What is best local LLMs for strategic reasoning?

Hey guys, I work with AI safety and much of our work is with huge models like Opus, Astra,... Recently we carried out a series of experiments and found out that models like Sonnet 4.5 and GLM 5.3 are able to notice...