China's Z.AI has confirmed its ownership of the highly popular Ox Alpha AI model, now rebranded as GLM-5.3-Flash. The company announced plans to price the model competitively, making it a low-cost, high-efficiency alternative to premium AI offerings. This launch also features Zhipu's new open-weight model, GLM-5.3-Flash, which was stealth-tested on a cluster of 100,000 Chinese-made chips. The development signals China's growing dominance in both open-weight models and the hardware required to run them.Zhipu's shares saw a significant increase following this announcement, closing over 12% higher. Industry experts note a trend of Chinese labs releasing advanced open models at substantially lower prices than Western competitors. The model experienced massive usage on platforms like OpenRouter during its preview, processing trillions of tokens. Initially released uncredited, Ox Alpha quickly garnered attention for its capabilities in coding and agentic tasks, handling text, image, and video inputs.This deployment serves as a crucial test of China's domestic hardware capabilities for large-scale AI inference, aiming to reduce reliance on foreign processors amidst export controls. Zhipu utilized a specialized inference engine to optimize performance on its homegrown chips, achieving efficiency comparable to mainstream Nvidia accelerators. The company has collaborated with domestic chip developers, including Huawei, Cambricon, and Moore Threads, who have confirmed compatibility with GLM-5.3-Flash. GLM-5.3-Flash boasts a large parameter count but efficiently activates a smaller portion for requests, and is the first in its series to natively process visual information alongside text.