Shares in Alibaba Group (NYSE:BABA) rose 4% in pre-market trading in New York after the Chinese technology group unveiled the most capable version yet of its Qwen artificial intelligence model.
The Hangzhou-based ecommerce and cloud computing company said Qwen 3.8-Max was the strongest model in the series to date.
It carries 2.4 trillion parameters, the internal variables a model uses to store what it has learned, and which serve as rough shorthand for how capable a system is.
The model will be made fully open-source next week, meaning developers will be able to download and adapt it rather than merely access it through Alibaba's platforms.
Alibaba said the release delivered improvements across coding, research and long-running tasks, and that it functioned as a multimodal system able to interpret images as well as text.
On the Text Arena leaderboard, which ranks models on maths, coding and creative writing, Qwen 3.8-Max sits fifth, behind several models built by Anthropic, the US developer of the Claude series.
It ranks second on Vision Arena, which measures reasoning over visual inputs, according to rankings compiled by Alibaba itself.
The launch lands in an increasingly crowded Chinese field, where Beijing has pushed hard for self-sufficiency in advanced computing as it competes with the US.
Moonshot AI, a Chinese startup, recently released its Kimi K3 model with 2.8 trillion parameters.
Alibaba has staked its next phase of growth on artificial intelligence and cloud services.
Eddie Wu, chief executive, said earlier this year that AI-related products would become the main driver of revenue growth in the cloud division.
