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

Chinese AI startup Z.ai announced on August 14 that its open-source GLM-5.3 model had nearly reached the performance of Anthropic's restricted Mythos 5 in identifying software vulnerabilities. Z.ai claimed that GLM-5.3 scored 84.5% on CyberGym, a test measuring a model's ability to review code, pinpoint security flaws, and confirm their validity, which is slightly higher than Mythos 5's 83.8%.

However, GLM-5.3 fell short of Mythos 5 in converting discovered flaws into functional attacks, scoring 54.4% on the ExploitBench test, compared to Mythos 5's 78.0%. In a separate timed test, GLM-5.3 completed 105 attack-development tasks in two hours and 130 in six hours, while Mythos 5 completed 181 and 247 tasks, respectively.

Z.ai plans to release GLM-5.3 publicly within two weeks after conducting security assessments and enhancing its safeguards.