Mistral-Small-Instruct-2409: benchmark results
Mistral's 22B dense instruct model with function calling and a 32K context, under the Mistral Research License (September 2024). Provider: Mistral. Released 2024-09-17. Access: Open.
Aggregate Rating 1523 ± 13, rank #702 of 1656 rated models, from 234 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - PSC (multiple choice, 5-shot) | 92.63 | Binary F1 (%) | 96.2 |
| MT-Bench PL - Humanities | 10 | Judge Score (0-10) | 94.9 |
| Polish EQ-Bench | 72.85 | EQ-Bench Score (parseability-adjusted) | 93.1 |
| Open PL LLM - PoQuAD Open Book (generative, 5-shot) | 71.51 | Levenshtein Similarity (%) | 92.8 |
| Open PL LLM - PSC (generative, 5-shot) | 97.91 | Binary F1 (%) | 90.4 |
| ArgBench (Chain-of-Thought) - Argument Reasoning | 43 | Macro F1 (0-1, times 100) averaged over the argument reasoni | 90 |
| HREF | 42.87 | Average HREF Score (%) | 87.9 |
| Open PL LLM - EQ-Bench | 57.05 | EQ-Bench Score | 86.6 |
| MT-Bench PL - Reasoning | 7.9 | Judge Score (0-10) | 82.7 |
| MT-Bench PL - STEM | 9.65 | Judge Score (0-10) | 82.7 |
| Open LLM Leaderboard - GPQA | 33.31 | Score | 81.7 |
| MT-Bench PL - Coding | 7.1 | Judge Score (0-10) | 81.6 |
Interactive version: theaggregate.ai/model?slug=mistral-small-instruct-2409 · How It Works · Data refreshed daily, snapshot 2026-10-20.