MiniMax

MiniMax M3

MiniMax frontier coding and agentic model with native multimodality, sparse attention, and a 1M-token context window.

Best for: Long-horizon coding, autonomous agents, tool use, image and video understanding, document analysis, and sustained multi-step work.

Released

2026-06-01

Capabilities

  • ✓reasoning
  • ✓Temperature control
  • ✓Web search
  • ✓Adjustable reasoning effort
  • ✓Image understanding
  • ✓video_input
  • ✓File/document understanding
Try MiniMax M3 in EVA

We use analytics cookies to understand how EVA is used and improve it. No chat content is ever collected.