The open-weight model MiniMax M2.7 has achieved performance parity with proprietary models in core agent tasks, according to an independent evaluation by LangChain. The model excels in file operations, tool calling, and instruction following, offering a significant cost reduction of approximately 20 times and a speed increase of 2- to 4-fold. For instance, while Opus 4.6 incurs a daily cost of $250 for 10 million tokens, MiniMax M2.7 operates at just $12 per day. This development, supported by benchmarks like SWE-Rebench and Terminal Bench 2.0, positions open models as viable alternatives for production environments, facilitating collaboration with proprietary systems.