SauerkrautLM-Mixtral-8x7B-Instruct [float16] [rev: 330eb185920d]: benchmark results
A VAGOsolutions DPO-aligned derivative of mistralai/Mixtral-8x7B-Instruct-v0.1 using German preference data. It remains distinct from the Sauerkraut release initialized from pretrained Mixtral and preserves any explicit evaluator precision/revision label. Provider: VAGOsolutions. Access: Open.
Aggregate Rating 1473 ± 33, rank #2772 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 71.37 | Accuracy (%) (5-shot) | 94.9 |
| Open LLM Leaderboard v1 - HellaSwag | 87.75 | Normalized accuracy (%) (10-shot) | 87.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 70.48 | Normalized accuracy (%) (25-shot) | 86.1 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 65.71 | MC2 (%) (0-shot) | 83.9 |
| Open LLM Leaderboard v1 - WinoGrande | 81.22 | Accuracy (%) (5-shot) | 77.8 |
| Open LLM Leaderboard v1 - GSM8K | 60.8 | Accuracy (%) (5-shot) | 73 |
Interactive version: theaggregate.ai/model?slug=sauerkrautlm-mixtral-8x7b-instruct-float16-rev-330eb185920d · How It Works · Data refreshed daily, snapshot 2026-11-08.