      "source": {
        "url": "https://arxiv.org/abs/2603.24621",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "semi-private",
        "notes": "Table 2 of the paper: semi-private leaderboard scores at release."
      }
    },
    {
      "system": "GPT 5.4 (High)",
      "developer": "OpenAI",
      "value": 0.2,
      "date": "2026-03-24",
      "source": {
        "url": "https://arxiv.org/abs/2603.24621",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "semi-private",
        "reasoning_effort": "high",
        "notes": "Table 2 of the paper: semi-private leaderboard scores at release."
      }
    },
    {
      "system": "GPT-6 Astra (max)",
      "developer": "OpenAI",
      "value": 62.7,
      "date": "2026-09-03",
      "source": {
        "url": "https://arcprize.org/blog/astra",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "semi-private",
        "scaffold": "Standard harness",
        "reasoning_effort": "max",
        "cost_usd_per_task": 474.51,
        "notes": "ARC Prize verified run; $26,098 total across 55 semi-private environments (cost per environment shown). Best Standard-harness result."
      }
    },
    {
      "system": "GPT-6 Astra (high)",
      "developer": "OpenAI",
      "value": 99.9,
      "date": "2026-09-03",
      "source": {
        "url": "https://arcprize.org/blog/astra",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "semi-private",
        "scaffold": "Provider Adapter",
        "reasoning_effort": "high",
        "cost_usd_per_task": 342.13,
        "notes": "ARC Prize verified run; $18,817 total across 55 semi-private environments (cost per environment shown). Best Provider Adapter result; verified table shows 99.95%."
      }
    }
  ]
}