MiniMax-Text-01
A hybrid MoE using Lightning Attention, with 456B total and 45.9B active parameters, trained at 1M context and extrapolatable to 4M.
FAMILY / MINIMAX
MiniMax M
An open language-model line centered on efficient MoE, long context, coding, and agents.
- First model
- 2025
- Models
- 7
- Latest dated version
- MiniMax-M3
01 / RELEASE SEQUENCE
Family release sequence
02 / GENERATIONS
What changed
MiniMax-M1
An open reasoning model with 1M context using hybrid linear and softmax attention.
MiniMax-M2
Shifted MiniMax’s main line toward efficient coding and agent execution.
MiniMax-M2.1
Improved real-world coding, tool use, and agent reliability.
MiniMax-M2.5
A high-efficiency update for coding, tools, and workplace productivity.
MiniMax-M2.7
Strengthened agent teams, self-evolution, and end-to-end software productivity.
MiniMax-M3
MiniMax’s current open flagship with 1M context, native image/video understanding, coding, and computer use.