MiniMax M2.5: Faster. Stronger. Smarter. Built for Real-World Productivity.
Summary
MiniMax announces MiniMax M2.5, an RL-trained model claiming state-of-the-art performance in coding, agentic tool use, search, and office work, with SWE-Bench Verified at 80.2% and other benchmarks indicating strong results. It emphasizes speed and cost efficiency, with two variants and a new Forge RL framework for scalable agentic tasks.