Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
MiniMax M2.7: Open Source Self-Evolving AI Model Excelling in Real-World Software Engineering Benchmarks
MiniMax has open sourced its groundbreaking MiniMax M2.7 model, a self-evolving Mixture-of-Experts (MoE) AI agent that achieves state-of-the-art results in professional software engineering benchmarks. With a 56.22% accuracy on SWE-Pro and 57.0% on Terminal Bench 2, MiniMax M2.7 matches top proprietary models like GPT-5.3-Codex, marking a significant milestone in accessible, high-performance AI.
This model is designed not only for code generation but also for complex reasoning and autonomous self-improvement, dramatically reducing production debugging times to under three minutes. Developers and organizations stand to benefit hugely from MiniMax M2.7’s ability to handle up to 50% of reinforcement learning workflows autonomously and its support for multi-agent collaboration.
This open-source release could reshape AI-assisted software development by making frontier-grade agentic models widely available to the developer community. Explore the technical details and access the model weights on Hugging Face to start integrating MiniMax M2.7 into your projects.