MiniMax H3
Parameters
Undisclosed
Context Length
Undisclosed
Modality
Text
Architecture
Undisclosed
License
other
Release Date
28 Jul 2026
Knowledge Cutoff
-
API Pricing (per 1M)
-
Evaluation Benchmarks
No evaluation benchmarks for MiniMax H3 available.
Rankings
Overall Rank
-
Coding Rank
-
About MiniMax H3
MiniMax H3 is a multi-modal and video generation foundation model developed by MiniMax AI. It provides advanced high-fidelity visual synthesis and temporal consistency.
Technical Specifications
Architecture specifications are undisclosed for proprietary models.
Attention
Attention Structure
Undisclosed
Attention Heads
Undisclosed
Key-Value Heads
Undisclosed
Attention Head Dimension
Undisclosed
Position Embedding
Undisclosed
RoPE Theta
Undisclosed
Sliding Window Attention
Undisclosed
Sliding Window Size
Undisclosed
Sliding Window Ratio
Undisclosed
Linear Attention
Undisclosed
Linear Attention Ratio
Undisclosed
Normalization
Undisclosed
Activation Function
Undisclosed
Dimensions
Hidden Dimension Size
Undisclosed
Number of Layers
Undisclosed
FFN Intermediate Size (Dense)
Undisclosed
Multi-Token Prediction Heads
Undisclosed
Tokenizer
Vocabulary Size
Undisclosed
About MiniMax M3
MiniMax's flagship M3 model family, released June 1, 2026, is powered by MiniMax Sparse Attention (MSA) architecture, offering 1M context capabilities at exceptionally low compute cost and optimized for long-horizon agentic workflows.