Z.ai released GLM-5.3 on August 14, 2026, and the model immediately identified a significant security flaw in Cursor, the AI-powered code editor. On the CyberGym vulnerability discovery benchmark, GLM-5.3 scored 84.5%, edging out Mythos 5 (83.8%) and GPT-5.6 Sol (83.6%). On ExploitBench, which measures a model’s ability to reason about and execute exploits, the model achieved 54.4%, roughly double the score of its predecessor GLM-5.2. Since the release of GLM-5.2, Z.ai’s models have collectively identified 2,436 vulnerabilities across 269 projects, including 1,097 classified as critical or high severity, some of which had gone undetected since 1981.
Source: Read the original article
