🕒 Created

Z.ai Releases New GLM-5.3-Flash Model Using Domestic Chinese Semiconductors

Z.ai has released a new large language model, GLM-5.3-Flash, which the company claims is 10 times more cost-efficient than its predecessor. The model, which debuted anonymously as 'Ox Alpha' on the OpenRouter platform, ranks 10th on the Artificial Analysis Intelligence Index. Z.ai claims to have utilized 100,000 China-made chips to handle all online requests for the model, highlighting China's push for technological self-sufficiency in the face of U.S. restrictions on advanced semiconductors. To improve efficiency, Z.ai implemented sparse attention and linear attention mechanisms to reduce hardware overhead and memory usage. The model was trained on a dataset of 30 trillion tokens using mHC technology to optimize the workflow. The release of GLM-5.3-Flash positions Z.ai as a significant competitor to Western frontier models from companies like OpenAI and Anthropic. Following the announcement, Z.ai's Hong Kong-listed shares rose by more than 8% in Thursday trading.

Sources