Junyang Lin, the former head of Alibaba Group’s Qwen (Tongyi Qianwen) AI model, has officially concluded his tenure at the company. On the early morning of March 7, 2026, Lin posted an emotional farewell on his WeChat Moments, marking the end of his journey with Qwen.
“Not until these past few days did I realize how many people in this world love me. Today is my last day, and when everyone applauded for me, I really held back my tears. No matter what others say about me, at least in my heart, I truly feel I did my best for my brothers, for Alibaba Cloud, and for the Group, although many things weren’t done perfectly—sorry.”
In the comments section below the post, he added: “For the company and for Qwen, this is all I can do. You all must keep going!”
Lin’s departure has drawn significant attention, stemming from his March 4 announcement on social media where he stated: “me stepping down. bye my beloved qwen.”
On the same day, Qwen’s post-training lead Bowen Yu and core contributor to Qwen 3.5/VL/Coder, Kaixin Li, also revealed their resignations. Kaixin Li wrote in his farewell post: (Please modify this quote if needed: “From Alibaba Qwen, signing off. Grateful to work with such amazing people, proud of our impact. Onward!”)
Earlier in January, Binyuan Hui, the former Qwen Code lead and initiator of OpenDevin, had already left to join Meta Platforms, Inc.
Alibaba’s Official Response
On March 5, Alibaba Group addressed Lin’s resignation through an internal email from CEO Eddie Yongming Wu. The email stated: “The company has decided to approve Junyang Lin’s resignation and thanks him for his contributions in the position. Jingren will continue to lead the Tongyi Lab to advance subsequent work. At the same time, the company will establish a foundation model support group, coordinated by me, Jingren, and Yu Fan to align Group resources in supporting foundation model construction.”
The “Jingren” referred to is Alibaba Cloud CTO Jingren Zhou, who also serves as the head of Tongyi Lab.
However, this email did not quell external analyses and speculations regarding the concentrated departures of multiple core technical personnel from Qwen, particularly concerning potential divergences between Lin and Alibaba.
According to Alibaba insiders, Lin’s resignation is related to Qwen’s overall strategic shift. The company believes it needs to recruit more technical talent, which to some extent involves adjustments to Lin’s original scope of responsibilities. After multiple rounds of communication, Lin chose not to accept and proactively proposed resignation.
In response to the escalating discussions about “Qwen model core team collective resignation” and “open-source strategy adjustments,” Alibaba Group stated on the evening of March 5: “Currently, the Qwen model team is stable, with no ‘collective resignation’ occurring. All products and services are operating normally. Qwen will adhere to its open-source strategy. The foundation model team has never been set DAU or other commercialization KPIs. The goal of the Qwen large model is to continuously pursue the upper limit of model intelligence and achieve AGI.”
Junyang Lin’s Background and Contributions
Early Life and Education
Born in 1993, Junyang Lin was Alibaba’s youngest P10-level technical expert. He earned his undergraduate degree in computer science from Peking University and his master’s degree from Peking University’s School of Foreign Languages. Lin’s research areas cover natural language processing and multimodal representation learning, with multiple papers published in top conferences such as NeurIPS, ICML, and ACL.
Career at Alibaba
After graduating in 2019, Lin joined Alibaba DAMO Academy as a senior algorithm engineer, participating in the development of ultra-large-scale pre-trained models M6 and the general unified multimodal pre-trained model OFA. At the end of 2022, Alibaba integrated its AI teams from language and vision at DAMO Academy into Alibaba Cloud, establishing Tongyi Lab. At that time, Lin was formally appointed as the technical lead for the Tongyi Qianwen (Qwen) series of large models.
Industry Reactions and Talent Hunt
Lin’s exit has rapidly ignited an industry-wide talent competition. A related executive from Google DeepMind’s development team, Omar Sanseviero, called out to the Qwen team on social media: “if any of you want a new home to build great models and contribute to the open models ecosystem, please reach out! Lots of exciting things in the roadmap and so much to build ahead of us”.
Another large model company, Zhipu AI, recently included a special note in its latest GLM team recruitment: “high-priority interview fast-track for certain major factory’s certain team.”
Developers have lamented “the end of an era,” while Hugging Face’s Asia-Pacific ecosystem lead, Tiezhen Wang, described the change as a “huge loss.” Qwen project-related practitioners have left messages of tribute, filled with regret.
Currently, Junyang Lin has not announced his next career move. In the fiercely competitive AI technology landscape where iterations are measured in weeks, balancing strategic expansion with talent retention, and coordinating open-source ideals with commercial realities, is not only a challenge Alibaba must address but a common issue faced by all technology companies.
