Z.ai’s GLM-5.3 Outperforms Top AI Models in CyberGym Vulnerability Detection

Z.ai’s GLM-5.3 model slightly outperforms Anthropic and OpenAI in CyberGym vulnerability detection, signaling strong Chinese AI capabilities.

Chinese AI firm Z.ai has introduced its latest model, GLM-5.3, which achieved an 84.5% score on CyberGym’s vulnerability detection benchmark. This result places it slightly ahead of Anthropic’s Mythos 5, which scored 83.8%, and OpenAI’s GPT-5.6 Sol at 83.6%, according to KrASIA.

The CyberGym benchmark measures AI models' ability to identify cybersecurity vulnerabilities, an increasingly critical capability as cyber threats grow more sophisticated. Z.ai’s marginal lead suggests a competitive edge in AI-driven security tools among leading global developers.

For Japanese investors and tech companies closely following AI advancements, this development highlights the rising influence of Chinese AI innovations in cybersecurity, a sector increasingly integrated with FX and equities markets due to evolving digital infrastructure risks.

#newsroom

By Mansa Kane

Tokyo Overnight

Japan markets, decoded. Get the brief in your inbox.

No spam. Unsubscribe anytime.