Ministral-3-3B-Instruct-2512: benchmark results
Mistral's December 2025 compact multimodal Ministral 3 Instruct checkpoint, with a 3.4B language model and 0.4B vision encoder. Provider: Mistral. Released 2025-12-02. Access: Open.
Aggregate Rating 1478 ± 12, rank #855 of 1656 rated models, from 354 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Polish Summarization - PSC | 25.82 | chrF3++ (%) | 100 |
| EuroEval Spanish NLU - ScaLA ES | 47.08 | Matthews correlation coefficient (%) | 94.2 |
| EuroEval Italian NLU - ScaLA IT | 53.69 | Matthews correlation coefficient (%) | 93.5 |
| EuroEval Portuguese NLU - ScaLA PT | 44.48 | Matthews correlation coefficient (%) | 91.6 |
| EuroEval French NLU - ScaLA FR | 56.71 | Matthews correlation coefficient (%) | 90.7 |
| UnpredictaBench - Code | 17.23 | KS@100 (%): share of the 238 Code problems (distributions in | 84.2 |
| EuroEval Catalan NLU - ScaLA CA | 45.23 | Matthews correlation coefficient (%) | 83.9 |
| EuroEval Albanian NLU - MultiWikiQA SQ | 59.06 | Character F1 (%) | 80.4 |
| EuroEval Portuguese NLU | 55.7 | NLU Average Score (%) | 76.9 |
| UnpredictaBench - Text | 21.25 | KS@100 (%): share of the 160 Text problems (named or describ | 76.3 |
| EuroEval Lithuanian NLU - Atsiliepimai | 45.88 | Matthews correlation coefficient (%) | 76.1 |
| EuroEval Italian NLU | 56.39 | NLU Average Score (%) | 75.4 |
Interactive version: theaggregate.ai/model?slug=ministral-3-3b-instruct-2512 · How It Works · Data refreshed daily, snapshot 2026-10-20.