Mistral-7B-Base-SFT-DPO: benchmark results
Princeton NLP's 7B Mistral checkpoint tuned with supervised fine-tuning and DPO preference optimization, released alongside their SimPO preference-optimization research. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1354 ± 19, rank #1377 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 51.43 | Normalized accuracy (%) | 90.5 |
| Open Korean LLM - KorNAT-Harmless | 66.64 | Normalized accuracy (%) | 86.4 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.25 | Normalized accuracy (%) | 75.9 |
| Open Korean LLM - Ko-IFEval | 32.83 | Strict accuracy, prompt/instruction mean (%) | 63.9 |
| Open Portuguese LLM - FaQuAD NLI | 70.58 | Macro F1 (%) | 61.6 |
| Open LLM Leaderboard - MuSR | 41.22 | Score | 52.3 |
| Open Portuguese LLM - ASSIN2 RTE | 89.42 | Macro F1 (%) | 50.2 |
| Open Korean LLM - Ko-EQ-Bench | 3.76 | EQ-Bench score | 50 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 17.97 | Exact match, flexible extract (%) | 49.1 |
| Open Korean LLM Leaderboard | 34.49 | Average Score (%) | 47.9 |
| Open Ko-LLM Leaderboard | 34.49 | Average (%) | 47.8 |
| Open LLM Leaderboard - IFEval | 44.03 | Score | 47.3 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-base-sft-dpo · How It Works · Data refreshed daily, snapshot 2026-10-23.