Alibaba said on September 22 that it plans to develop artificial-intelligence models with between 5 trillion and 10 trillion parameters, a scale well above its current flagship Qwen 3.8 Max model, which Reuters reported has about 2.4 trillion parameters.
At the Apsara Conference in Hangzhou, Chief Executive Eddie Wu said Qwen 4 is already in training and that later Qwen 4.5 and Qwen 5 models are intended to scale toward the 5–10 trillion parameter range. He presented the work as part of Alibaba's longer-term push toward more capable AI systems, including research into recursive self-improvement.
Alibaba also unveiled the Zhenwu V900 AI chip, developed by its T-Head semiconductor unit. The company described it as China's most powerful AI chip and said it delivers roughly three times the performance of the earlier M890. Reuters reported that mass production is scheduled for the first quarter of 2027. Those performance claims are Alibaba's own; no independent benchmark establishing the claimed advantage was cited in the initial report.
The company said the Zhenwu V900 can support very large clusters, with configurations of up to 500,000 cards for advanced model training.
Alibaba also set a target of expanding Alibaba Cloud's global data-centre capacity to more than 20 gigawatts by 2032. That is a future capacity target rather than capacity already in service.
The 5–10 trillion parameter models are likewise development plans, not completed systems available today. Model size alone does not determine real-world capability, efficiency or safety, which also depend on architecture, training methods, data, compute and deployment choices.
The announcements show Alibaba expanding across the AI stack — models, chips and cloud infrastructure — as Chinese technology companies continue building domestic computing capacity amid U.S. restrictions on access to some advanced chips.