MiniMax M2.5
MiniMax M2.5 is a frontier mixture-of-experts model with 230 billion total parameters and 10 billion active parameters, trained via large-scale reinforcement learning across 200,000+ real-world digital environments. It achieves state-of-the-art performance in coding, agentic tool use, web search, and office productivity — scoring 80.2% on SWE-Bench Verified and 76.3% on BrowseComp — while completing complex agentic tasks 37% faster than its predecessor at a fraction of the cost of comparable frontier models.
Features
Serverless API
Minimax 2.5 is available via sciforium' serverless API, where you pay per token. There are several ways to call the sciforium API, including sciforium' Python client, the REST API, or OpenAI's Python client.
DocsAgentic Capabilities
Built for complex multi-step tasks with native agentic architecture. Excels at research synthesis, code generation, and iterative problem solving.
Docs