Chinese AI company Zhipu claims its new is a better bug-finder than Anthropic, OpenAI

Chinese AI company Zhipu unveiled a new model called GLM-5.3 last week, boasting bug-finding abilities comparable to those of American models such as Anthropic's Mythos and OpenAI's GPT-5.6 Sol. The model showcased its prowess by excelling in the CyberGym benchmark, a test designed to evaluate a model's capability in resolving real-world cybersecurity challenges.

GLM-5.3 surpassed Fable 5 and GPT-5.6 Sol, demonstrating superior vulnerability discovery and the ability to form coherent plans for complete exploitation chains. Within a span of just a few months, the model demonstrated a remarkable capacity to uncover previously unnoticed vulnerabilities, dating back to around 40 years. The findings were extensive, encompassing various systems, operating systems, browser engines, open-source infrastructure, web applications, and network protocols.

Despite performing less favorably compared to western models on other security and coding benchmarks, the exceptional bug-finding prowess of GLM-5.3 indicates that China is rapidly catching up in terms of software security and is not far behind in its capability to identify vulnerabilities in rival software.