Infinity-Instruct-3M-0625-Mistral-7B: benchmark results
BAAI's SFT-only fine-tune of Mistral 7B v0.1 on the Infinity-Instruct 3M and 0625 datasets, tuned without RLHF. Provider: BAAI. Released 2024-07-09. Access: Open.
Aggregate Rating 1442 ± 18, rank #1039 of 1656 rated models, from 22 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 47.03 | Strict accuracy, prompt/instruction mean (%) | 91.6 |
| Open Portuguese LLM - FaQuAD NLI | 81.11 | Macro F1 (%) | 91.5 |
| Open Korean LLM - KorNAT-Helpful | 50.45 | Normalized accuracy (%) | 85.9 |
| Open Ko-LLM Leaderboard | 41.27 | Average (%) | 81.7 |
| Open Korean LLM Leaderboard | 41.17 | Average Score (%) | 81.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 44.12 | Exact match, flexible extract (%) | 79 |
| Open Portuguese LLM - ASSIN2 RTE | 91.95 | Macro F1 (%) | 72.8 |
| Open Korean LLM - Ko-EQ-Bench | 25.5 | EQ-Bench score | 71.7 |
| Open LLM Leaderboard - IFEval | 58.67 | Score | 70.5 |
| AlpacaEval 2.0 | 31.42 | LC Win Rate (%) | 70.3 |
| Open LLM Leaderboard - MuSR | 42.72 | Score | 67.5 |
| Open Korean LLM - KorNAT-Harmless | 63.55 | Normalized accuracy (%) | 64.2 |
Interactive version: theaggregate.ai/model?slug=infinity-instruct-3m-0625-mistral-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.