Mistral-7B-Instruct-DPO: benchmark results
Princeton NLP's 7B Mistral Instruct checkpoint tuned with DPO preference optimization, released as part of their SimPO preference-optimization research line. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1353 ± 19, rank #1380 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GPQA-Diamond | 25.76 | Normalized accuracy (%) | 81.3 |
| Open Korean LLM - KorNAT-Harmless | 64.71 | Normalized accuracy (%) | 76.5 |
| Open Korean LLM - KorNAT-Helpful | 48.88 | Normalized accuracy (%) | 74.1 |
| Open Korean LLM - Ko-IFEval | 35 | Strict accuracy, prompt/instruction mean (%) | 72 |
| Open LLM Leaderboard - IFEval | 51.76 | Score | 61.7 |
| Open Portuguese LLM - FaQuAD NLI | 70.07 | Macro F1 (%) | 60.5 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 25.93 | Exact match, flexible extract (%) | 59.6 |
| Open Portuguese LLM - ASSIN2 RTE | 89.73 | Macro F1 (%) | 51.8 |
| Open Korean LLM - Ko-EQ-Bench | 4.94 | EQ-Bench score | 50.7 |
| Open Portuguese LLM - BLUEX | 51.46 | Accuracy (%) | 49.6 |
| Open Portuguese LLM - ENEM | 59.69 | Accuracy (%) | 46 |
| Open Korean LLM Leaderboard | 33.31 | Average Score (%) | 41 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-instruct-dpo · How It Works · Data refreshed daily, snapshot 2026-10-23.