System Card← leaderboard
Mistral

Mistral

#26of 28 · general

Large 4 (preview)

1.1Tparams · MoE
1Mcontext
$1.36 / $4.18per 1M in / out
2026-10-06released
50100Coding 27 · #21 of 28Agents 17 · #26 of 29Cyber 0 · #22 of 24Engineering 8 · #24 of 26Math 20 · #22 of 26Health 6 · #24 of 24Legal 27 · #20 of 27Professional 7 · #27 of 28Writing 10 · #17 of 21 · one sourceInstruct 17 · #21 of 25Hallucination 15 · #24 of 26Multimodal 9 · #22 of 24
Coding27#21
Agents17#26
Cyber0#22
Engineering8#24
Math20#22
Health6#24
Legal27#20
Professional7#27
Writing10#17
Instruct17#21
Hallucination15#24
Multimodal9#22

History

rating1,4101,4151,4201,4251,4301,435Jan 2026Apr 2026Jul 2026Oct 2026Mistral Large 3 · $0.5/M in · 2025-12-02 · still served · text 1,414 ratingMistral Medium 3.5 · $1.5/M in · 2026-04-28 · still served · text 1,426 ratingMistral Large 4 · $0.68/M in · 2026-10-06 · still served · text 1,429 ratingMedium 3.5Large 3
  • Mistral
  • older → newer
$ / M input tokens3 of 176 priced models · price is a third-party claim, not a vendor quote

Measured

≈ imputed · amber dot = lab-self-reported among the sources
GPQA Diamond88.1%#21Epoch AI
HLE35.0%#26Artificial Analysis
LiveBench71.8%#18LiveBench
SciCode54.2%#18Artificial AnalysisEpoch AILLM Stats
CritPt10.6%#24Artificial AnalysisEpoch AI
GDPval46.1%#21Artificial Analysis
Arena1429#21Arena
AA Index38.4#19Artificial Analysis
Vals Index48.0#21Vals.ai
Epoch Capabilities Index150.0#20Epoch AI

Neighbors

Qwen Lite→ weaker math · cheaperMiniMax→ weaker coding · cheaperKimi Code→ weaker math · weaker codingStep→ more factual · cheaperMiMo Flash→ stronger agent · cheaper