Benchmarks · coding
SWE-rebench, issues 15 May – 1 Jul 2026 (Nebius)
A model resolves 111 fresh GitHub issues from 65 repositories, created between 15 May and 1 July 2026, inside one fixed minimal agent scaffold, run five times. Maintained by Nebius. Score: Resolved rate (mean over five runs) on the task window's issues, fixed ReAct-style scaffold, 128K context.
13 models · 13 results · data as of 2026-10-29 · swe-rebench.com ↗
Leaderboard
Fable 5 high
All results
13 rows| # | Model | Effort | Score | Harness |
|---|---|---|---|---|
| 1 | Fable 5 | high | 64.5% | SWE-rebench fixed ReAct scaffold, tool mode |
| 2 | Grok 4.5 | high | 63.8% | SWE-rebench fixed ReAct scaffold, tool mode |
| 3 | Opus 5 | high | 63.4% | SWE-rebench fixed ReAct scaffold, tool mode |
| 4 | GLM 5.2 | high | 62.9% | SWE-rebench fixed ReAct scaffold, tool mode |
| 5 | GPT-5.6 Sol | medium | 62.3% | SWE-rebench fixed ReAct scaffold, tool mode |
| 6 | Sonnet 5 | high | 56.8% | SWE-rebench fixed ReAct scaffold, tool mode |
| 7 | MiniMax M3 | max | 47.2% | SWE-rebench fixed ReAct scaffold, tool mode |
| 8 | MiMo V2.5 Pro | xhigh | 46.5% | SWE-rebench fixed ReAct scaffold, tool mode |
| 9 | GPT-5.6 Luna | medium | 43.6% | SWE-rebench fixed ReAct scaffold, tool mode |
| 10 | DeepSeek V4 Pro | high | 40.2% | SWE-rebench fixed ReAct scaffold, tool mode |
| 11 | Qwen 3.6 27B | max | 31.2% | SWE-rebench fixed ReAct scaffold, tool mode |
| 12 | Qwen 3.6 35B-A3B | max | 24.7% | SWE-rebench fixed ReAct scaffold, tool mode |
| 13 | Qwen 3.5 35B-A3B | low | 17.1% | SWE-rebench fixed ReAct scaffold, tool mode |