06/10 2026
330

By Guo Tian
Source / Jiedian Finance
Amidst speculation about the imminent listings of Zhipu AI and MiniMax on the STAR Market, The Wall Street Journal reported that StepOn Star is gearing up to submit its listing application to the Hong Kong Stock Exchange as early as June 8, aiming to become the third major large-scale model company to go public.
This marks the second company led by AI veteran Yin Qi to pursue an IPO, following Qianli Technology. Since Yin Qi took the helm, StepOn Star has accelerated its momentum, securing two rounds of financing this year totaling over 20 billion yuan, with investors proposing a cornerstone valuation of approximately 10 billion USD.
While StepOn Star may not boast a high profile among consumer-facing applications, failing to rank in the top 10 for monthly active users of AI-native apps, its models are already operational on 42 million mobile phones and automotive terminals from another perspective. Its strength lies not in having the most chatbot users but in the close integration of its models with hardware.
Betting on AI + Terminals

The once-unified "AI Six Tigers" have significantly diverged under the influence of DeepSeek.
Zhipu AI and MiniMax made their debut on the Hong Kong Stock Exchange in January this year, with both companies also advancing towards dual listings in the "A+H" markets. They stand out as prominent large-scale model startups, with Zhipu AI's market capitalization nearing 500 billion HKD and MiniMax's approaching 150 billion HKD as of press time.
Baichuan and 01.AI have stepped back from the competition for general-purpose large models. Meanwhile, Yuezhi Anymian and StepOn Star are progressing at different speeds. Yuezhi Anymian recently secured 2 billion USD in financing, boasting over 10 billion yuan in cash reserves, and is dismantling its red-chip structure, showing no immediate urgency for an IPO. In contrast, StepOn Star is clearly picking up the pace—completing its joint-stock transformation and red-chip dismantling in April, closing its Pre-IPO round in May, and expected to submit its application in June.
Industry insiders suggest that while StepOn Star and Yuezhi Anymian are latecomers in this wave of IPOs, the key lies in the strategies they employ when they finally enter the market.
Zhipu AI focuses on government and enterprise privatization deployments, MiniMax breaks through via overseas markets and consumer-facing products, Yuezhi Anymian pursues a pure foundational model route, and DeepSeek maintains stickiness through low costs and open-source strategies. StepOn Star has chosen AI + terminals—not just developing models but embedding model capabilities into hardware terminals like smartphones and automobiles.
Let's first examine foundational models. The capability of foundational models directly impacts the performance of applications and products. However, a practical issue arises: if models are solely sold as APIs, developers can easily switch to cheaper alternatives with minimal migration costs.
To date, StepOn Star has released a series of models from Step1 to Step3. On June 4, Step 3.7 Flash secured multiple top rankings on Artificial Analysis. Officially defined as an "Agent production-grade model," Step3.7Flash reflects StepOn Star's strategic shift towards Agent capabilities.
Evidently, pure technology does not constitute an absolute competitive advantage. The rapid rise and fall of glm, kimi, and mimo over the past year serve as vivid examples. StepOn Star, though a latecomer, is no slouch technically.
Shifting focus to consumer-facing products, QuestMobile data reveals that as of March 2026, the top three AI-native apps by monthly active users are Doubao, Qianwen, and DeepSeek. StepOn Star's previously launched role-playing product "Maopao Ya" has been discontinued and integrated into the StepOn AI app, which failed to crack the top 10. Currently, the main battleground among major model companies has shifted from chatbots to coding. Monthly active users for chatbots do not equate to commercial value; terminal-side revenue offers more stable and promising commercial prospects than consumer-facing traffic.
What truly sets StepOn Star apart is its integration of models into hardware terminals. Running on mobile chips or embedded in automotive cockpits creates significantly higher switching costs and more predictable revenue streams.
The data speaks volumes: StepOn Star has partnered with leading smartphone manufacturers such as OPPO, Honor, and ZTE, with an installation base exceeding 42 million units and nearly 20 million daily services. Terminal Agent API calls have grown for three consecutive quarters by the end of 2025, with a cumulative increase of nearly 170%. Platform API usage has surged nearly 20-fold over the past year. Revenues reached approximately 500 million yuan in 2025 and are projected to hit 1.2 billion yuan in 2026.
This is the fundamental difference between StepOn Star and other companies. It is not betting on users opening an app to chat with AI but on its models running at the core when users use their phones or drive their cars.
Yin Qi's AGI Journey: A Chinese xAI + Tesla Narrative

StepOn Star's bet on closed loops is intertwined with Yin Qi's experience at Megvii and his unique AGI philosophy.
The pursuit of AGI is a favorite topic among Chinese AI entrepreneurs, yet consensus on the most effective path remains elusive.
DeepSeek firmly believes that large language models are the key to AGI and has not heavily invested in multimodal outputs. In contrast, Yin Qi sees the integration of AI with the physical world as essential for achieving AGI, leading him to forge an industrial closed loop.
Public information indicates that Yin Qi assumed the role of StepOn Star's chairman in January 2026, though he was reportedly one of the initial founders. He also serves as chairman of Qianli Technology.
Qianli Technology, formerly Lifan Technology, underwent restructuring in 2020, with Geely becoming its controlling shareholder. In July 2024, Yin Qi invested over 2.4 billion yuan to acquire a 19.91% stake in Qianli Technology, becoming its second-largest shareholder. He was elected chairman in November of the same year, leading the company's transformation into an AI + automotive enterprise. In 2025, Geely merged its Zeekr intelligent driving team, Geely Research Institute, and Megvii's Maichi Intelligent Driving into Qianli Intelligent Driving, involving approximately 3,000 personnel.
A complete chain has taken shape: StepOn Star provides model capabilities, Qianli Technology handles mass production and delivery, and Geely offers terminal scenarios and vehicle platforms. Yin Qi has also stated that Tesla's Grok + FSD route represents the only path forward for autonomous driving, leading to comparisons of StepOn Star + Qianli + Geely as China's version of xAI + Tesla.
Having already ventured into automobiles and smartphones, the next steps include wearables and robots.
Based on Yin Qi's insights into the AI industry, he is highly optimistic about VLA (Vision-Language-Action). He believes that future wearables and robots will feature numerous motion components or actuators, making VLA, which coordinates the brain with actuators, a key differentiator for StepOn Star.
Yin Qi's obsession with closed loops also stems from his experience at Megvii. During the AI Four Dragons era, Megvii raised 14 billion yuan in financing, incurred 15 billion yuan in losses, and withdrew its IPO application after five and a half years. This experience led him to conclude that the theory of "assembling the plane while jumping off the cliff" is untenable. Therefore, at StepOn Star, he has aimed to integrate models, terminals, and scenarios from the outset rather than first developing the strongest model and then seeking commercialization.
Yin Qi's approach is not confined to PowerPoint slides; he continuously paves the way with concrete actions.
According to public information, Qianli Technology plans to achieve 1-1.3 million units of ASD (Assisted Self-Driving) installations by the end of 2026 and aims for 8 million units by 2028.
In addition to tangible commercialization goals, Yin Qi has also expanded his network. The investor structure in the latest funding round reflects his logic, with Pre-IPO participants primarily from the terminal hardware supply chain: ZTE has previously collaborated with StepOn Star on AI functionality; Huaqin Technology and Longcheer are the world's top two mobile phone ODM manufacturers; and OmniVision specializes in image sensors.
From Jiedian Finance's perspective, terminal manufacturers' willingness to invest indicates that the strategy of terminal-side implementation is indeed effective.
StepOn Star's listing application comes at a time when companies are flocking to go public.
Chen Yiting, CEO of the Hong Kong Stock Exchange, revealed in January at Davos that over 350 companies were awaiting listing; this number rose to 410 by the end of May.
From the perspective of the large model vertical, market enthusiasm remains high. Zhipu AI and MiniMax's stock prices have continued to rise post-listing, boosting confidence among latecomers. DeepSeek is rumored to have secured nearly 50 billion yuan in its first external funding round, while Yuezhi Anymian's valuation is speculated to reach 30 billion USD.
Compared to the AI 1.0 era of the AI Four Dragons, the large model era offers a more favorable environment for commercialization, applications, and capital markets.
Particularly as market sentiment shifts from speculation to performance, StepOn Star seems to have arrived at a more opportune moment.
However, submitting the listing application is just the beginning.
The key questions are whether API calls for large models can continue to increase, whether terminal-side revenue can sustain growth, and whether Yin Qi's integrated industrial chain can secure more deeply bound terminal partners beyond Geely. While the closed-loop narrative offers stability, Yin Qi still needs to expand his network to unlock greater imagination. Fortunately, StepOn Star's ceiling appears higher than that of pure chatbot companies.
*The featured image was generated by AI.