System Card← leaderboard
MiniMax

MiniMax

#27of 28 · general

MiniMax M3

428Bparams · MoE
1Mcontext
$0.30 / $1.2per 1M in / out
2026-06-01released
weights ↗
50100Coding 5 · #28 of 28Agents 3 · #29 of 29Cyber 2 · #21 of 24Engineering 2 · #25 of 26Math 4 · #25 of 26Health 21 · #19 of 24Legal 13 · #26 of 27Professional 12 · #25 of 28Writing 1 · #21 of 21 · one sourceInstruct 12 · #23 of 25Hallucination 21 · #20 of 26 · 47% of the recipeMultimodal 7 · #24 of 24
Coding5#28
Agents3#29
Cyber2#21
Engineering2#25
Math4#25
Health21#19
Legal13#26
Professional12#25
Writing1#21
Instruct12#23
Hallucination21#20
Multimodal7#24

History

ECI145.50145.75146.00146.25146.50146.75147.00Apr 2026MiniMax M2.5 · $0.27/M in · 2026-02-12 · still served · Epoch Capabilities Index 146.6 ECI pointsMiniMax M2.7 · $0.21/M in · 2026-03-18 · still served · Epoch Capabilities Index 145.8 ECI pointsMiniMax M3 · $0.23/M in · 2026-06-01 · still served · Epoch Capabilities Index 146.9 ECI pointsM2.7M2.5
  • MiniMax
  • older → newer
$ / M input tokens3 of 176 priced models · price is a third-party claim, not a vendor quote

Usage

3.4TtokensAug 2 – Oct 6▼ 10% 7d
1T2T3T4TAug 2Sep 1Oct 1

Geography

United States12.1%
China8.5%
Germany5.3%
Brazil5.1%
Spain5.0%
Colombia3.5%
Mexico3.2%
India2.9%
Argentina2.9%
Indonesia2.6%

Measured

≈ imputed · amber dot = lab-self-reported among the sources
GPQA Diamond91.8%#13Epoch AIVals.ai
HLE39.0%#20Artificial Analysis
MMLU-Pro84.2%#9Vals.ai
LiveBench68.6%#21LiveBenchLLM Stats
SWE-bench V77.8%#9LLM StatsVals.ai
LiveCodeBench82.2%#7Vals.ai
SciCode47.1%#25Artificial AnalysisEpoch AI
CritPt3.7%#27Artificial AnalysisEpoch AI
Terminal-Bench53.6%#20Vals.ai
APEX Agents32.7%#22Epoch AILLM Stats
GDPval37.1%#23Artificial Analysis
Arena1441#19Arena
AA Index29.2#25Artificial Analysis
Vals Index36.5#23Vals.ai
Epoch Capabilities Index146.9#21Epoch AI

Neighbors

Kimi Code→ less factual · pricierQwen Lite→ stronger agent · pricierMistral→ stronger coding · pricierStep→ stronger agent · pricierGLM Flash→ stronger agent · cheaper