Mistral-7B-Base-SFT-IPO: benchmark results
Princeton NLP's 7B Mistral checkpoint tuned with supervised fine-tuning and IPO preference optimization, released alongside their SimPO preference-optimization research. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1352 ± 19, rank #1386 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 35.66 | Strict accuracy, prompt/instruction mean (%) | 73.6 |
| Open Korean LLM - KorNAT-Harmless | 63.4 | Normalized accuracy (%) | 62.2 |
| Open Portuguese LLM - FaQuAD NLI | 70.37 | Macro F1 (%) | 60.8 |
| Open Korean LLM - Ko-EQ-Bench | 11.72 | EQ-Bench score | 57.6 |
| Open LLM Leaderboard - IFEval | 48.3 | Score | 56.3 |
| Open Korean LLM - Ko-GPQA-Diamond | 23.74 | Normalized accuracy (%) | 55.6 |
| Open Portuguese LLM - OAB Exams | 41.64 | Accuracy (%) | 48.6 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 16.53 | Exact match, flexible extract (%) | 47 |
| Open Portuguese LLM - ASSIN2 RTE | 88.88 | Macro F1 (%) | 46.8 |
| Open Korean LLM Leaderboard | 33.78 | Average Score (%) | 44.2 |
| Open Ko-LLM Leaderboard | 33.78 | Average (%) | 44 |
| Open Portuguese LLM - BLUEX | 48.82 | Accuracy (%) | 43.5 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-base-sft-ipo · How It Works · Data refreshed daily, snapshot 2026-10-23.