miqu-1-70B-sf: benchmark results
Community dequantization of 'miqu-1-70b', a leaked early Mistral Medium prototype retrained from Llama 2 70B, whose origin Mistral's CEO confirmed (February 2024). Provider: Other. Released 2024-02-01. Access: Open.
Aggregate Rating 1509 ± 26, rank #757 of 1656 rated models, from 121 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - WinoGrande | 85.32 | Accuracy (%) (5-shot) | 98.1 |
| Open LLM Leaderboard v1 - MMLU | 75.49 | Accuracy (%) (5-shot) | 96.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 73.04 | Normalized accuracy (%) (25-shot) | 95.2 |
| EuroEval Spanish NLU - Sentiment Headlines ES | 51.27 | Matthews correlation coefficient (%) | 94.5 |
| Open LLM Leaderboard v1 - HellaSwag | 88.61 | Normalized accuracy (%) (10-shot) | 93 |
| EuroEval Swedish NLU - Swerec | 79.32 | Matthews correlation coefficient (%) | 92.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 69.38 | MC2 (%) (0-shot) | 88.9 |
| Open LLM Leaderboard - GPQA | 35.07 | Score | 87.9 |
| Open Portuguese LLM - ENEM | 74.67 | Accuracy (%) | 87.2 |
| Open LLM Leaderboard - MuSR | 45.82 | Score | 87 |
| Open Portuguese LLM - ASSIN2 RTE | 93.1 | Macro F1 (%) | 86.4 |
| Open Portuguese LLM - BLUEX | 64.81 | Accuracy (%) | 85.7 |
Interactive version: theaggregate.ai/model?slug=miqu-1-70b-sf · How It Works · Data refreshed daily, snapshot 2026-10-20.