Zhipu's GLM-5.3 coding model unexpectedly develops advanced vulnerability discovery capabilities

According to CSO Online, Chinese AI developer Zhipu launched GLM-5.3, a coding-focused model that developed advanced cybersecurity capabilities faster than expected. On the CyberGym vulnerability identification benchmark, GLM-5.3 scored 84.5%, ahead of Anthropic's Mythos 5 (83.8%) and OpenAI's GPT-5.6 Sol (83.6%), though it trails both on exploitation tasks measured by ExploitBench (54.4% versus 78% and 76.5% respectively).