MiniMax-M1
AI Model
Reasoning
Current
Open reasoning with 'lightning attention' and a million-token window.
You're looking at an AI model record: who built it, what it can do, and the hardware and data behind it.
MiniMax-M1 (June 2025) pairs a 456B/46B-active MoE with hybrid lightning attention for efficient million-token contexts — an architectural bet beyond the vanilla transformer, released openly.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Developed by
Competes with
Sources
Record reviewed as of 2026-01-15 — see Methodology.