appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Wency Chen

Alibaba Group Holding has made its next-generation flagship artificial intelligence model Qwen3.8-Max widely accessible to global users ahead of an open-weights release next week.
The move marks Alibaba’s return to open-sourcing its top-tier AI models after keeping several recent flagship releases proprietary earlier this year. It also signals the firm’s entry into a recent round of powerful releases by Chinese developers aggressively narrowing the gap with leading US labs.
The massive 2.4-trillion-parameter model supports a context window of up to 1 million tokens, meaning the model can process up to 750,000 words or their equivalent for each query.
It is available to global developers via Alibaba Cloud’s Model Studio application programming interfaces (APIs), as well as through QwenWork, the company’s workplace AI agent platform, Alibaba announced on Monday. The company’s Hong Kong-listed shares rose 7 per cent following the release, closing at HK$125.20 the same day.
QwenWork, which entered public beta on the same day through web and desktop apps, is an all-in-one productivity platform for individuals and businesses, targeting Tencent’s WorkBuddy, Moonshot AI’s Kimi Work and overseas rivals including Claude Cowork and ChatGPT Work.
A multimodal foundation model, Qwen3.8-Max can process various types of data from lengthy documents to television series and live streams to build searchable knowledge bases, according to Alibaba, which owns the South China Morning Post.
The company also said the model could recreate software applications from screenshots, generate interactive games and educational animations, and convert two-dimensional floor plans into 3D visualisations.


Alibaba has positioned Qwen3.8-Max as particularly suited for autonomous coding, complex research and other long-running tasks.
According to the company, Qwen3.8-Max ranked fifth on Text Arena and second on Vision Arena – two user-voted AI leader boards – trailing only select offerings from Anthropic’s Claude series.
The model was initially previewed on July 19, when Alibaba noted its performance was “second only” to Anthropic’s Fable 5, making it available at the time on its Token Plan subscription service, as well as its Qoder and QoderWork agentic platforms.
Under the hood, Qwen3.8-Max features a sparse mixture-of-experts architecture, paired with a hybrid attention mechanism. While the model contains 2.4 trillion total parameters, it activates just 95 billion – or about 4 per cent – during inference, balancing scale with computing efficiency, Alibaba said.
The release comes as competition among Chinese AI developers shifts heavily towards coding agents, multimodal systems and software capable of executing complex, multi-step workflows.
Last month, Moonshot AI introduced Kimi K3, a 2.8 trillion-parameter multimodal model with a 1 million-token context window. Meanwhile, DeepSeek recently upgraded its V4-Flash API to improve agent capabilities and began testing a software harness to manage tool-based tasks.
Powered by Alibaba’s Qwen models, the newly released QwenWork can draft documents, analyse data, generate audio and video, and build websites, the company said. The platform combines Alibaba’s existing agent tools QoderWork, MuleRun and Wukong, and will soon be integrated into its DingTalk workplace app.


Chinese AI firm Z.ai faces reputation hit after users spot unauthorised uploads

In a first, top European design prize awarded to €20 uniform for China’s delivery drivers

Patriot dedicated to Hong Kong: business heavyweights remember Tung Chee-hwa

‘In hard times, your honour shines’: Jack Ma pays tribute to Tung Chee-hwa

Alibaba

Alibaba’s AI model Qwen3.8-Max made widely accessible ahead of open-weights release

The logo of Qwen is seen during the World Artificial Intelligence Conference in Shanghai on July 17. Photo: AFP

ByteDance and Alibaba to disable humanlike AI custom agents as new rules loom

Doubao, a ChatGPT-like conversational bot developed by TikTok owner ByteDance, informed users on Friday night that its agent feature would go offline on July 15. Photo: Simon Song

From fried chicken to flight plans: Alibaba wants Qwen to become China’s digital fixer

The latest upgrade to Alibaba’s Qwen platform brings consumer brands into its AI ecosystem, challenging rivals including Tencent in the battle for China’s next digital gateway. Photo: Shutterstock

Alibaba is betting that AI agents, rather than traditional apps, will become the primary way consumers access online services in the future. Photo: Reuters

Alibaba unveils new Qwen model, custom chips in bid to become China’s ‘AI factory’

Alibaba Grouop Holdings unveiled its new large language model Qwen3.7-Max which it said will be a robust foundation for AI models on Wednesday. Photo: Handout

What a visit to China’s AI labs reveals about the battle for global soft power

The ink-wash map of AI, with chipmakers as mountains on islands and theories flowing like rivers. Photo: Wency Chen

A ‘silicon chip’ unlocks the door at a Shanghai AI research institute. Photo: Wency Chen

Ant Group is expanding into AI healthcare with its AQ medical assistant. Photo: Wency Chen

How sell-off at flashy AI-focused US hedge fund is a wake-up call for Chinese investors

Digital screens display trading numbers on the floor of the New York Stock Exchange (NYSE) on July 23, 2026. Photo: AFP

Situational Awareness founder Leopold Aschenbrenner was a former OpenAI researcher. Photo: X.com/leopoldasch

Employees of SK Hynix attend the company’s opening bell ceremony at the Nasdaq market on the day of its IPO in New York City, July 10, 2026. Photo: Reuters

Tech

Tech Trends

News

China Future Tech

Flagship 2.4 trillion parameter model highlights native multimodal capabilities and the ability to carry out long-horizon agentic tasks


Flagship 2.4 trillion parameter model highlights native multimodal capabilities and the ability to carry out long-horizon agentic tasks.


Alibaba’s AI model Qwen3.8-Max made widely accessible ahead of open-weights release

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Flagship 2.4 trillion parameter model highlights native multimodal capabilities and the ability to carry out long-horizon agentic tasks

Flagship 2.4 trillion parameter model highlights native multimodal capabilities and the ability to carry out long-horizon agentic tasks