The model is expected to be released widely in two weeks
Z.ai's GLM-5.3 scored 84.5% on CyberGym, surpassing Claude Fable 5 and GPT-5.6 Sol, with 2,400+ security vulnerabilities discovered. Chinese open-weight models are matching U.S. frontier models in cyber capabilities, commoditizing security testing tools as American labs reduce public releases.
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Sarah Grillo/Axios
Chinese open-weight models are closing in on U.S. frontier models' ability to find and exploit security flaws — and they could land in the public's hands more quickly.
Why it matters: Cyber-capable AI models can bring huge gains to both defenders and hackers eager to scale up their operations.
Two new models from Chinese firms compete with top US mainstream and frontier models. Should cyber-defenders be worried?
As cutting-edge models like Zhipu AI’s GLM-5.2 rapidly close the performance gap with frontier models, policy analysts warn a…
As Chinese AI models grow in capability and popularity among US companies, the arguing over what should be done about them has…
Andrew Ng and others argue accessible systems outperform overly restricted platforms on cybersecurity and transparency.
The latest open-source AI model is raising fresh concerns that AI-powered hacks are becoming more accessible.
The future of AI may be dictated not by who builds the smartest model, but who builds the models everyone runs on. Right now,…
