SWE-bench Multilingual
Code · 2025-05-06
SWE-bench Multilingual evaluates software-engineering agents on 300 curated issue-resolution tasks drawn from 42 repositories across nine programming languages. Agents modify a repository snapshot to resolve a real GitHub issue, and success requires both issue-specific and regression tests to pass.
Top models (higher is better)
| Model | Score |
|---|---|
| DeepSeek V4.1 Flash | 98.2 |
| DeepSeek V4 Pro 0813 | 96.4 |
| Claude Opus 5.5 | 95.5 |
| Claude Fable 5 | 94.8 |
| GPT-5.6 Terra | 94.4 |
| GPT-5.6 Sol Pro | 94.2 |
| Grok 4.6 | 92.7 |
| GPT-5.6 Luna | 92.4 |
| Kimi K3 | 91.4 |
| Claude Fable 5.1 | 91.1 |
| Claude Sonnet 5.5 | 90.9 |
| GLM-5.3 Flash | 88.9 |
| Grok 4.5 | 87.7 |
| GLM-5.3 | 87.6 |
| Claude Mythos Preview | 87.3 |
| Gemini 3.8 Flash | 85.8 |
| Claude Haiku 5.5 | 85.1 |
| GPT-6.1 Sol | 84.9 |
| Opus 4.8 | 84.4 |
| Grok 4.7 | 84.2 |
| Gemini 3.7 Flash | 82.6 |
| GPT-6 Sol | 81.7 |
| Qwen3.8-Flash-Next | 81.0 |
| GLM-5.2 | 78.6 |
| Laguna S 2.1 | 78.5 |
| Beam 1.0 501B A23B | 78.3 |
| Qwen3.7-Max (source unspecified) | 78.3 |
| GPT-6 Luna | 78.2 |
| Sonnet 4.6 | 78.1 |
| Opus 4.6 | 77.8 |
| GPT-5.5 | 77.7 |
| Inkling | 77.7 |
| GPT-5.4 | 77.0 |
| Opus 4.5 | 76.2 |
| DeepSeek-V4-Pro | 76.2 |
| MiniMax M3 | 76.0 |
| Hy3 | 75.8 |
| Kimi K2.7 Code | 75.6 |
| Gemini 3.6 Flash | 75.1 |
| Gemini 3 Flash Preview | 72.7 |