MiniMax M2 has been released on Ollama's cloud platform as a model designed for coding and agentic workflows. It demonstrates highly competitive performance across mathematics, science, instruction following, coding, and tool use according to Artificial Analysis benchmarks.
The model features 10 billion activated parameters out of a total of 230 billion, which supports lower latency and higher throughput for interactive agents and batched sampling. Its composite score ranks #1 among open-source models globally in general intelligence evaluations.
Engineered for end-to-end developer workflows, MiniMax M2 excels at multi-file edits and test-validated repairs. It shows strong performance on Terminal-Bench and (Multi-)SWE-Bench tasks across various programming languages.
In agent evaluations using BrowseComp-style metrics, the model consistently locates difficult sources, maintains traceable evidence, and recovers from flaky steps when executing complex toolchains involving shells, browsers, retrieval systems, and code runners.
Users can access the model via Ollama's cloud API or local deployment. Configuration examples are provided for VS Code integration, Zed editor support on Windows, and direct API usage with authentication.
