SWE-Marathon: leaderboard
Archived SWE-Marathon v1.0 pass@1 leaderboard. Distinct agent scaffolds retain their source scores and source version metadata; these are not current v1.1 results.
Metric: Pass@1 (%). Source: www.swe-marathon.org. Source: www.swe-marathon.org. Saturation forecast: Around October 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 26 | #37 |
| 2 | Claude Fable 5 | 24 | #16 |
| 3 | GLM-5.2 | 13 | #73 |
| 4 | Claude Opus 4.7 | 11 | #49 |
| 5 | Gemini 3.5 Flash | 7 | #60 |
| 6 | GPT-5.5 | 6 | #31 |
| 7 | Gemini 3.1 Pro (Preview) | 4 | #58 |
| 8 | DeepSeek V4 Pro | 4 | #102 |
| 9 | GLM-5.1 | 1 | #97 |
| 10 | Kimi K2.6 | 0 | #104 |
| 11 | MiniMax-M2.7 | 0 | #258 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=swe-marathon · How It Works · Data refreshed daily, snapshot 2026-10-20.