MiniMax-VL-01
MiniMax-VL-01 adds a 303M-parameter vision encoder and MLP projector to MiniMax-Text-01. The model supports image-grounded conversation, document understanding, and visual reasoning while inheriting the hybrid-attention language backbone. It was trained in stages on image-caption and instruction data, with public weights and reference inference code available. Model weights use the MiniMax Model License, separate from the MIT license on code. Historical launch/API claims should not be interpreted as guarantees that the original endpoint still serves this checkpoint.
2025-01-15
456B total, 45.9B activated; 303M vision encoder
Mixture of Experts (MoE) with hybrid lightning/softmax attention and vision encoder
MiniMax Model License
Specifications
- Parameters
- 456B total, 45.9B activated; 303M vision encoder
- Architecture
- Mixture of Experts (MoE) with hybrid lightning/softmax attention and vision encoder
- License
- MiniMax Model License
- Type
- multimodal
- Modalities
- textimage
Benchmark Scores
Advanced Specifications
- Model Family
- MiniMax M
- Finetuned From
- MiniMax-Text-01
- API Access
- Not Available
- Chat Interface
- Not Available
Capabilities & Limitations
- Capabilities
- long-context understandingreasoninginstruction followingvisual reasoningdocument understanding
- Known Limitations
- Custom weight-license terms apply.The four-million-token research setting is not a universal hosted context guarantee.
- Notable Use Cases
- long-document researchself-hosted assistantsvisual question answering