China's Z.ai to rival Anthropic, OpenAI in coding with new AI model

Z.AI Co. is set to upgrade its flagship AI model, joining a series of Chinese open-weight offerings seeking to outcompete Anthropic PBC and OpenAI.

 

The next iteration from the Beijing-based company, GLM-5.3, will bring improved coding capabilities that will help it close the gap on AI leaderboard toppers like Anthropic’s Fable 5, the company said in a statement. Z.ai, also known as Zhipu, is iterating fast: GLM-5.3 is built atop the same roughly 700-billion-parameter base model as its predecessor released in June.

 

Z.ai plans to release the new large language model’s weights — the underlying parameters that allow developers to download and customize the system — within two weeks. It shared benchmark scores showing its latest release will be much better than the already celebrated GLM-5.2 and run close to, sometimes ahead of, Fable 5.

 

Despite those claims, Z.ai’s shares slid as much as 9% in Hong Kong Friday, while close rival MiniMax Group Inc. fell 16%. Analysts attributed the losses to cautious investors cashing in recent gains ahead of the pair’s quarterly results, given the enormous volatility in AI names in recent weeks.

 

The US and China are engaged in a white-hot race for dominance in emerging technologies including AI. San Francisco-based Anthropic and OpenAI are widely recognized for having the most capable models, though recent weeks have seen Chinese companies achieve comparable performance at a lower cost.

 

Moonshot’s Kimi, DeepSeek’s latest V4 models, and Alibaba Group Holding Ltd.’s Qwen have each ranked highly in artificial intelligence tests. Z.ai is amping up the competition by also releasing its product with a permissive license that aims to attract the most users and developers.

 

The opportunity for Z.ai to grab more market share is opening up. DeepSeek on Thursday announced a price increase of up to fourfold for its V4 Flash and Pro models, though the blended price for V4 Pro remains lower than that of GLM-5.2, according to Artificial Analysis. The third-party evaluator gives DeepSeek’s top model and GLM-5.2 the same intelligence score of 53, trailing Kimi K3 as well as US frontier offerings such as Claude Fable 5 and OpenAI’s GPT 5.6.   

 

Z.ai is one of China’s AI pioneers and the world’s first large language model maker to go public. At its peak following GLM-5.2’s release, its market value surged to $137 billion, surpassing internet giants such as PDD Holdings Inc. and NetEase Inc. It has since fallen to roughly $80 billion — still a tenfold gain since its January listing in Hong Kong.

 

Z.ai has completed construction of a data center where it has installed at least 10,000 Chinese-made chips to help develop and deploy its GLM models, Bloomberg reported in July. Its annual recurring revenue — the predictable yearly sales from developers and businesses tapping its services — reached $1 billion in the same month.