MiniMax launches MiniMax-M2.5 for coding and agentic tasks
MiniMax introduced MiniMax-M2.5, a 229B text-generation model trained with reinforcement learning across hundreds of thousands of real-world environments. The company says the model targets coding, tool use, search and office tasks, and reported 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench and 76.3% on BrowseComp. MiniMax also claims it can run continuously for an hour at 100 output tokens per second for $1.
MiniMax-M2.5 adds another model focused on coding and agentic workloads, with notably low claimed operating costs that may interest developers building tool-using applications.
MiniMaxAI/MiniMax-M2.5 Text Generation • 229B • Updated 2 days ago • 908k • • 1.53k
pipeline_tag: text-generation license: other license_name: modified-mit license_link: https://github.com/MiniMax-AI/MiniMax-M2.5/blob/main/LICENSE library_name: transformers new_version: MiniMaxAI/MiniMax-M2.7
Join Our 💬 WeChat | 🧩 Discord community.
MiniMax Code | ⚡️ API | MCP | MiniMax Website