Zhipu's GLM-5.3 Launches With Cybersecurity Edge, Already Finding Real Vulnerabilities

Chinese AI startup Zhipu, also known as Z.ai, released GLM-5.3 — a 700-billion parameter flagship model that beat Anthropic's Mythos 5 and OpenAI's GPT-5.6 Sol on the CyberGym cybersecurity benchmark. The model scored 84.5% success rate identifying and validating security flaws from source code, edging past Mythos (83.8%) and GPT-5.6 Sol (83.6%). Within hours of launch, GLM-5.3 reportedly found a "potentially serious vulnerability" in Cursor, the AI coding startup...