appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Xinmei Shen

The Hong Kong Generative AI Research and Development Centre (HKGAI) has officially launched a new DeepSeek-based large language model that can run on domestic chips, as the government-backed lab seeks to commercialise its products and export Chinese AI overseas.
The HKGAI-V3 model, built on DeepSeek V4, has achieved “significant improvements” in efficiency and agentic capabilities, the centre said on Wednesday.
The home-grown model delivered over tenfold improvement in the efficiency of token compression, which reduced the number of tokens a model needs to process and therefore its response time, and an almost hundredfold increase in uninterrupted agent runtime, according to the lab.
Agent Workshop, an AI agent platform the HKGAI developed based on HKGAI V3, for instance, operated stably for up to 28 hours without interruption in a single session to produce a research report, HKGAI director Guo Yike said on Wednesday at the launch event.

HKGAI, established in October 2023 under the government’s InnoHK funding programme, focuses on developing AI models and products tailored to Hong Kong’s needs. It operates HKChat, the city’s first Cantonese-enabled chatbot trained to offer information on local services, laws and regulations, and several AI productivity tools for government departments.
HKGAI has further “embedded” Hong Kong’s “cultural DNA” and “internalised” the city’s local knowledge into the new V3 model, making it more aligned with the linguistic context and habits of local users, the centre said.
HKGAI’s launch of its first V1 model last year showed that Hong Kong “was not missing out on this technological wave represented by artificial intelligence”, Hong Kong’s Secretary for Innovation, Technology and Industry Sun Dong said on Wednesday.
“Today, with the release of V3, I say with even greater pride: we are not only present – we are gradually moving to the forefront,” Sun said.
Guo, also the provost at the Hong Kong University of Science and Technology, told the South China Morning Post last month that the HKGAI-V3 had been optimised to operate on both Western and domestic hardware, including Huawei Technologies’ Ascend 910C chips.

Its ability to run on Chinese chips would allow local government bodies and enterprises to operate AI systems with their own preferred infrastructure rather than relying on public cloud platforms, Guo said in May.
Hong Kong’s public sector has been increasingly adopting mainland Chinese technology solutions amid growing risks of US export control measures, which treat the special administrative region the same as the mainland.
HKGAI on Wednesday did not mention its ability to run on Chinese AI chips, only saying that it could provide “diversified inference computing power and intelligent services”.
A proponent for the idea of sovereign AI – broadly defined as the ability for a jurisdiction to develop and operate AI systems on its own infrastructure – Guo is now pushing for HKGAI to replicate its experience localising DeepSeek for Hong Kong to overseas markets.
“Expanding abroad is definitely a crucial strategy for AI development,” Guo said. “We have the ability to [ensure] a model aligns with a society’s governance and political environments. Having this capability is vital for going global.”
The lab, aiming to move towards commercialisation as it is set to complete its incubation under the InnoHK programme in September, said it was open to collaborating with partners across computing infrastructure, industry solutions and applications.
HKGAI was already working with companies including Huawei, server maker Inspur Group, and China’s three major telecommunications operators – China Mobile International, China Unicom Global, and China Telecom Global, it said.


B-2 Spirit, AI cooperation, no US-China conflict: key Xi-Trump takeaways so far

Singapore’s finance firms pledge to train 80,000 local staff in AI skills

Don’t let AI take over our human creative expression

Alibaba unveils ‘pragmatic’ AI road map to drive monetisation, infrastructure efficiency

Artificial intelligence (AI)

Hong Kong launches DeepSeek-based AI model designed to run on domestic chips

Guo Yike, provost of HKUST and director of HKGAI, delivers remarks at the launch event on Wednesday. Photo: Dickson Lee

A robot performance at the launch of HKGAI V3 model in Hong Kong on Wednesday. Photo: Dickson Lee

Officials on stage during the launch of the HKGAI V3 model in Hong Kong on Wednesday. Photo: Dickson Lee

DeepSeek nears US$7b haul in first-ever funding round, with backing from Tencent, CATL

DeepSeek long rejected external funding to focus on pursuit of artificial general intelligence. Photo: AFP

An entity named Hangzhou Chengce Business Consultation was formed on May 19, sharing DeepSeek’s registered address. Photo: Shutterstock Images

Tencent shares jump 10% on expectations of AI agent within WeChat super app

Tencent has repeatedly stressed its cautious approach to adding an agent to its super app. Photo: Reuters

Tech

Tech Trends

News

China Future Tech

Agent Workshop operated stably for up to 28 hours without interruption in a single session to produce a research report


Agent Workshop operated stably for up to 28 hours without interruption in a single session to produce a research report.


Hong Kong launches DeepSeek-based AI model designed to run on domestic chips

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Agent Workshop operated stably for up to 28 hours without interruption in a single session to produce a research report

Agent Workshop operated stably for up to 28 hours without interruption in a single session to produce a research report