MiniMax M3.1 Flash
Active Parameters
428B
Context Length
1.05M
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
120 KB/tok
License
other
Release Date
17 Oct 2026
Knowledge Cutoff
-
API Pricing (per 1M)
Input: $0.15 · Output: $0.60
Hardware & System Requirements
Precision
Context
Workload Memory: 843.49 GB
FP16 · 8K context
Base Weights: 840.05 GB
KV Cache: 0.94 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
8x NVIDIA H200 141GB
843.5 / 1015.2 GB usable · TP 8
Determine requirements for your specific workload:
Upgrade to Pro
Runtime requirements
Checkpoint format
FP16
CUDA
>= 11.8
PyTorch
>= 2.1
Transformers
>= 4.52.4
| Architecture | Precision | Attention kernel |
|---|---|---|
Blackwell, Hopper | Native FP16 | FlashAttention-3 |
Ada Lovelace, Ampere | Native FP16 | FlashAttention-2 |
Architecture Diagram
Evaluation Benchmarks
Rank
#79
| Benchmark | Score | Rank |
|---|---|---|
Web Development | 1584 | 35 |
Agent Arena | -2.76 | 50 |
Intelligence Index | 0.29 | 91 |
Rankings
Overall Rank
#79
Coding Rank
#68
About MiniMax M3.1 Flash
MiniMax M3.1 Flash is an efficient sparse multimodal foundation model built by MiniMax. It provides accelerated conditional generation and vision-language processing across complex multi-turn workflows.
Specifications
Attention
Attention Structure
DeepSeek Sparse Attention
Attention Heads
64
Key-Value Heads
4
Attention Head Dimension
128
Position Embedding
ROPE
RoPE Theta
10,000,000
Indexer Key Dimension
128
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Base Precision
FP16
Hidden Dimension Size
6,144
Number of Layers
60
FFN Intermediate Size (Dense)
12,288
Tokenizer
Vocabulary Size
200,064
Mixture of Experts
Total Expert Parameters
23.0B
Number of Experts
128
Active Experts
4
Shared Experts
1
FFN Intermediate Size (per Expert)
3,072
Dense Layers Before MoE
3
Resources
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.