ModelBest, a Chinese startup building lightweight artificial intelligence models designed to run entirely on smartphones and laptops, has raised more than $700 million this year following a C+ series round, according to a post by one of its investors.
The fundraising takes the company's valuation to roughly $2.8 billion.
ModelBest did not disclose the size of the latest tranche.
The company works in edge AI, meaning models small enough to run on the device itself rather than sending requests to a remote data centre.
The appeal is faster responses, better privacy because data never leaves the handset, and less dependence on cloud infrastructure that is expensive to rent and, in China, harder to source at the leading edge.
The money arrives during a burst of activity across the sector.
StepFun has pushed into the same territory with its StepX Neo agentic smartphone and its Step Edge family of on-device models.
Apple has been moving in parallel, shipping on-device models and licensing Apple Intelligence in China.
ModelBest has also been advertising technical progress.
In May the company said its MiniCPM5-1B, a model with one billion parameters, had been trained entirely on Huawei Ascend hardware.
Parameters are the internal values a model learns during training, and a billion is small by frontier standards, where leading systems run to hundreds of billions.
ModelBest said the model topped Artificial Analysis's intelligence index among open-weights models below two billion parameters.
The Huawei detail carries weight beyond the benchmark.
Training without Nvidia silicon is the outcome US export controls were designed to prevent, and demonstrating it on a small model is the cheapest way to prove the toolchain works.
The funding gives ModelBest the resources to test whether the same approach scales.
