China's Z.ai says new model nears Anthropic's Mythos 5 in cyber-defence tests

Chinese AI startup Z.ai claims its open-source GLM-5.3 model has come close to Anthropic's restricted Mythos 5 in testing software vulnerabilities. Z.ai reported that GLM-5.3 scored 84.5% on CyberGym, a test assessing the model's ability to review code, identify security flaws, and confirm their legitimacy, just slightly above Mythos 5's 83.8%.

However, GLM-5.3 trailed Mythos 5 in converting discovered flaws into working attacks, with GLM-5.3 scoring 54.4% on the ExploitBench test compared to Mythos 5's 78.0%. Z.ai also noted that GLM-5.3 completed 105 attack-development tasks in two hours and 130 in six hours, while Mythos 5 managed 181 and 247 tasks, respectively. In response to concerns about AI systems finding and exploiting software flaws, Z.ai plans to release GLM-5.3 publicly within two weeks after completing security assessments and enhancing safeguards, with sensitive cybersecurity functions accessible only to verified users through a "trusted access" program.