Chinese AI start-up ModelBest kicks off pre-IPO tutoring process for mainland listing

Chinese AI startup ModelBest has initiated its pre-IPO preparation process for a listing in mainland China, capitalizing on the increasing demand for compact AI models suitable for devices such as smartphones, laptops, and automobiles. The four-year-old firm also aims to tackle China's scarcity of advanced American computing chips by developing lightweight systems optimized for local processors.

ModelBest has engaged Citic Securities to assist with its pre-listing preparations, which typically span at least three months and cover areas like corporate governance, risk management, audit readiness, and information disclosure practices.

Founded in August 2022, ModelBest operates within the burgeoning niche of "edge AI," which emphasizes smaller, more efficient models that run directly on end-user hardware. This approach bypasses latency, reduces cloud bandwidth costs, and safeguards user privacy. ModelBest's co-founder and chief technology officer, Zeng Guoyang, highlighted this strategy in a recent interview with Chinese tech outlet 36Kr, stating that domestic compute power and data availability pose the primary challenges.

In May, ModelBest unveiled MiniCPM5-1B, a 1-billion-parameter model trained exclusively on Huawei Technologies' Ascend chips. According to Artificial Analysis, this model ranks as the top-performing open-weight model with fewer than 2 billion parameters, surpassing Alibaba Group Holding's Qwen3.5-2B. ModelBest's lightweight AI solutions have been integrated into various consumer electronics, automotive systems, and robotics applications, collaborating with industry leaders such as Huawei, Samsung Electronics, SAIC Motor, Changan Automobile, and Unitree Robotics, as reported on their WeChat social media platform.

Although ModelBest has not disclosed its planned fundraising amount, timeline, or listing location, its recent Series C+ funding round in June saw over 5 billion yuan ($738 million) in capital raised, propelling its valuation to 20 billion yuan, as per Shanghai Mejoy Capital, one of its investors. The company's public listing plans align with a global trend towards local AI deployment on everyday hardware.

In May, Meta Platforms introduced Muse Glimmer, a 30-billion-parameter open-weight model designed for local use on personal computers. Apple also disclosed two on-device models in June, with one boasting 3 billion parameters. Earlier this month, Shanghai-based StepFun unveiled StepX Neo, claiming to be the world's first "agentic smartphone" powered by StepEdge AI models.