Hypersphere raises hundreds of millions of RMB for AI model research

Shenzhen-based AI startup Hypersphere Technology has raised hundreds of millions of renminbi in angel funding, according to a report published by TechNode on October 26. The reported backers include Legend Capital, CAS Star, SenseTime Guoxiang Capital and WestSummit Capital.¹

The company was founded by Weiyang Liu, a professor at the Chinese University of Hong Kong, and is working on what it calls “Infinite Models”. The idea is to combine specialist models into a coordinated system rather than relying on one general-purpose model for every task.¹

How Hypersphere says its approach works

According to the report, Hypersphere plans to build its own expert models while also using capabilities from existing open and proprietary models. A routing and aggregation layer would then match individual tasks with the models best suited to handle them.¹

That puts the company in a familiar but increasingly important part of the AI market: building systems that decide which model should do the work, rather than treating one large model as the answer to every problem. The company’s reported research areas include model pretraining, reinforcement-learning post-training, parameter-efficient fine-tuning and model merging.¹

The funding report does not give a precise amount, valuation or timetable for a product launch. It also does not establish whether Hypersphere’s approach has been independently benchmarked against competing model-routing systems. Those details matter because the practical value of a multi-model architecture will depend on how reliably it improves results, cost or speed outside a research setting.

What the funding means

The reported investment gives Hypersphere room to hire, train models and build the infrastructure needed to test its “Infinite Models” concept. It also shows that investors remain willing to back new approaches to foundation-model development, even as the market becomes more crowded.

For now, the clearest takeaway is that Hypersphere has attracted a substantial early vote of confidence, not that it has already produced a breakthrough system. More evidence will be needed before the company’s model-coordination claims can be judged against established AI platforms.

Our opinion

Hypersphere’s idea is more interesting than another promise to build a simply larger model. Routing work between specialist systems could make AI more efficient if the coordination layer is good enough, but it also creates another engineering problem: the system must choose correctly, combine outputs sensibly and keep costs under control. The funding gives the company a chance to prove that thesis. It is a notable start, but the evidence will be in the benchmarks and the products, not the size of the angel round.

¹ TechNode reports the funding, investors, founders and “Infinite Models” approach; the company’s independently verified financial terms and performance benchmarks were not included in the report.