China’s Z.ai unveils GLM-5.3, claims chart-leading scores

Chinese AI firm Z.ai has introduced its latest open-weights model, GLM-5.3, boasting performance that rivals the newest AI launches from OpenAI and Anthropic. The new model showcases a significant leap in capability over its predecessor GLM-5.2, particularly excelling in complex coding and long-horizon tasks with a 50 percent improvement in in-house code benchmarks.

Z.ai plans to release GLM-5.3's weights in two weeks following a thorough safety evaluation and hardening process. The model demonstrated exceptional prowess in exploiting vulnerabilities, scoring 84.5 percent on CyberGym, surpassing Anthropic’s Mythos 5 and OpenAI’s GPT-5.6 Sol. Despite these impressive claims, Z.ai's shares experienced a nearly 4 percent decline in market value following the announcement.

Z.ai, a Hong Kong-listed company, went public in January this year with an initial market capitalization of nearly $7 billion, which has since dropped to around $75 billion. Experts caution that the company remains on an unsustainable commercial footing due to the escalating costs of agentic AI.