Author's Description
MiniMax-M1 is a large-scale, open-weight reasoning model designed for extended context and high-efficiency inference. It leverages a hybrid Mixture-of-Experts (MoE) architecture paired with a custom "lightning attention" mechanism, allowing it to process long sequences—up to 1 million tokens—while maintaining competitive FLOP efficiency. With 456 billion total parameters and 45.9B active per token, this variant is optimized for complex, multi-step reasoning tasks. Trained via a custom reinforcement learning pipeline (CISPO), M1 excels in long-context understanding, software engineering, agentic tool use, and mathematical reasoning. Benchmarks show strong performance across FullStackBench, SWE-bench, MATH, GPQA, and TAU-Bench, often outperforming other open models like DeepSeek R1 and Qwen3-235B.
Key Specifications
Supported Parameters
This model supports the following parameters:
Features
This model supports the following features:
Performance Summary
MiniMax-M1 demonstrates exceptional reliability, consistently providing usable responses with minimal technical failures, ranking in the 99th percentile. However, this reliability comes with a trade-off in speed, as the model tends to have longer response times, placing it in the 3rd percentile across benchmarks. Its pricing is positioned at premium levels, ranking in the 17th percentile. Across specific benchmarks, MiniMax-M1 exhibits strong performance in accuracy, particularly excelling in Email Classification with perfect 100% accuracy, making it the most accurate model at its price point and among models of comparable speed. It also shows high accuracy in General Knowledge (99.8%), Ethics (99.0%), Coding (91.0%), and Reasoning (92.0%). Its primary weakness lies in Instruction Following, where its accuracy drops to 62.6%. While its cost per query varies, it generally falls within the 8th to 25th percentile for cost-effectiveness on individual benchmarks. The consistently high duration across all benchmarks reinforces its slower processing speed. Overall, MiniMax-M1 is a highly reliable and accurate model, particularly for knowledge-intensive and complex reasoning tasks, but users should account for its longer response times and premium pricing.
Model Pricing
Current Pricing
Feature | Price (per 1M tokens) |
---|---|
Prompt | $0.3 |
Completion | $1.65 |
Price History
Available Endpoints
Provider | Endpoint Name | Context Length | Pricing (Input) | Pricing (Output) |
---|---|---|---|---|
Minimax
|
Minimax | minimax/minimax-m1 | 1M | $0.3 / 1M tokens | $1.65 / 1M tokens |
Novita
|
Novita | minimax/minimax-m1 | 1M | $0.55 / 1M tokens | $2.2 / 1M tokens |
Benchmark Results
Benchmark | Category | Reasoning | Free | Executions | Accuracy | Cost | Duration |
---|
Other Models by minimax
|
Released | Params | Context |
|
Speed | Ability | Cost |
---|---|---|---|---|---|---|---|
MiniMax: MiniMax M1 (extended) Unavailable | Jun 17, 2025 | — | 128K |
Text input
Text output
|
★ | ★ | $$$$ |
MiniMax: MiniMax-01 | Jan 14, 2025 | ~456B | 1M |
Text input
Image input
Text output
|
★★★ | ★★★★ | $$$ |