Inside Alibaba’s AI acceleration: A deep dive into Qwen 3.8 Max and QwenWork

Alibaba has announced the launch of Qwen 3.8 Max, a new model it has described as “the most powerful” in its Qwen series. In fact, a quick look at the LLM Arena, that real-time ranking generated by users around the world, suggests that the Chinese giant has hit the mark with this model.

For example, on the very day of its launch, it was already ranked fifth overall in the text capabilities category (the ability to understand and generate human language), behind only Anthropic’s Claude models. It also stands out in areas such as vision (2nd) and coding (4th), although it does not make the top spots in fields such as image and video generation, where Alibaba already has other alternative models.

The new Qwen 3.8 Max, in detail

If we zoom in a bit on the new model, we can see that it is a massive-scale system. For example, if you remember, a few days ago we told you about the disruptive arrival of Kimi K3, an open model that caused a major stir after being accused of distilling the capabilities of certain closed models such as Claude Fable. Kimi K3 has 2.8 trillion parameters, while Qwen 3.8 Max comes in only slightly behind, with 2.4 trillion.

Alibaba says the new model features a “state-of-the-art” sparse mixture-of-experts (MoE) architecture combined with a hybrid attention mechanism, allowing it to activate only 95 billion parameters. In this way, it can combine cutting-edge intelligence for complex tasks with “significantly” lower computational costs and latency compared with traditional models of a similar scale.

On the other hand, and although, as is now often the case with new AI launches, it is a multimodal model, Qwen 3.8 Max stands out especially in areas such as text. Still, Alibaba highlights the power of Qwen 3.8 Max in two specific fields in its announcement: coding and visual tasks.

In coding, the company explains that the new model can work “independently for extended periods without human intervention.” In fact, in internal testing, Alibaba says the model autonomously carried out a real-world software engineering project for 16 days, a self-evolving agent framework that has been released as open source on GitHub. It is also capable of building complete web front-end projects from a single screenshot of the user interface.

In the visual domain, Alibaba boasts that Qwen 3.8 Max can seamlessly process 100-page documents, full television series, or 100-hour livestreams, turning them into interactive, searchable knowledge bases. It also highlights its capabilities in fields as diverse as video editing, generating immersive educational animations from text prompts, transforming 2D floor plans into detailed 3D interior visualizations, and creating interactive games directly from natural-language requests.

Meet Qwen3.8-Max: A New Bar for Coding and Cowork. pic.twitter.com/mJaA8vr6xT

— Qwen (@Alibaba_Qwen) August 3, 2026

Alibaba later published a battery of benchmarks comparing Qwen 3.8 Max with some of the leading models on the market. According to the company, the model delivers benchmark-leading results in areas such as reasoning, programming, tool use, and agentic tasks. As with any benchmark produced by the developer itself, these results will also need to be validated through independent evaluations.

QwenWork arrives

Qwen 3.8 Max is already available through Alibaba Cloud Model Studio APIs for developers worldwide, but it will also be available to test in the new QwenWork, Alibaba’s latest all-in-one AI agent platform for the workplace.

QwenWork, whose beta version has just been unveiled, is an all-in-one AI agent platform for the workplace, designed to boost productivity with advanced AI models. The idea is to turn the AI model into a complete workspace, where a single agent can search for information, generate documents, create images, code, or collaborate with other users without the need to switch tools.

“Thanks to its native multimodal capabilities for image, video, and audio generation, QwenWork streamlines creative workflows by eliminating the need to switch between multiple AI tools,” Alibaba explains. For now, QwenWork is available for public beta testing in China through its official website.

China’s AI now

The speed at which major Chinese models are evolving is becoming difficult to keep up with. Just a few days ago, the spotlight was on Kimi K3; now Alibaba is responding with Qwen 3.8 Max and a new agent-based productivity platform. And both are competing (at least according to LLM Arena, and at least in some rankings) with cutting-edge American models such as GPT Sol and Claude Fable.

What do you think the next step will be in this race to control AI?

Image: ChatGPT