Alibaba has introduced Qwen3.8-Max, describing it as the largest and most capable artificial intelligence model the company has built. The launch places Alibaba in direct competition with Moonshot AI, whose Kimi K3 model debuted last month with a larger parameter count.
Qwen3.8-Max contains 2.4 trillion parameters, the numerical settings a model learns from data to recognise patterns and perform tasks. Kimi K3, by comparison, holds 2.8 trillion parameters.
While a higher parameter count does not guarantee better performance, the figure is widely regarded as an indicator of the scale of computing power and data behind an AI system.
Chinese technology firms routinely disclose parameter counts to attract developers, since most of their models are open-weight and can be downloaded and adapted. OpenAI, Anthropic and Google follow a different approach, keeping parameter counts of their closed-source models undisclosed.
Qwen3.8-Max debuted on the crowdsourced model-comparison platform Arena.AI, where it became the top-ranked Chinese model among text-based systems.
It still trails Anthropic-developed models, including Claude Fable 5 and three Opus variants. In the category evaluating models that analyse images and other visual content, Qwen3.8-Max secured second place globally, behind only a Claude Fable 5 variant.
Both Qwen3.8-Max and Kimi K3 are capable of processing text, images and video, with each able to handle up to one million tokens in a single instance.
Tokens refer to fragments of data, typically parts of words, and a higher token capacity allows a model to process extensive material at once, such as lengthy legal documents, large codebases or hundreds of pages of text.
According to Alibaba, Qwen3.8-Max is built on a mixture-of-experts architecture, which distributes tasks among specialised components rather than activating the full model for every request. This design allows only 95 billion parameters to be used at any given time, lowering operating costs and reducing response time.
Alibaba added that the model successfully completed a software-engineering project within 16 days. Qwen3.8-Max is scheduled for release next week through Alibaba Cloud’s Model Studio platform.
