Alibaba’s Qwen Shake-up: Alibaba Cloud CTO Zhou Jingren Steps In After Sudden Exit

Leadership Changes at Qwen (通义千问)

Alibaba (阿里巴巴) has reportedly moved quickly to steady its flagship large language model unit, Qwen (通义千问), after the abrupt departure of a key leader last week. Chinese outlet LatePost, via ifeng, reported that Alibaba Cloud (阿里云) CTO and Tongyi Lab (通义实验室) head Zhou Jingren (周靖人) will temporarily take charge of Qwen’s model organization to streamline resources and accelerate iteration. Liu Dayiheng (刘大一恒), a veteran of the Qwen effort, will continue to lead pre-training and reportedly assume oversight of post-training and coding teams, all reporting to Zhou. Lin Junyang (林俊旸) reportedly resigned suddenly last week.

Who is Zhou? A Microsoft alumnus, he joined Alibaba in 2015, led frontier R&D in iDST and DAMO Academy (达摩院), and worked on deployments at Ant Group (蚂蚁集团) and Taobao (淘宝). Since late 2022 he has served as Alibaba Cloud CTO and head of Tongyi Lab, and in 2025 reportedly became an Alibaba Partner, the company’s top collective decision-making body. Under his watch, Alibaba has leaned into open source and expanded its ModelScope (魔搭社区) platform for sharing models.

Why It Matters

Qwen sits at the core of Alibaba’s AI strategy and is widely used across Chinese academia, labs, and startups. Any leadership change can ripple through that ecosystem. Alibaba has signaled it will stick with an open-source approach for Qwen, a stance that differentiates it from several U.S. peers and could help broaden adoption despite resource constraints facing Chinese model developers.

Geopolitics lurk in the background. U.S. export controls have tightened access to cutting-edge AI chips in China, pushing local players to optimize training efficiency and data curation. Liu’s reputed strength in data-centric training is seen as crucial for squeezing more out of limited compute—an increasingly strategic advantage.

What Remains Unclear

The core collaboration model—keeping pre-training, post-training, and coding closely integrated within one team—reportedly remains intact. But what about adjacent bets? LatePost noted it is not yet clear how Qwen’s work on vision generation, speech models, and embodied AI will be reorganized. As Alibaba recalibrates under Zhou’s interim stewardship, will it double down on its open-source roadmap or pivot toward more proprietary offerings? The answer could shape how China’s AI stack competes globally.