MiniMax-01
MiniMaxยทminimax/minimax-01
Ultra-long context 456B MoE model supporting native 4M token context windows with linear computational efficiency via Lightning Attention.
Open WeightsOpenTool CallingStructured OutputsNatural Language ProcessingArtificial Intelligence
Context
4,194,304
Output Limit
8,192
Linked Papers
12
Citations
380
Release Date
Jan 15, 2025
Model Specifications
- Lab / Vendor
- MiniMax
- Model Family
- MiniMax
- Modality
- text, image
- Architecture
- Lightning Attention Linear-Complexity MoE
- Parameters
- 456B MoE (45.9B active)
- Context Window
- 4,194,304
- Max Output
- 8,192
- Input / Output Price
- $0.2 / $1.1 per 1M
- License
- Apache 2.0
Research & Evaluation Papers
Academic evaluations, empirical benchmarks, and publications for MiniMax-01.
Sort