Mixtral 8x7B (v0.1): benchmark results
Mistral's Apache-2.0 sparse-MoE model (46.7B total/12.9B active, 32K context) that beat Llama 2 70B on most benchmarks at launch. Provider: Mistral. Released 2023-12-11. Access: Open.
Aggregate Rating 1434 ± 13, rank #1083 of 1656 rated models, from 172 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Portuguese NLU - MultiWikiQA PT | 77.35 | Character F1 (%) | 94 |
| EuroEval French NLU - FQuAD | 74.48 | Character F1 (%) | 93.3 |
| European LLM Leaderboard - Zero-Shot Accuracy | 56.24 | Average Accuracy (%) | 92.7 |
| ARC Challenge (AI2) | 87.3 | Accuracy (%) | 90.8 |
| EuroEval Finnish NLU - Tydiqa FI | 72.27 | Character F1 (%) | 89.5 |
| HellaSwag | 86.7 | Accuracy (%) | 89.5 |
| URIAL-Bench - Coding | 5.3 | Judge Score (0-10) | 88.9 |
| OpenBookQA | 85.8 | Accuracy (%) | 88.1 |
| EuroEval Spanish NLU - MLQA ES | 66.3 | Character F1 (%) | 87.5 |
| Open PL LLM - PolQA Open Book (generative, 5-shot) | 89.51 | Levenshtein Similarity (%) | 87.4 |
| Open Portuguese LLM - OAB Exams | 54.9 | Accuracy (%) | 87.4 |
| EuroEval Italian NLU - SQuAD IT | 72.85 | Character F1 (%) | 87.1 |
Interactive version: theaggregate.ai/model?slug=mixtral-8x7b-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.