SWE-bench Verified: leaderboard
The 'Verified' subset of SWE-bench - 500 human-validated GitHub issues ensuring correct problem statements and test coverage. The most trusted variant for measuring real-world software engineering ability.
Metric: Resolved (%). Source: huggingface.co. Source: www.swebench.com. Saturation forecast: Around January 2027. 47 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 (High) | 76.8 | #83 (Claude Opus 4.5) |
| 2 | Gemini 3 Flash (High) | 75.8 | #99 (Gemini 3 Flash) |
| 3 | Claude Opus 4.6 | 75.6 | #63 |
| 4 | GPT-5.2 (High) | 72.8 | #110 (GPT-5.2) |
| 5 | GPT-5.2 Codex | 72.8 | #95 |
| 6 | GPT-5.2 (2025-12-11) (High) | 71.8 | |
| 7 | Claude Sonnet 4.5 (High) | 71.4 | #142 (Claude Sonnet 4.5) |
| 8 | Kimi K2.5 (High) | 70.8 | #144 (Kimi K2.5) |
| 9 | Claude Sonnet 4.5 | 70.6 | #142 |
| 10 | DeepSeek V3.2 (High) | 70 | #205 (DeepSeek V3.2) |
| 11 | Gemini 3 Pro | 69.6 | #80 |
| 12 | GPT-5.2 | 69 | #110 |
| 13 | Claude Opus 4 (20250514) | 67.6 | #154 |
| 14 | Claude Haiku 4.5 (High) | 66.6 | #280 (Claude Haiku 4.5) |
| 15 | GPT-5 (Medium) | 65 | #98 (GPT-5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=swe-bench-verified · How It Works · Data refreshed daily, snapshot 2026-10-20.