OpenEvals - SWE-bench Verified: leaderboard
Hugging Face OpenEvals official leaderboard covering SWE-bench Verified software issue resolution.
Metric: Resolved (%). Source: huggingface.co. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 76.4 | #146 |
| 2 | MiniMax-M2.5 | 75.8 | #303 |
| 3 | Step 3.5 Flash | 74.4 | #270 |
| 4 | GLM-5 | 72.8 | #141 |
| 5 | Qwen 3.5 27B | 72.4 | #187 |
| 6 | Qwen 3.5 122B A10B | 72 | #173 |
| 7 | Kimi K2.5 | 70.8 | #144 |
| 8 | Qwen3 Coder Next | 70.6 | #326 |
| 9 | DeepSeek V3.2 | 70 | #205 |
| 10 | Qwen 3.5 35B A3B | 69.2 | #256 |
| 11 | GPT-OSS-120B | 47.9 | #341 |
| 12 | GPT-OSS-20B | 37.4 | #518 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openevals-swe-bench-verified · How It Works · Data refreshed daily, snapshot 2026-10-20.