RankingSWE-bench Verified · source release snapshot; version as labeled
SWE-bench Verified · source release snapshot; version as labeled
- Bucket
- Coding
- Unit
- percent
- Direction
- Higher is better
- Version
- source release snapshot; version as labeled
- Display harness
- Mistral Medium 3.5 model card performance charts
- Board
- https://huggingface.co/mistralai/Mistral-Medium-3.5-128B
Compare published benchmark results with category weights →
Models
| Model | Score | Harness | Evidence | Source-recorded date |
|---|---|---|---|---|
| Devstral 2Mistral | 72.2%Reported settings & sourceMistral previous coding model comparison, published model-card harness settings. First-party chart numeric label, visually verified. Mistral Medium 3.5 model card performance charts · images/image3.png · reviewed 2026-09-06 | Mistral Medium 3.5 model card performance charts | lab self-report | 2026-09-06 |
| Devstral Small 2Mistral | 68%Reported settings & sourceMistral previous coding model comparison, published model-card harness settings. First-party chart numeric label, visually verified. Mistral Medium 3.5 model card performance charts · images/image3.png · reviewed 2026-09-06 | Mistral Medium 3.5 model card performance charts | lab self-report | 2026-09-06 |