March 4, 2026 — Alibaba Group’s top scientist in charge of its Qwen large language model, Junyang Lin, announced his sudden resignation on social media today, stating simply: “me stepping down. bye my beloved qwen.”
His abrupt departure comes amidst a significant internal reorganization of the Qwen development team and follows rare public remarks on AI strategy by Alibaba founder Jack Ma, signaling a period of intense transition for the Chinese tech giant’s artificial intelligence ambitions.
Sudden Exit Follows Leadership Rally
Lin’s self-announcement caught industry observers by surprise. Just one day prior, on March 3rd, Jack Ma convened a rare meeting with the core management of both Alibaba and its affiliate Ant Group at the YunGu School in Hangzhou. Ma emphasized the rapid, historic pace of the AI revolution, stating technological iterations are now measured in weeks and urged the leadership to adapt quickly. The high-level gathering, which included Chairman Joseph Tsai, CEO Eddie Wu, and Ant’s top executives, underscored the strategic importance of AI to the conglomerate.
When contacted by reporters, an Alibaba Cloud spokesperson declined to comment on Lin’s resignation.
Restructuring and Internal Dissent Cited as Factors
According to a report by LatePost, Lin’s departure is linked to a planned reorganization of the Qwen team that would have significantly reduced his managerial scope. Lin reported to Jingren Zhou, Alibaba Cloud’s CTO and head of the Tongyi Lab, the AI research unit housing the Qwen team.
The Tongyi Lab plans to dismantle the current “vertically integrated” Qwen team, which covered different training processes and modalities, and split it into separate, horizontally divided teams focused on pre-training, post-training, text, and multimodal tasks. These teams would still fall under the Tongyi Lab.
This planned disaggregation reportedly conflicted with Lin’s technical vision. Throughout 2025, Lin had repeatedly advocated for tighter integration and communication between pre-training, post-training, infrastructure, and training teams. An anonymous source also claimed that some Alibaba executives were not fully satisfied with the recently released Qwen-3.5 model, privately describing it as a “half-finished product.”
Wider Team Exodus and Rebranding
Lin is not the only high-profile departure. Bowen Yu, the head of post-training for Qwen, has also formally left the company. His responsibilities will be taken over by Hao Zhou, a former senior researcher at Google DeepMind who joined the Tongyi Lab earlier this year. Zhou reports directly to Jingren Zhou.
Furthermore, Bin Yuan Hui, the head of Qwen Code, left Alibaba in January 2026 to join Meta.
The leadership changes coincide with a major brand consolidation. On March 2nd, Alibaba announced it was unifying its AI branding under the single name “Qwen” in English and “千问” (Qianwen) in Chinese, retiring the various names like “Tongyi Qianwen” to avoid confusion. “Tongyi Lab” will remain the name of the overarching AI organization.
Lin’s Legacy and a Pattern of Turnover
Lin, born in 1993, joined Alibaba’s DAMO Academy in 2019 after studies at Peking University. He was instrumental in launching the Tongyi Qianwen model in 2023 and later spearheaded the open-sourcing of the Qwen series. In 2025, he oversaw the release of the flagship Qwen3-Max, a trillion-parameter model that outperformed international peers on several benchmarks.
His exit is the latest in a series of high-level departures from the Tongyi Lab. Former technical lead Chang Zhou left for ByteDance in August 2024; speech team leader and DAMO founding member Zhijie Yan (P10) departed in February 2025; and applied vision team lead Liefeng Bo left in April 2025.
Lin’s departure marks a pivotal moment for Alibaba’s AI efforts as it seeks to navigate internal restructuring and fierce global competition.
Leave a Reply