Arena-Hard v2 - release-reported (judge not stated)
Chat & Writing
Arena-Hard v2.0 scores reported by a model release without stating the judge and baseline configuration. Kept as a diagnostic. Higher is better.
Top models (higher is better)
| Model | Score |
|---|---|
| Mistral Medium 3.1 | 67.3 |
| Mistral Large 3 675B Instruct 2512 | 66.7 |
| Mistral Small 4 | 58.3 |
| Mistral Small 3.2 24B Instruct 2506 | 43.1 |
| Mistral Small 3.1 24B Instruct 2503 | 19.6 |