MiniMax · Released Feb 28, 2026
MiniMax-M3
Million-token open weights model with low inference cost and strong long-sequence associative retrieval.
OpenRating Index
45.6
0–100 composite
Reasoning Index
63.2
0–100 composite
Cost / Task
$0.14
benchmark run
Output speed
116 tok/s
blended tok/s
Time to first token
0.5s
median latency
Input price
$0.1
per 1M tokens
Output price
$0.4
per 1M tokens
Context window
1M
tokens
Strengths
Million-token context window with open weights
Low per-task cost ($0.14) for document-scale search
Strong performance on conversational storytelling and creative text
How we rate MiniMax-M3
OpenRating scores every model on the same axes. The OpenRating Index is a weighted composite of reasoning, knowledge, and agentic benchmarks, normalised to a 0–100 scale. Speed is blended output measured in tokens per second under standardised load, with time-to-first-token reported separately. Prices are list input and output prices per million tokens, and Task Cost measures end-to-end workload spending.
OpenRating Index
45.6
Reasoning Index
63.2
Token Pricing
Input per 1M tokens
$0.1
Output per 1M tokens
$0.4
Cost of 1M input + 1M output
$0.5
Task Cost Economics
Empirical cost breakdown for a standard OpenRating Index benchmark task ($0.14 total).
Related models
Compare MiniMax-M3 with the models it is most often measured against.