Zhipu AI Unveils GLM-5.3, Boosting Cybersecurity with Open-Weights Coding Model

Zhipu AI’s GLM-5.3 coding model improves performance by 50 percent and has identified over 2,400 vulnerabilities, with open-source release imminent.

Zhipu AI has introduced GLM-5.3, a new open-weights coding model that offers a 50 percent performance improvement over its predecessor through post-training alone, according to The Decoder. This model is positioned as one of the most powerful open-source coding tools available.

In practical application, GLM-5.3 has already aided security teams in identifying 2,436 vulnerabilities across 269 projects, showcasing its potential impact in cybersecurity. The Decoder also reports that the model weights will be released as open source in two weeks, enabling broader access and adoption.

For Japanese markets, where cybersecurity and AI-driven automation are increasingly critical, GLM-5.3’s open-source availability could accelerate innovation and security measures within tech and financial sectors.

#newsroom

By Mansa Kane

Tokyo Overnight

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

No spam. Unsubscribe anytime.