Zhipu AI Open-Sources GLM-5.3 Weights for Coding and Cybersecurity

Zhipu AI has officially open-sourced the weights of its GLM-5.3 model, allowing for local deployment, fine-tuning, and commercial use. The model is designed to excel in complex coding, defensive cybersecurity, and long-horizon agent tasks. Scoring 60 on the Artificial Analysis Intelligence Index, GLM-5.3 matches the performance of top-tier closed-source models like Claude Fable 5 while significantly lowering deployment costs. This release strengthens the open-source AI ecosystem by providing state-of-the-art capabilities in coding and cybersecurity for free. The model achieves a 50% improvement in coding over GLM-5.2 on the Z.ai Code Bench and requires a safety review only for organizations with annual revenues exceeding $10 billion using it for external services. Additionally, Zhipu AI conducted an extra two-week safety evaluation prior to release due to the model's advanced cybersecurity capabilities.