The Qwen4 family was officially announced for the first time. At the opening of the Cloud Intelligence Conference 2026, the new head of LLM at Qwen, Liu Dai Heng, announced the upcoming Qwen4 family. It will include: Qwen4-Max Qwen4-Flash Qwen4-Plus Qwen4-27B In the future, Qwen intends to move on to training models on a scale of 5–10 trillion parameters. In parallel, Alibaba introduced the Zhenwu V900 AI chip, which, according to the company, is three times more productive than the previous generation. On its basis, they expect to build clusters of up to 500 thousand accelerators, and by 2032, the capacity of Alibaba data centers should increase to 20 GW.