Zhipu announced today that the GLM-5.3 API is officially available starting today. The model specializes in complex coding, defensive cybersecurity, and long-horizon tasks.
According to the company, GLM-5.3 scored 60 on the globally authoritative Artificial Analysis Intelligence Index (AA Intelligence Index), placing it in the capability range of frontier models worldwide. It is on par with closed-source flagship models such as Claude Fable 5 and GPT-5.6 Sol, and is tied with Kimi K3 for first place among open-source models.
While achieving intelligence at the same level, GLM-5.3 significantly lowers the barrier to accessing frontier-model capabilities through a smaller parameter scale and lower usage costs.
In the chart, the vertical axis represents the Artificial Analysis Intelligence Index, while the horizontal axis represents the average cost of completing a single task. GLM-5.3 occupies the most attractive high-intelligence, low-cost range, with the lowest per-task cost among frontier flagship models, substantially advancing the intelligence-cost Pareto frontier for frontier models.
According to the company, GLM-5.3 pursues not only a higher intelligence ceiling but also greater intelligence per unit cost. As the API becomes officially available, frontier intelligence will enter the real-world workflows of more developers, businesses, and knowledge workers.
IT Home learned that GLM-5.3 and the previous-generation GLM-5.2 use the same base model, with performance improvements coming primarily from post-training.
GLM-5.3 has currently been integrated into coding platforms such as ZCode and included in the GLM Coding Plan. API access is now available, with pricing unchanged from GLM-5.2. The model weights will be open-sourced next Friday.