Z.ai's GLM-5.3 Outperforms Anthropic's Mythos 5 on CyberGym Benchmark

Chinese AI firm Zhipu AI (HKG: 2513), trading internationally as Z.ai, has made a bold claim about its latest model, GLM-5.3. According to the company, its new model outperforms Anthropic's Mythos 5 on the CyberGym benchmark, which measures a model's ability to check for security vulnerabilities.

The CyberGym benchmark is used to evaluate AI models' capabilities in identifying and preventing cyber threats. Z.ai's GLM-5.3 reportedly surpasses Mythos 5 in this regard, potentially indicating significant advancements in the field of artificial intelligence.

Z.ai has not provided further details on how GLM-5.3 outperforms Mythos 5 or what specific improvements it offers. However, the claim suggests that Chinese AI firms are making notable strides in developing more effective and efficient models for security applications.