{
  "generated": "2026-09-08",
  "benchmark": "paperbench",
  "results": [
    {
      "system": "BasicAgent claude-3.5-sonnet",
      "developer": "Anthropic",
      "value": 21.0,
      "date": "2025-04-02",
      "source": {
        "url": "https://github.com/openai/frontier-evals/tree/main/project/paperbench",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "BasicAgent",
        "notes": "Repository leaderboard, 3 runs, +/-0.8."
      }
    },
    {
      "system": "IterativeAgent o1-high (24h limit)",
      "developer": "OpenAI",
      "value": 24.4,
      "date": "2025-04-02",
      "source": {
        "url": "https://github.com/openai/frontier-evals/tree/main/project/paperbench",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "IterativeAgent",
        "reasoning_effort": "high",
        "notes": "Repository leaderboard, 3 runs, +/-0.7, 24-hour limit."
      }
    },
    {
      "system": "IterativeAgent o1-high (36h limit)",
      "developer": "OpenAI",
      "value": 26.0,
      "date": "2025-04-02",
      "source": {
        "url": "https://github.com/openai/frontier-evals/tree/main/project/paperbench",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "IterativeAgent",
        "reasoning_effort": "high",
        "notes": "Top of the repository leaderboard; 3 runs, +/-0.3, 36-hour limit."
      }
    },
    {
      "system": "IterativeAgent o1-high",
      "developer": "OpenAI",
      "value": 43.4,
      "date": "2025-04-02",
      "source": {
        "url": "https://github.com/openai/frontier-evals/tree/main/project/paperbench",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "code-dev",
        "scaffold": "IterativeAgent",
        "reasoning_effort": "high",
        "notes": "PaperBench Code-Dev leaderboard, 3 runs, +/-0.8."
      }
    }
  ]
}