MiniMax M3

Name: MiniMax M3 model facts
Creator: OSMEXO

minimax/minimax-m3

MiniMax
Family · MiniMax M3
Released 1 Jun 2026
Open weights
Text · Image · Video → Text
Tools · JSON · Vision

Not on OSMEXO yet

Request this model

Spec & list price

Context1M

Max output512K

Input / 1M$0.30

Output / 1M$1.204× input

Scores · standing

Intelligence29.2#69/489

Coding58.6#48/196

Usage

Tokens / day80.6B▼ 58% 7d

Usage rank#37

Requests / day634.7K

Tool calls / req0.41

Per request · 30 days

Avg input55.9K3.8× median

Avg output1.0K1.5× median

Cache hit91%1.3× median

Reasoning share66%1.2× median

Tool-call failures3.6%1.9× median

Consumption

Peak 219.0B 26 OctLow 49.0B 30 OctRank today #37

News & trends

50 mentions this week ▲ 400%this model · videos, posts, articles · 30 days

Trending · 14 days

News

vllm v0.32.0vLLM · 18 Oct
vllm v0.25.0vLLM · 11 Jul
Serving MiniMax-M3 for efficient inference: Unlocking 1M-Token Context and Multimodality Without RegretsTogether AI · 2 Jun

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis

IFBench#3 of 328

82.9%

GPQA Diamond#14 of 445

92.9%

Long-context reasoning#19 of 393

83.0%

τ²-Bench#58 of 319

88.9%

Coding index#48 of 196

58.6

Terminal-Bench 2.1#46 of 182

65.2%

SciCode#75 of 160

47.1%

Terminal-Bench Hard#30 of 311

42.4%

Speed · median across 13 providers, latest day

Output speed#110/33486 tok/s1.3× faster than median

Fastest provider168 tok/sbest of 13 providers

First token#2381.64 s1.6× longer than median

First answer token#6418 sthinking included

#3 cheapest of 15 with a similar score

Value

Intelligence Index

Blended price per 1M tokens (log)

Technical details

Tokenizer: Other
License: other
Parameters: 427B
Moderation: none
Downloads · likes: 195.4K · 1.6K

Request parameters 19

tools
tool_choice
response_format
structured_outputs
reasoning
include_reasoning
temperature
top_p
top_k
min_p

+9 more

max_tokens
stop
seed
frequency_penalty
logit_bias
logprobs
presence_penalty
repetition_penalty
top_logprobs

When to use

Best for

Data analysis and reasoning over big inputs
Instruction-heavy structured output tasks
Very long context question answering
Image and video understanding in chat
Self-hosted deployments on open weights

Think twice for

Browser automation agents
Long autonomous agent runs
Image or video generation

Research papers

Same family