Zhipu CEO Zhang Peng
TMTPOST -- Beijing-based Zhipu AI has officially entered the IPO pipeline, becoming the first among China's "Big Model Six Tigers" to begin listing preparations, according to a regulatory filing published by the China Securities Regulatory Commission (CSRC) on Monday.
The company, which evolved from Tsinghua University's computer science department and is known for its ChatGLM and GLM large models, is being counseled by China International Capital Corporation (CICC).
The listing process began with internal preparations in April, with formal counseling expected to run from August to October this year. This timeline suggests a potential IPO filing with the Shanghai or Hong Kong Stock Exchange by the end of 2025.
Zhipu's listing attempt signals the beginning of a high-stakes race between U.S. and Chinese AI giants to tap public markets and secure long-term financing for AI model scaling.
Founded in 2019, Zhipu has raised over 16 billion yuan (around $2.2 billion), attracting major backers including Hillhouse, Qiming Venture Partners, and tech giants like Alibaba, Tencent, and Xiaomi. Its product line has rapidly evolved from early GLM models to today's GLM-4 series, featuring the just-released GLM-4-32B-0414, now China's fastest open-source inference model at 200 tokens per second.
Zhipu's open-source ChatGLM series has been downloaded over 30 million times globally, with 150,000+ GitHub stars. The Qingyan AI assistant app has over 25 million users and strong commercial adoption across sectors, from finance to automotive.
CEO Zhang Peng reaffirmed the company's IPO ambitions, emphasizing its "full-stack independence" across computing platforms—including compatibility with Ascend, Hygon, Sunway, and dozens of domestic GPU chips—and asserting that U.S. export controls would not materially affect business operations.
Meanwhile, in a strategic countermove, OpenAI released its new GPT-4.1 series—standard, mini, and nano—on April 15, boasting 1 million token context windows and a 26% cost reduction compared to GPT-4o. These models, aimed at developers via API access, outperform their predecessors on coding benchmarks like SWE-bench and are part of OpenAI's longer-term plan to build "Agent engineers" capable of end-to-end software development.
CEO Sam Altman said GPT-4.1 mini and nano are OpenAI's most cost-effective offerings to date. With ChatGPT usage reportedly now covering nearly 800 million users globally, OpenAI's momentum remains unmatched.
Fresh off a $40 billion funding round led by SoftBank, OpenAI is now valued at $300 billion, making it the world's second-most valuable private tech company after SpaceX. Industry insiders suggest that an OpenAI IPO is “inevitable,” as investor Mark Klein of SuRo Capital hinted it would become part of their portfolio.
With Zhipu and OpenAI both advancing toward IPOs and doubling down on product innovation, the stage is set for an AI model showdown—not just on technical performance, but in the global capital markets.
As Zhang Peng put it: “The path to AGI is just beginning. What matters most is bringing large model capabilities into the real world to solve complex problems.”
With the IPO race now officially underway, the world will be watching who turns promise into profit first.
