Zhipu, MiniMax, and Moonshot are competing to become the “first stock in the large-model sector.”

On one side, there’s model R&D that continues to burn cash; on the other, scaled profitability remains elusive. Faced with the patience-testing demands of the capital markets and the fierce reshuffling within the industry, three major large-model unicorns—Zhipu, MiniMax, and Moonshot—have simultaneously chosen to kick off their race toward an IPO.

According to reports, MiniMax, headquartered in Shanghai, plans to launch its IPO in Hong Kong as early as January 2026. Zhipu, meanwhile, completed its listing guidance filing with the Beijing Securities Regulatory Bureau in April 2025, becoming the first company among the “Big Model Six Little Tigers” to initiate the IPO process and is currently considering shifting its focus to the Hong Kong stock market. Meanwhile, Yue Zhi An Mian, whose valuation is expected to reach US$4 billion, also plans to kick off its IPO journey in the second half of 2026 following the completion of a new round of financing worth hundreds of millions of dollars.

Different paces: Three unicorns with distinct sprint rhythms

Based on currently available public information, the three unicorns are at three different stages on their journey toward the public capital markets. Their progress and plans clearly outline a path—from launch to sprint.

Zhipu’s progress is the clearest so far. On April 14, 2025, the company officially filed its guidance registration with the Beijing Securities Regulatory Bureau, with China International Capital Corporation serving as the guidance institution. This marks Zhipu as the first among the “Six Little Tigers” of large-scale AI models to initiate the IPO process. According to the filing information, the guidance program will last seven months and is expected to be completed by October 2025.

Following closely behind is MiniMax. According to relevant reports, the company is preparing to list on the Hong Kong stock exchange and has already retained financial advisors for this purpose. The IPO could raise hundreds of millions of dollars. Sources indicate that the IPO might take place as early as January 2026. Previously, insiders revealed that MiniMax is also exploring the possibility of listing on the A-share market, underscoring its multifaceted considerations regarding capital markets.

A reporter from the Science and Technology Innovation Board Daily learned from relevant sources that, influenced by the high listing requirements and relatively slow pace of advancement on the STAR Market, Zhipu may abandon its previous plan to list on the STAR Market and instead focus on pursuing a Hong Kong stock market listing. Meanwhile, driven by broader macroeconomic forces and leveraging Shanghai’s local resource advantages, MiniMax—whose core team has ties to SenseTime and boasts greater flexibility in commercial operations—may be ahead in terms of its listing progress.

However, a relevant source also told a reporter from the Science and Technology Innovation Board Daily that Zhipu has a small probability of going public before the end of the year.

Also actively preparing for an IPO is Moonlit, according to a report by the Science and Technology Innovation Board Daily. A reporter previously learned from insiders that Moonlit is about to complete a round of U.S. dollar financing, which could boost its valuation to around 4 billion U.S. dollars. It is understood that prospective investors include IDG Capital and Tencent, and the funding amount could reach several hundred million U.S. dollars. In addition, Moonlit may launch its IPO in the second half of next year.

Two years after the surge in large-model startups, leading companies are collectively entering a critical phase of rushing toward the capital markets.

Capital’s Favor: Valuation Surge Behind Massive Funding Rounds

The ability of these three companies to pursue IPOs is closely tied to the massive funding they’ve secured over the past several years. The continuous influx of capital has not only supported their high R&D costs but has also driven their valuations to impressive levels.

MiniMax is highly regarded in the capital market. In July 2025, the company nearly completed a new round of Series C funding totaling nearly $300 million, with participation from investors including the Shanghai State-owned Capital Mother Fund. Post-investment, the company’s valuation exceeded $4 billion (approximately RMB 30 billion).

Since its establishment in 2019, Zhipu has completed more than 15 rounds of financing to date, attracting investment from well-known institutions such as Hillhouse Capital, Qiming Venture Partners, and Junlian Capital, as well as internet giants including Meituan, Tencent, and Xiaomi. In just the six months since December 2024, Zhipu has secured over 5 billion RMB in funding in a concentrated series of financing rounds. According to available data, its latest valuation has reached 40 billion RMB.

Moreover, according to the aforementioned sources, capital operations are not a company’s core competitiveness; rather, technological strength is the key.

In terms of technological competition, on December 9, Zhipu AI open-sourced its core AI agent model, AutoGLM, late at night. This model is regarded by the industry as the world’s first AI agent equipped with “Phone Use” capabilities, enabling it to reliably complete complex multi-step processes—such as placing food delivery orders and booking airline tickets—that involve dozens of steps.

In addition, during the Multimodal Open Source Week, Zhipu released a series of core technologies and models covering visual understanding, device control, speech processing, and video generation.

Zhang Peng, CEO of Zhipu, recently revealed to a reporter from the Science and Technology Innovation Board Daily that the company’s software tools and model business aimed at developers—known as the GLM Coding Plan—has generated annual recurring revenue (ARR) exceeding 100 million RMB (equivalent to 14 million USD). In addition, Zhipu expects its total revenue to grow by more than 100% in 2025. Zhang Peng stated that the company aims to increase the revenue contribution from its API business to half of its total revenue. Currently, Zhipu’s API business platform serves over 2.7 million paying customers, including some of China’s largest technology companies.

Moonshot, which achieved a rapid valuation leap in a short period of time, has already begun preparations for the next-generation model, according to a reporter from the STAR Market Daily. Earlier, during Reddit’s “Ask Me Anything” session, a user asked the company’s founder, Yang Zhilin, about the release date of the K3 model. In response, Yang said the model would be released “before Sam’s trillion-dollar data center is completed.” Here, “Sam” refers to Sam Altman, CEO of OpenAI, who had previously pledged to invest over 1.4 trillion U.S. dollars in data center construction.

It’s worth noting that, compared to their U.S. counterparts, Chinese startups in the large-model space are not yet highly valued. According to relevant reports, OpenAI’s valuation has already reached 500 billion U.S. dollars in its latest funding round. OpenAI is also making a push toward the public market; as reported by foreign media in October, OpenAI plans to file for an IPO as early as next year, with a potential valuation of up to 1 trillion U.S. dollars.

An IPO is clearly far from being the end point for companies developing large-scale AI models. An industry insider frankly told a reporter from the STAR Market Daily that even after securing an IPO, these companies may still find it difficult to be seen by the market as absolute protagonists capable of independently rewriting the rules. They need to continuously demonstrate to the capital markets not only that they can create excellent models, but also that they can establish sustainable revenue models and differentiated competitive barriers.

“Zhipu needs to demonstrate that its leading model capabilities can be translated into stable profits; MiniMax, on the other hand, must prove that its products—such as video generation—can replicate or even surpass Talkie’s success and establish a robust technological moat. As for Moonshot AI, the challenge lies in whether its technological idealism can attract a sufficiently large and willing-to-pay professional user base.”