MiniMax M2 on OpenClaw: Setup, Pricing, and Performance...

The article describes MiniMax's M2 family of large language models, utilizing a Mixture of Experts architecture for high performance at low inference cost. The M2.7 model achieves 90% of frontier model quality at 7% of the cost, with benchmark results comparable to Claude Sonnet 4.