MiniMax Releases M2: A High-Performance Open-Source AI for Coding
The company has published the source code for its new MiniMax-M2 model, a powerful AI optimized for complex software development and autonomous agent tasks.

Impressive Benchmark Performance
In independent tests, MiniMax-M2 outperformed models like Grok 4 Fast, Gemini 2.5 Pro, Claude Opus 4.1, the latest DeepSeek V3, and Qwen3-235B.
While GPT-5-high and Grok 4 still lead the rankings, MiniMax-M2's strong performance in benchmarks testing scientific knowledge, mathematical reasoning, and coding skills marks it as a significant new contender.
Key Features & Access
- Optimized for Coding: The model is engineered to handle complex software development and programming challenges.
- Advanced Agent Functions: It excels at autonomous tasks, such as browsing the web and searching for information.
- Open Source: The model is freely available on Hugging Face and GitHub.
- API Availability: Developers can integrate M2 via its API, with pricing set at $0.3 per 1M input tokens and $1.2 per 1M output tokens.
- Technical Specs: MiniMax-M2 is a 230-billion parameter model that actively uses 10 billion parameters during inference.
Try It For Free
You can test MiniMax-M2 for a limited time in the MiniMax AI agent chatbot. After the promotional period, the free version will revert to the previous MiniMax-M1 model, which was introduced in June 2025.
225
Comments
Please log in to add comments.
Loading comments...