⚙️
Inference Parameters
Temperature: 1.0
Top_p: 0.95
Optimal for creativity and diversity while maintaining logical coherence.
MiniMax-M1 Model Details
Complete technical specifications, architecture details, and performance benchmarks of the world's first open-source hybrid attention reasoning model.
Model Overview
We introduce MiniMax-M1, the world's first open-weight, large-scale hybrid-attention reasoning model. MiniMax-M1 is powered by a hybrid Mixture-of-Experts (MoE) architecture combined with a lightning attention mechanism.
The model is developed based on our previous MiniMax-Text-01 model, which contains a total of 456 billion parameters with 45.9 billion parameters activated per token. Consistent with MiniMax-Text-01, the M1 model natively supports a context length of 1 million tokens, 8x the context size of DeepSeek R1.
Furthermore, the lightning attention mechanism in MiniMax-M1 enables efficient scaling of test-time compute – For example, compared to DeepSeek R1, M1 consumes 25% of the FLOPs at a generation length of 100K tokens. These properties make M1 particularly suitable for complex tasks that require processing long inputs and thinking extensively.
Detailed Performance Benchmarks
Comprehensive evaluation results across multiple categories
| Category | Task | MiniMax-M1-80K | MiniMax-M1-40K | DeepSeek-R1 | Claude 4 Opus | OpenAI-o3 |
|---|---|---|---|---|---|---|
| Mathematics | AIME 2024 | 86.0 | 83.3 | 79.8 | 76.0 | 91.6 |
| AIME 2025 | 76.9 | 74.6 | 70.0 | 75.5 | 88.9 | |
| MATH-500 | 96.8 | 96.0 | 97.3 | 98.2 | 98.1 | |
| Coding | LiveCodeBench | 65.0 | 62.3 | 55.9 | 56.6 | 75.8 |
| FullStackBench | 68.3 | 67.6 | 70.1 | 70.3 | 69.3 | |
| Reasoning | GPQA Diamond | 70.0 | 69.2 | 71.5 | 79.6 | 83.3 |
| ZebraLogic | 86.8 | 80.1 | 78.7 | 95.1 | 95.8 | |
| MMLU-Pro | 81.1 | 80.6 | 84.0 | 85.0 | 85.0 | |
| SWE-bench Verified | 56.0 | 55.6 | 49.2 | 72.5 | 69.1 | |
| Long Context | OpenAI-MRCR (128k) | 73.4 | 76.1 | 35.8 | 48.9 | 56.5 |
| OpenAI-MRCR (1M) | 56.2 | 58.6 | -- | -- | -- | |
| LongBench-v2 | 61.5 | 61.0 | 58.3 | 55.6 | 58.8 |
Usage Recommendations
Optimal settings for different scenarios
💬
General Purpose
System Prompt:
"You are a helpful assistant."
For summarization, translation, Q&A, creative writing.
🔢
Mathematical Tasks
System Prompt:
"Please reason step by step, and put your final answer within \boxed{}."
For calculation and logical deduction problems.
Function Calling
Advanced capabilities for tool integration
Function Calling Support
The MiniMax-M1 model supports function calling capabilities, enabling the model to identify when external functions need to be called and output function call parameters in a structured format.
Tool Integration
Structured Output
Agentic Applications
API & Chatbot
For general use and evaluation, we provide online services and development tools.
Online Chat
Developer API
MCP Server
Citation
@misc{minimax2025minimaxm1scalingtesttimecompute,
title={MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention},
author={MiniMax},
year={2025},
eprint={2506.13585},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2506.13585},
}