It’s not only the US AI labs that are seeing churn among their researchers — Chinese AI labs are seeing some movement too.
Junyang Lin, the tech lead of Alibaba’s Qwen AI team and arguably the public face of one of China’s most successful open-source AI model families, announced his departure in a brief post on X: “me stepping down. bye my beloved qwen.” The post, from his account @JustinLin610, quickly racked up hundreds of retweets and thousands of likes, a testament to the following he had built over years of being one of the most visible researchers in the Chinese AI open-source community.
The Face of Qwen on X
To followers of the global AI open-source scene, Junyang Lin — who goes by “Justin” in English-language contexts — was not just a researcher working behind the scenes. He was the public communicator, the hype man, and the technical voice of the Qwen project on social media. On X, he regularly announced model releases, shared benchmark results, engaged with developers, and responded to the global community building atop Qwen’s models. In an era when AI labs compete as much on developer mindshare as on benchmark scores, Lin’s active presence on the platform gave Qwen a human face in a way that few Chinese AI projects have managed with international audiences.
His departure leaves a notable void. Alibaba has not yet announced who, if anyone, will step into the public-facing role he occupied. Interestingly, X account Chen Chang, who is a contributor to Qwen, hinted that Lin’s departure wasn’t his choice. “I’m truly heartbroken. I know leaving wasn’t your choice. Just last night, we were side by side launching the Qwen3.5 small model. I honestly can’t imagine Qwen without you,” he said.
Just yesterday, Lin had thanked Elon Musk after Musk had praised Qwen’s small models release, saying it had managed to show “impressive intelligence density”.
Right along side Lin’s departure, another Qwen researcher said that they were leaving. “Signing off from @Alibaba_Qwen. Grateful for the chance to work with such brilliant minds. Proud of our impact. Onwards and upwards!” wrote Kaixin Li.
Another Qwen researcher Binyuan Hui also appears to have left. His X profile says “former MTS at Qwen”.
Building Qwen from the Ground Up
Lin joined Alibaba in 2019 as a Senior Algorithm Engineer working on NLP and multimodal research for search and recommendation. He eventually rose to Senior Staff Algorithm Engineer and, from 2023, served as the formal tech lead of the Qwen team. Before Qwen, he was a core developer on several influential Alibaba AI projects: the massive mixture-of-experts model M6; OFA (Unifying Architectures, Tasks, and Modalities), which was published at ICML 2022 and became a widely cited work in multimodal pretraining; and Chinese-CLIP, which remains one of the most widely used Chinese-language vision-language models, with its GitHub repository accumulating over 2,000 stars.
On Google Scholar, Lin has amassed over 42,000 citations, with the Qwen3 technical report alone drawing nearly 9,000 citations — a remarkable figure for a model technical report and a sign of the research community’s deep engagement with the work his team produced.
Qwen’s Rise to Global Prominence
The Qwen project Lin led grew into one of the most impactful open-source AI initiatives in the world. Alibaba launched the first Qwen model in beta in April 2023, opening it publicly that September. From those early 7-billion-parameter releases, the family expanded rapidly into a sprawling portfolio: language models, vision-language models (Qwen-VL), audio models, math-specialized models, coding models, and eventually the reasoning-focused QwQ series. By the time Qwen3 launched in April 2025, the models had accumulated over 600 million downloads and spawned more than 170,000 derivative models on Hugging Face — surpassing even Meta’s Llama in that metric.
Fortune magazine recognized Alibaba on its 2025 Change the World List specifically for its open-source AI work, and the Qwen models became frequently cited examples of how Chinese AI labs had narrowed the gap with their American counterparts.
A Candid Voice on China’s AI Constraints
Lin was also a notably candid public voice on the structural challenges facing Chinese AI labs. At a high-profile AI summit at Tsinghua University in January 2026 — where he shared a stage with leading figures from Zhipu AI, Moonshot AI, and Tencent — he acknowledged plainly that the US compute infrastructure likely exceeds China’s by one to two orders of magnitude. He noted that while American labs could afford to plow resources into next-generation research, Alibaba’s team was often consuming the bulk of its available compute just meeting delivery requirements. Despite those constraints, he argued, China’s resource limitations had pushed its researchers toward creative solutions, particularly in algorithm-hardware co-design.
He also delivered a keynote at ICLR 2025, walking the global research community through the technical underpinnings of Qwen2.5 and its specialized variants.
What Comes Next
Lin’s post gave no indication of where he is heading next, and Alibaba has not commented publicly on the transition. The timing comes just weeks after the release of Qwen3.5, the latest model in the family, so Alibaba appears to have a steady pipeline in place regardless of the leadership change.
Whether Lin is departing to found a startup, join another lab, or pursue independent research remains to be seen. Given his visibility in the community and his track record, his next move is likely to attract significant attention — both in China and among the international open-source AI community he helped cultivate.