SWE-Bench Verified
Unknown
500 manually verified GitHub issues confirmed solvable by human engineers. High-quality subset of SWE-bench.
Benchmark Stats
Models29
Papers29
Metrics1
SOTA History
Not enough data to show trend.
resolve-rate
resolve-rate
Higher is better
| Rank | Model | Source | Score | Year | Paper |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.5 First model to break 80%. With agent scaffolding. | Editorial | 80.9 | 2026 | Source |
| 2 | Claude Opus 4.6 Simple bash + file edit scaffold. | Editorial | 80.8 | 2026 | Source |
| 3 | Gemini 3.1 Pro With custom agent scaffold. | Editorial | 80.6 | 2026 | Source |
| 4 | GPT-5.2 Thinking GPT-5.2 with thinking mode and agent scaffold. | Editorial | 80 | 2026 | Source |
| 5 | Claude Sonnet 4.6 Simple bash + file edit scaffold. | Editorial | 79.6 | 2026 | Source |
| 6 | Gemini 3 Flash With custom agent scaffold. | Editorial | 78 | 2026 | Source |
| 7 | claude-sonnet-45 | Editorial | 77.2 | 2026 | Source |
| 8 | GPT-5.1 GPT-5.1 with agent scaffold. | Editorial | 76.3 | 2026 | Source |
| 9 | Gemini 3 Pro With custom agent scaffold. | Editorial | 76.2 | 2026 | Source |
| 10 | GPT-5 GPT-5 with agent scaffold. | Editorial | 74.9 | 2026 | Source |
| 11 | Claude Haiku 4.5 128K thinking budget. Simple scaffold. | Editorial | 73.3 | 2026 | Source |
| 12 | claude-sonnet-4 | Editorial | 72.2 | 2026 | Source |
| 13 | Devstral 2 Devstral 2 with SWE-agent. | Editorial | 72.2 | 2026 | Source |
| 14 | o3 | Editorial | 69.1 | 2026 | Source |
| 15 | o4-mini With agent scaffolding. | Editorial | 68.1 | 2026 | Source |
| 16 | Grok 3 With custom agent scaffold. | Editorial | 63.8 | 2026 | Source |
| 17 | gemini-25-pro | Editorial | 63.8 | 2026 | Source |
| 18 | claude-37-sonnet | Editorial | 62.3 | 2026 | Source |
| 19 | DeepSeek R1-0528 Self-reported score with Agentless framework. | Editorial | 57.6 | 2026 | Source |
| 20 | o3-mini | Editorial | 55.8 | 2026 | Source |
| 21 | gpt-41 | Editorial | 54.6 | 2026 | Source |
| 22 | Gemini 2.5 Flash With custom agent scaffold. | Editorial | 54 | 2026 | Source |
| 23 | deepseek-r1 | Editorial | 49.2 | 2026 | Source |
| 24 | claude-35-sonnet | Editorial | 49 | 2026 | Source |
| 25 | o1 | Editorial | 48.9 | 2026 | Source |
| 26 | deepseek-v3 | Editorial | 42 | 2026 | Source |
| 27 | gpt-4o | Editorial | 41.2 | 2026 | Source |
| 28 | Claude 3.5 Haiku Standard scaffold. | Editorial | 40.6 | 2026 | Source |
| 29 | deepseek-v25 | Editorial | 37 | 2026 | Source |