Download
MiniMax-M2.5 is a state-of-the-art foundation model extensively trained with reinforcement learning across hundreds of thousands of real-world environments. It delivers leading performance in coding, agentic tool use, search, and complex office workflows, achieving top benchmark scores such as 80.2% on SWE-Bench Verified and 76.3% on BrowseComp. Designed to reason efficiently and decompose tasks like an experienced architect, M2.5 plans features, structure, and system design before generating code. The model supports full-stack development across web, mobile, and desktop platforms, covering the entire lifecycle from system design to testing and code review. With native serving speeds of up to 100 tokens per second, it completes complex agentic tasks significantly faster than previous versions while maintaining high token efficiency. M2.5 is built to be highly cost-effective, enabling continuous deployment of powerful AI agents at a fraction of the cost of other frontier models.
Features
- State-of-the-art performance in coding benchmarks, including SWE-Bench Verified and Multi-SWE-Bench.
- Advanced agentic tool calling and search capabilities, excelling in BrowseComp, Wide Search, and RISE evaluations.
- Architect-style task decomposition that plans specifications, structure, and UI before implementation.
- Full-stack development support across 10+ programming languages and multiple platforms (Web, Android, iOS, Windows).
- High-speed inference with up to 100 tokens per second and 37% faster task completion compared to M2.1.
- Ultra cost-efficient deployment, with hourly operation costs as low as $0.30–$1 and significantly lower pricing than competing frontier models.
Project Samples
