Mistral Nemo — benchmark results
Mistral's 12B Apache-2.0 model built with NVIDIA, with a 128K context, designed as a drop-in upgrade for Mistral 7B. Provider: Mistral. Released 2024-07-18. Access: Open.
Unified ELO 1470 ± 10, rank #907 of 1776 rated models, from 341 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGC-Bench - puntuguese | 1.32 | Dataset z-score | 97.6 |
| AGC-Bench - hypobench | 0.43 | Dataset z-score | 96.3 |
| LIBRA - ruQasper | 24.47 | Dataset Total Score (%) | 92.9 |
| LIBRA - ruSciPassageCount | 12.83 | Dataset Total Score (%) | 92.9 |
| AGC-Bench - creatset | 1.3 | Dataset z-score | 86.6 |
| AGC-Bench - c3_crosstalk | 0.89 | Dataset z-score | 79.3 |
| LIBRA - MatreshkaNames | 32.22 | Dataset Total Score (%) | 78.6 |
| LIBRA - ruTREC | 58.4 | Dataset Total Score (%) | 78.6 |
| AGC-Bench - humor_transfer | 0.88 | Dataset z-score | 77.2 |
| AGC-Bench - pun_eval | 0.66 | Dataset z-score | 76.8 |
| AGC-Bench - ocw_connections | 0.88 | Dataset z-score | 75.6 |
| LIBRA - ruBABILongQA3 | 16 | Dataset Total Score (%) | 71.4 |
Interactive version: theaggregate.ai/model?slug=mistral-nemo · How the rankings work · Data refreshed daily, snapshot 2026-07-22.