Developing2 outlets · over 18 hours
GLM-5.2 lags frontier safety, refuses no cyber tasks
GLM-5.2, Z.ai's open-weight model, trails OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.7 by months on cyber and bio capabilities, yet refused none of SaferAI's offensive tasks. Claude Opus 4.7 refused so consistently SaferAI couldn't complete its cyber benchmark. Z.ai published no safety framework for GLM-5.2.
The story so far2 updates, latest first.
Tracked by ByteBrief