1 related articles
GLM-5.2 tops open-weight models in coding with a 74.4 Frontiers-WE score, beating GPT-5.5. Its MIT license enables local deployment, and the gap with closed-source flagships is closing fast.