MiniMax M2.1
MiniMax M2.1 is a 229-billion-parameter Mixture-of-Experts model released in December 2025 under a modified MIT license. It was tuned for robustness in coding, tool use, instruction following, and long-horizon planning across agentic workflows. The model shows strong multilingual programming, scoring 72.5 on SWE-bench Multilingual and 74.0 on SWE-bench Verified, with 88.0 on MMLU-Pro and 83.0 on GPQA-D. It reaches an average VIBE score of 88.6 (with 91.5 on VIBE-Web and 89.7 on VIBE-Android), reflecting robust full-stack and native mobile development capability. The sparse architecture supports faster inference and responsive agentic loops, and MiniMax positions M2.1 as matching or approaching leading proprietary coding models on real-world software engineering tasks.
Key info
Best for
How MiniMax M2.1 scores on each use case.
Benchmarks
Independent benchmark scores: the Intelligence Index and the individual evaluations behind it, grouped by skill.
Available routes
No routes currently available — MiniMax M2.1 isn't routed through the Opper gateway right now. It may return.
Contact us about this model →