Sulat.com
AI models
Deep Infra logo

Model details

MiniMax-M3

MiniMax-M3 sits at the top of MiniMax's text model family and is pitched as a frontier-class open-weight release. Both the Fireworks launch coverage and the official product page position it as a capable long-context multimodal model aimed at developers building serious applications rather than casual chat. It is featured alongside MiniMax's other text models on the official site, signaling its role as the flagship text offering in the lineup.

Fireworks' launch post frames MiniMax-M3 around three themes: a dedicated MSA (Sparse Attention) section explaining how it handles long contexts, a section on frontier coding and agentic capabilities, and a direct comparison against the prior MiniMax M2.7 generation. That combination tells a clear product story, namely that MiniMax-M3 is designed for teams who need to push past standard context windows on coding, tool use, and multi-step reasoning tasks. The open-weight framing also makes it a practical fit for organizations that want frontier-style behavior with the flexibility to host or fine-tune on their own terms.

Deep InfraMiniMaxAI/MiniMax-M3minimax

Quick Info

Powered by
Provider
Deep Infra
Model key
MiniMaxAI/MiniMax-M3
Release date
Jun 1, 2026
Last updated
Jun 1, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.28
Output token cost
$1.10

Limits

Output tokens
512,000 tokens
Context window
524,288 tokens

Latest news about MiniMax-M3

Videos about MiniMax-M3

Recent tweets and retweets from Deep Infra

More models around MiniMax-M3