China's Z.ai claims its new GLM-5.3 model is close to Anthropic's Mythos 5 level in cybersecurity tests

Chinese artificial intelligence (AI) startup Z.ai has announced that its upcoming open-source model, GLM-5.3 , demonstrated vulnerability detection capabilities on par with Anthropic ’s restricted cybersecurity model, Mythos 5 . Citing internal benchmark data shared by the company, a news article by Reuters says that GLM-5.3 scored 84.5% on CyberGym, which is an industry benchmark evaluating a model's ability to review source code, detect security flaws, and verify their authenticity. In comparison, Z.ai reported an 83.8% score for Anthropic's Mythos 5.