{
  "generated": "2026-09-08",
  "benchmark": "swe-bench",
  "results": [
    {
      "system": "RAG + Claude 2",
      "developer": "Anthropic",
      "value": 1.96,
      "date": "2023-10-10",
      "source": {
        "url": "https://www.swebench.com/#test",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "RAG baseline",
        "notes": "Paper baseline (BM25 retrieval, no agent), checked by the SWE-bench team."
      }
    },
    {
      "system": "SWE-agent + GPT 4 (1106)",
      "developer": "OpenAI",
      "value": 12.47,
      "date": "2024-04-02",
      "source": {
        "url": "https://www.swebench.com/#test",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "SWE-agent"
      }
    },
    {
      "system": "OpenHands + CodeAct v2.1 (claude-3-5-sonnet-20241022)",
      "developer": "Anthropic",
      "value": 29.38,
      "date": "2024-11-03",
      "source": {
        "url": "https://www.swebench.com/#test",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "OpenHands CodeAct v2.1"
      }
    },
    {
      "system": "SWE-agent 1.0 (Claude 3.7 Sonnet)",
      "developer": "Anthropic",
      "value": 33.83,
      "date": "2025-02-27",
      "source": {
        "url": "https://www.swebench.com/#test",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "SWE-agent 1.0"
      }
    },
    {
      "system": "Sonar Foundation Agent + Claude 4.5 Opus",
      "developer": "Anthropic",
      "value": 52.62,
      "date": "2025-12-19",
      "source": {
        "url": "https://www.swebench.com/#test",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "split": "full",
        "scaffold": "Sonar Foundation Agent",
        "notes": "Top of the Full leaderboard; submission not run or checked by the SWE-bench team."
      }
    }
  ]
}
