【#Tech24H】On August 19, Zhipu launched the API for its next-generation foundational model, GLM-5.3. The model excels in complex coding, defensive cybersecurity, and long-horizon tasks. Currently, cutting-edge models are shifting from competing solely on larger parameter scales to competing on effective intelligence per unit of cost. It is reported that GLM-5.3, with a smaller parameter count, higher parameter efficiency, and lower inference cost, reaches the level of frontier models, achieving the lowest per-task cost among flagship models and significantly advancing the “intelligence-cost” Pareto Frontier, benefiting the accelerated deployment and largescale adoption of AI applications. GLM-5.3 does not rely on simply scaling up parameters; it shares the same base model as its predecessor GLM-5.2, with performance improvements coming primarily from post-training. This demonstrates that advancing frontier capabilities does not necessarily depend on larger parameter sizes or greater computational investment. The model has already been integrated into official products such as ZCode and GLM Coding Plan, with API pricing remaining consistent with GLM-5.2, and the model weights will be open-sourced next Friday. [ By Zhang Liyan | Tang Ruohan ]