Ministral 3 14B — benchmark results

Mistral's Apache-2.0 14B dense model from the Mistral 3 family, with image understanding and single-GPU focus (December 2025). Provider: Mistral. Released 2025-12-02. Access: Open.

Unified ELO 1545 ± 7, rank #638 of 1776 rated models, from 595 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval English Summarization - CNN Dailymail46.05Score (%)100
EuroEval Finnish NLU - Tydiqa FI78.61Reading comprehension Score (%)100
Konkur 1404 - Foreign Language44.23Accuracy (%, text-only)100
EuroEval Italian NLU - SQuAD IT78.66Reading comprehension Score (%)99.8
EuroEval Dutch NLU - SQuAD NL86.09Reading comprehension Score (%)99.7
EuroEval French NLU - Allocine97.8Sentiment classification Score (%)99.7
EuroEval Icelandic NLU - NQII65.72Reading comprehension Score (%)99.7
EuroEval Danish NLU - Multi Wiki QA DA82.94Reading comprehension Score (%)99.6
EuroEval Estonian NLU - MultiWikiQA ET77.94Reading comprehension Score (%)99.6
EuroEval Lithuanian NLU - MultiWikiQA LT76.12Reading comprehension Score (%)99.6
EuroEval Polish NLU - Polemo298.2Sentiment classification Score (%)99.6
EuroEval Faroese NLU - FoQA79.94Reading comprehension Score (%)99.5

Interactive version: theaggregate.ai/model?slug=ministral-3-14b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.