SWE-rebench V2 - Python: leaderboard
Metric: pass@1 (%, 0-100) on the 60 Python tasks, averaged over three runs; the SWE-rebench V2 diagnostic sample: 300 issue-resolution tasks drawn at random from the collection, 60 each in Python, JavaScript, Go, Rust and Scala, with pre-built Docker environments and fail-to-pass tests; mini-SWE-agent with default generation settings, three independent runs per task; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 36.1 | #79 |
| 2 | GLM-4.7 | 27.2 | #185 |
| 3 | MiniMax-M2.1 | 26.1 | #283 |
| 4 | Gemini 3 Flash | 25.6 | #93 |
| 5 | DeepSeek V3.2 | 23.3 | #198 |
| 6 | GPT-5.2 (Medium) | 20.6 | #105 (GPT-5.2) |
| 7 | GPT-OSS-120B | 8.9 | #330 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=swe-rebench-v2-python · How It Works · Data refreshed daily, snapshot 2026-10-11.