Mistral-7B-Instruct-IPO: benchmark results
Princeton NLP's 7B Mistral Instruct checkpoint tuned with IPO preference optimization, released as part of their SimPO preference-optimization research line. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1365 ± 19, rank #1356 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 50.15 | Normalized accuracy (%) | 84.2 |
| Open Korean LLM - KorNAT-Harmless | 65.47 | Normalized accuracy (%) | 80.4 |
| Open LLM Leaderboard - MuSR | 43.24 | Score | 72.5 |
| Open Korean LLM - Ko-IFEval | 33.64 | Strict accuracy, prompt/instruction mean (%) | 68 |
| Open Korean LLM - Ko-EQ-Bench | 17.64 | EQ-Bench score | 63.5 |
| Open Portuguese LLM - FaQuAD NLI | 71.45 | Macro F1 (%) | 63.4 |
| Open LLM Leaderboard - IFEval | 49.29 | Score | 57.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 22.9 | Exact match, flexible extract (%) | 55.7 |
| Open Portuguese LLM - ASSIN2 RTE | 89.9 | Macro F1 (%) | 53 |
| Open Korean LLM Leaderboard | 34.84 | Average Score (%) | 49 |
| Open Ko-LLM Leaderboard | 34.84 | Average (%) | 48.8 |
| Open Portuguese LLM - ENEM | 59.83 | Accuracy (%) | 46.4 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-instruct-ipo · How It Works · Data refreshed daily, snapshot 2026-10-23.