mistral-small-latest — benchmark results

Rolling alias for Mistral's current Small - as of mid-2026 Mistral Small 4, a 119B MoE (6.5B active) hybrid unifying chat, reasoning and coding (256K context). Provider: Mistral. Released 2026-03-16. Access: Open.

Unified ELO 1519 ± 28, rank #732 of 1776 rated models, from 9 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Arcadia MMMU Open Ended35.85Accuracy (%)90
Turkish MMLU67Accuracy (%)69.2
CyberSecEval2 Prompt Injection28.59Accuracy (%)44.4
Arcadia MMMU Multiple Choice54.55Accuracy (%)36.4
OpenAI HumanEval82.32Accuracy (%)31.8
CyberSecEval2 Interpreter Abuse18.55Accuracy (%)22.2
Arcadia CommonsenseQA82.23Accuracy (%)10
CyberSecEval2 Vulnerability Exploit35.91Accuracy (%)10
GDM InterCode CTF35.44Accuracy (%)0

Interactive version: theaggregate.ai/model?slug=mistral-small-latest · How the rankings work · Data refreshed daily, snapshot 2026-07-22.