appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Xinmei Shen

As the global artificial intelligence race pushes the limits of scale with massive, trillion-parameter models, a parallel contest is heating up over smaller systems.
Instead of relying on sprawling, power-hungry cloud data centres, a growing number of tech companies in China and elsewhere are turning to lightweight models designed to run entirely on smartphones or laptops.
Proponents said this localised approach would be a game-changer, promising faster processing, enhanced data privacy and lower operating costs.
Investors are buying in. Beijing-based ModelBest, a prominent player in on-device AI, raised more than 5 billion yuan (US$738 million) this year, according to a WeChat post by investor Shanghai Mejoy Capital.
ModelBest’s latest C+ Series funding round, completed earlier this week, propelled the firm’s valuation to 20 billion yuan, according to the post.
The start-up did not disclose the size of its latest funding round and did not immediately respond to a request for comment on Thursday.
The funding milestone comes amid a flurry of activity in China’s so-called local AI or edge AI sector.
Earlier this week, Shanghai-based StepFun – backed by tech giant Tencent Holdings – unveiled the StepX Neo, which it claimed to be the world’s first agentic smartphone.
Designed around autonomous AI agents, the phone was built to interpret complex user requests and execute tasks independently, StepFun said. The company has yet to announce a launch date or pricing.
A day earlier, StepFun introduced Step Edge, a family of on-device AI models offering “ultra-low local latency” as short as 0.1 seconds, secure local data processing, and the ability to handle simple tasks even with weak or non-existent network connections.
The rise of edge AI in China comes amid Beijing’s push for technological self-reliance. While home-grown AI chips still face bottlenecks when training frontier-scale models, they are better suited to training lightweight alternatives.
In May, ModelBest revealed that its billion-parameter on-device model, MiniCPM5-1B, was trained entirely on Huawei Technologies’ Ascend hardware. The model topped independent AI benchmarking and insights platform Artificial Analysis’ intelligence index for open-weights models under 2 billion parameters.
Apple, which this week secured a long-awaited licence to operate its Apple Intelligence service on iPhones in China, has accelerated its own on-device strategy.
In June, the US tech giant announced two on-device models: the 3 billion-parameter AFM 3 Core, and an advanced 20 billion-parameter model that selectively activates between 1 and 4 billion parameters based on user tasks.
Apple was also evaluating technology from Silicon Valley start-up PrismML, which claimed earlier this week to have successfully run Alibaba Group Holding’s 27 billion-parameter Qwen3.6 model locally on an iPhone, according to a CNBC report on Tuesday. Alibaba owns the South China Morning Post.
Alibaba said on Wednesday its Qwen models would be “integrated into Apple Intelligence experiences” in China, while Baidu said it was working with Apple to develop localised AI features.


Philippine tribe vows to stand ground against US-led Pax Silica: ‘listen to us’

As the US weighs restrictions on cloud computing, how will China’s AI sector adapt?

Chinese AI chipmaker Hygon plots expansion from data centres to robotics

For China’s hi-tech IPOs, it’s Hong Kong vs the mainland – but how much does it matter?

Artificial intelligence (AI)

Giant models grab headlines, but some Chinese start-ups bet on smaller phone-ready AI

On-device models promise faster processing, enhanced data privacy and lower operating costs, according to proponents. Photo: Shutterstock

AI investor mania: China’s DeepSeek chases US$70 billion valuation in fresh round

DeepSeek has launched a new financing round shortly after closing a landmark deal. Photo: Getty Images

Chinese AI labs to challenge Thinking Machines Lab with new industry focus

Two Chinese AI labs, Yoolee AI and InfiX.ai, aim to challenge the likes of Thinking Machines Lab and Palantir with a focus on AI solutions tailored for specific industries, reduced costs and enhanced privacy. Photo: Shutterstock

AI scientist Yang Hongxia, a professor at Hong Kong Polytechnic University, has compared her AI start-up InfiX.ai to Thinking Machines Lab. Photo: Handout

Sam Altman signals OpenAI price war as rivalry with Anthropic, China heats up

OpenAI CEO Sam Altman speaks at an event in Tokyo, Japan, last year. Photo: Reuters

China works on AI safety benchmark as regulators target large model risks

Chinese authorities have started building a safety benchmark to evaluate artificial intelligence models. Photo: Shutterstock

In China, parents turn to AI to help children select a university degree

Graduates celebrate after a graduation ceremony at Renmin University of China in Beijing. Photo: Xinhua

Tech

Tech Trends

Business

China Business

Companies

A growing number of tech firms are pivoting to lightweight models designed to run entirely on smartphones and laptops


A growing number of tech firms are pivoting to lightweight models designed to run entirely on smartphones and laptops.


Giant models grab headlines, but some Chinese start-ups bet on smaller phone-ready AI

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}A growing number of tech firms are pivoting to lightweight models designed to run entirely on smartphones and laptops

A growing number of tech firms are pivoting to lightweight models designed to run entirely on smartphones and laptops