Mistral-7B-Instruct-KTO: benchmark results
Princeton NLP's 7B Mistral Instruct checkpoint tuned with KTO preference optimization, released as part of their SimPO preference-optimization research line. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1366 ± 21, rank #1337 of 1654 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GPQA-Diamond | 26.26 | Normalized accuracy (%) | 85.5 |
| Open Korean LLM - KorNAT-Harmless | 66.21 | Normalized accuracy (%) | 84.9 |
| Open Korean LLM - KorNAT-Helpful | 49.95 | Normalized accuracy (%) | 83 |
| Open Korean LLM - Ko-IFEval | 36.45 | Strict accuracy, prompt/instruction mean (%) | 75.5 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 25.02 | Exact match, flexible extract (%) | 58.5 |
| Open Korean LLM - Ko-EQ-Bench | 11.78 | EQ-Bench score | 57.7 |
| Open LLM Leaderboard - IFEval | 49.08 | Score | 57.4 |
| Open Portuguese LLM - ASSIN2 RTE | 90.32 | Macro F1 (%) | 56.4 |
| Open Portuguese LLM - FaQuAD NLI | 67.63 | Macro F1 (%) | 54.5 |
| Open Portuguese LLM - BLUEX | 52.71 | Accuracy (%) | 53.8 |
| Open Korean LLM Leaderboard | 35.2 | Average Score (%) | 50 |
| Open Portuguese LLM - ENEM | 59.76 | Accuracy (%) | 46.2 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-instruct-kto · How It Works · Data refreshed daily, snapshot 2026-10-24.