MiniMax M2
Open Weights
MiniMax
MiniMax's open-weight model built for maximum coding and agentic workflows, a 230B-parameter mixture-of-experts that activates only 10B parameters per token for efficient, sophisticated end-to-end tool use. It serves a 128K-token context window.
Strengths
- Built for maximum coding and agentic workflows
- Efficient 230B MoE with only 10B active parameters
- Sophisticated end-to-end tool use
- 128K-token context window
- Open weights for self-hosting
Caveats
- Text-only (no vision or audio input)
- Specialized for coding/agents over general chat
- MoE serving needs substantial GPU memory
- Vendor
- MiniMax
- Context Window
- 131,072 tokens
- License
- MIT
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Open weights - free to self-host; also served via the MiniMax API. Efficient 230B MoE with only 10B active parameters.