Ministral 8B: benchmark results
Mistral's 8B edge model ('les Ministraux') with a 128K context and interleaved sliding-window attention. Provider: Mistral. Released 2024-10-16. Access: Open.
Aggregate Rating 1441 ± 13, rank #1040 of 1656 rated models, from 108 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TextArena Balanced Subset | 28.81 | Trueskill (rating) | 90 |
| ViMU - Rhetoric Mechanism Identification | 31.87 | Rhetoric mechanism identification score (%): selecting the r | 66.7 |
| ViMU - Structured Subtext Understanding | 21.16 | Structured subtext understanding (%): the mean of the rhetor | 66.7 |
| CureMed-Bench | 42.87 | Logical accuracy (%; share of final answers a GPT-4o judge m | 62.5 |
| CureMed-Bench - Language Consistency | 46.93 | Language consistency (%; share of final answers written in t | 62.5 |
| TinyQABenchmark (core-en) | 80.8 | Exact Match | 50 |
| ViMU - Social Value Signal Identification | 10.45 | Social value signal identification score (%): selecting the | 40 |
| Software Engineering Arena - Model Arena | 998 | Elo Rating | 38.6 |
| SciHorizon-Gene - Overall | 45.98 | Published Mean Task Score (0-100) | 38.5 |
| Konkur 1404 - Mathematics | 33.91 | Accuracy (%, text-only) | 36.4 |
| SnakeBench | 18.7 | TrueSkill Rating | 34.9 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 90 | Questions answered with a valid query on the first attempt ( | 34.8 |
Interactive version: theaggregate.ai/model?slug=ministral-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.