On March 2, 2026, Alibaba had reason to celebrate. The Qwen team had just released four new small Qwen 3.5 models designed for smartphones and lightweight devices. Elon Musk liked the post and wrote: "Impressive intelligence density". The team could not have wished for better publicity. But a day later, a shock came.
A Reaction That Resembled a Funeral
Lin Junyang posted a short message on X: "me stepping down. bye my beloved qwen." No explanation. No context. Just a farewell to the project he had helped build since April 2023.
Lin joined Alibaba in July 2019. Born in 1993, he was one of the youngest senior technical leaders in the entire company. After the departure of Qwen's previous technology chief, Zhou Chang, Lin took over leadership of the team in 2022 and had directed the project's entire technical strategy ever since. Under his leadership, the Qwen model family grew from the original Tongyi models to Qwen 2.5 and Qwen 3.5 Max. Since 2023, the team has released more than 400 open-source models with parameter counts ranging from 500 million to 235 billion. Then, finally, came that tweet.
His colleagues reacted in a way rarely seen in the corporate world. Wenting Zhao, a research scientist on the Qwen team, wrote that it was the "end of an era." Yuchen Jin, CTO of the startup Hyperbolic, recalled working with Lin's team on model launches, with the team still online at 6 a.m. Beijing time. Tiezhen Wang of Hugging Face called the departure a "huge loss."
One of the project's contributors, Chen Cheng, wrote directly to Lin: "I know leaving wasn't your choice", adding that they had still been working together on the model release just hours earlier. That sentence says more than an entire press release. Or rather, it says more precisely because there was no press release. Alibaba did not respond to requests for comment at all.
Several junior researchers on the team resigned on the same day as Lin. One of them wrote a sentence that spread throughout the AI community: "Qwen is nothing without its people" - a direct echo of the words voiced by OpenAI employees during the company's leadership crisis in 2024.
me stepping down. bye my beloved qwen.
— Junyang Lin (@JustinLin610) March 3, 2026
Behind the Scenes: An Emergency Meeting and Unanswered Questions
In early March, Alibaba called an emergency meeting of the entire Tongyi Lab. Eddie Wu, CEO of Alibaba Group, addressed employees working on Qwen directly, approximately 12 hours after Lin's announcement. Employees, represented by reinforcement learning department head Liu Dayiheng, asked management pointed questions. They asked about the team's restructuring, the arrival of new member Hao Zhou from Google DeepMind, model development plans, and the future allocation of resources.
Management responded. Jiang Fang, Alibaba's chief people officer, admitted that internal communication had been inadequate. She assured employees that the restructuring meant expansion, not cuts. Zhou Jingren, CTO of Alibaba Cloud, faced uncomfortable questions: why could external customers purchase computing capacity without difficulty while internal teams ran into limits? The answer was vague - historical factors, and work was underway to address the issue.
Management's central message was clear: Qwen remains Alibaba's highest priority. The restructuring was reportedly unrelated to internal political disputes. But employees did not entirely see it that way.
Hao Zhou: A Successor, or Just a New Face?
A rumor circulated that Hao Zhou from Google DeepMind would come in to directly replace Lin. The reality is more complicated. His formal role and the organizational structure are still under discussion. Zhou joined Quark in January 2026 and then moved to Tongyi Lab. He reports directly to Zhou Jingren. Many employees believe he will lead Qwen's post-training work.
Zhou has a solid résumé: a bachelor's degree from the University of Science and Technology of China, a PhD from the University of Wisconsin-Madison, three years at Meta AI, and four years at Google DeepMind. He contributed to the development of Gemini 3.0 and worked on systems using multi-step reinforcement learning. On paper, he looks excellent. But replacing someone who built the project from the ground up and was its "spiritual leader" is a different challenge.
The Significance of the Departure
The numbers speak for themselves. Qwen's core team consists of just over 100 people. Including the other Tongyi Lab teams, the total headcount is in the hundreds. For comparison, ByteDance's Seed team, which handles foundation model training, has nearly 2,000 employees. Alibaba was therefore achieving comparable results with significantly fewer resources - precisely because of people like Lin.
One investor put it bluntly: "If this group really leaves, Qwen could be delayed by at least six months to a year while the team is rebuilt and brought up to speed." An AI executive at ByteDance described Lin as "a talent worth at least $100 million". Alibaba is also facing pressure from all sides. The Qwen app launched in November 2025, and the company is trying to catch up with ByteDance, whose Doubao is approaching 200 million daily active users. At the same time, it must keep pace with the development of cutting-edge models that form the backbone of Alibaba Cloud's commercial strategy.
Open Source
Some observers see Qwen's open-source approach as corporate charity. But that is an oversimplified view. Since 2023, Qwen has released more than 400 models for free - and there is clear business logic behind it.
The open models attracted developers, who provided feedback that accelerated development. Some of the companies and universities that deployed the models in production then migrated to paid Alibaba Cloud services. The direct financial benefit is difficult to measure - Meta faces the same problem with its Llama models, in which it has invested billions of dollars. But the reputational and community value is undeniable.
And that is precisely the value Lin Junyang helped build. He was the face of Qwen for the global developer community. He was the one who answered questions, who was online at 6 a.m. during model launches, and who connected the Chinese team with the world.
What Comes Next?
According to sources close to the situation, Alibaba is still in communication with Lin, and his final departure is not certain. As of March 4, the details of his potential departure had not yet been finalized. At the same time, Binyuan Hui, another member of the Qwen team, changed his profile on X to "formerly MTS @Alibaba_Qwen" - but it is unclear whether he has actually left the company or when the change occurred.
Alibaba faces a simple choice: either it quickly shows who will lead Qwen going forward, maintains the pace of model releases, and reassures the developer community - or the story will continue to grow. In the open world of AI, where community trust depends on regular releases, transparent documentation, and a responsive team, every silence is costly.
Can one company keep pace with OpenAI, Google, and Anthropic with a team of one hundred people? Lin Junyang proved that it can. The question is whether anyone else can do it.
Sources: eweek.com and bloomberg.com
