On July 28, Chinese AI startup Zhipu released its latest open-source model, GLM-4.5, marking its most powerful release to date and reinforcing its ambitions in the growing field of intelligent agent applications. The new model arrives amid a surge in model releases from Chinese developers, reflecting intensifying domestic competition in generative AI, report Reuters and Bloomberg.
Zhipu is one of China's so-called "AI tigers" backed by local governments and major tech firms such as Alibaba and Tencent. Its rapid progress in securing government contracts across multiple regions was recently highlighted by OpenAI as a signal of China's growing AI capabilities.
The GLM-4.5 update positions Zhipu among a wave of companies that are accelerating their open-source strategies. Other recent moves include Moonshot's release of Kimi K2 and StepFun's non-proprietary reasoning model, as developers seek to gain traction by offering free, customizable AI tools. According to state media outlet Xinhua, China had released 1,509 large language models as of July, accounting for over 40% of the global total of 3,755 models.
According to the South China Morning Post, the fierce competition is already shifting market dynamics. DeepSeek, once the most commonly used model provider in China, has seen its market share decline sharply, according to cloud platform PPIO. Meanwhile, Alibaba's Qwen models have surged in popularity, briefly surpassing DeepSeek in usage on PPIO's platform in late May.
Zhipu's latest release may also carry financial implications. Bloomberg reported that the company is weighing a shift of its planned IPO to Hong Kong, with expectations of raising around US$300 million.
Article edited by Jack Wu
