MiniMax M2
MiniMax M2 is a 230-billion-parameter Mixture-of-Experts model with only 10 billion active parameters per token, engineered for high-efficiency performance in coding and agentic workflows. The sparse architecture keeps capability competitive across developer-focused tasks while reducing the compute spent per token. The model scores 69.4 on SWE-bench Verified, 82 on MMLU-Pro, and 46.3 on Terminal-Bench, and its composite intelligence score ranks first among open-source models across mathematics, science, coding, and tool use. It is built for multi-file editing, compile-run-fix loops, test-validated repair, and agentic evaluations including BrowseComp and GAIA. M2 uses interleaved thinking wrapped in think tags, which should be preserved in conversation history for best results. Recommended sampling settings are temperature 1.0, top_p 0.95, and top_k 40.
Key info
Benchmarks
Independent benchmark scores — composite indices for reasoning, coding, and math, plus individual eval scores where available.
Available routes
No routes currently available — MiniMax M2 isn't routed through the Opper gateway right now. It may return.
Contact us about this model →