Alibaba Unveils Qwen3.8-Max, Its Largest AI Model, Amidst Growing Competition

On Monday, the Chinese company Alibaba unveiled what it described as its largest and most capable artificial intelligence model to date: Qwen3.8-Max. Its size is comparable to the model launched last month by its domestic rival, Moonshot AI.

Chinese tech companies—a dominant global force in open-weight AI models—are waging a fierce, fast-paced battle to develop more powerful systems without making their operation prohibitively expensive.

Qwen3.8-Max boasts 2.4 trillion parameters (the numerical settings a model learns from data and uses to recognise patterns, generate responses, and perform tasks), while Moonshot’s Kimi K3 model has 2.8 trillion.

A higher figure does not automatically make a model better, but it has become a closely watched indicator for measuring the scale of computing power and data underpinning advanced AI systems.

Chinese tech companies are keen to publish parameter counts to help their models gain traction within the developer community. Their models are typically "open-weight," meaning the underlying learned parameters—which allow developers to run or adapt the system—are available for download.

In contrast, OpenAI, Anthropic, and Google do not publish parameter counts for their closed-source models.

Qwen3.8-Max was introduced on Arena.AI, a model comparison platform based on crowdsourced evaluations. It immediately ranked as the top Chinese model in the text-based category, although it still trails behind Claude Fable 5 and three variants of Opus, all developed by Anthropic. However, in the Arena.AI rankings for AI models that analyse images and other visual content, Qwen3.8-Max took second place globally, surpassed only by a variant of Claude Fable 5.

Both Qwen3.8-Max and Kimi K3 are capable of handling text, images, and video, as well as processing up to one million tokens simultaneously. Tokens are data fragments—often parts of words or short words—and a high number indicates that the model can process vast amounts of information at once, such as lengthy legal documents, extensive software source code, or hundreds of pages of documentation.

Alibaba reported that its model utilises a "mixture-of-experts" design, which distributes the workload among specialised system components rather than activating the entire model for every request. Consequently, only 95 billion parameters are used simultaneously, reducing both costs and response times.

The tech giant noted that the model completed a software engineering project in 16 days. Qwen3.8-Max is scheduled to launch next week via Alibaba Cloud's Model Studio platform.