Alibaba Group CEO Yongming Wu announced in an internal memo that the company has approved the resignation application of Lin Junyang, the core leader of the Qwen project. Alibaba Cloud CTO and head of the Tongyi Lab, Jingren Zhou, will continue to lead the lab and advance its subsequent work.
Concurrently, Alibaba will establish a foundational model support team, to be coordinated by Yongming Wu himself, Jingren Zhou, and Alibaba Group CTO Yu Fan, to pool corporate resources in support of foundational model development.
Wu emphasized that technological progress is a relentless pursuit, and developing foundational large models is a critical future-oriented strategy for Alibaba.
This announcement indicates that Wu will personally oversee resource allocation for foundational models, signaling that the Qwen model team is set to receive increased GPU resource support, which is undoubtedly a positive development for the team.
Alibaba's strategic focus for Qwen has shifted from building technological influence to a dual emphasis on both foundational models and creating a super application gateway. This shift means key performance indicators are moving from academic metrics like open-source community activity and research paper citations to operational metrics such as user retention, scenario penetration, and commercial revenue.
Lin Junyang, aged 33, was one of Alibaba's youngest P10-level technical leaders and a key driver behind the Tongyi Qwen project. He was widely regarded as the "spokesperson" for the Qwen open-source ecosystem. Under his leadership, Qwen adhered to a "full-scale, full-modality" open-source strategy, achieving over 10 billion global downloads and spawning more than 200,000 derivative models. However, the commercial efficiency of open-source models has always been subject to debate.
Lin graduated in 2019 and joined the Alibaba DAMO Academy's Intelligent Computing Lab. A year later, when the Tongyi Qwen project was initiated, he became a core architecture member. He was promoted to technical lead in 2022, led the open-sourcing of the Qwen series in 2024, competing directly with models like GPT and Claude on global leaderboards, and became Alibaba's youngest P10 expert in May 2025.
Lin's departure is related to an organizational restructuring within the Tongyi Lab. The lab plans to split the Qwen team from a "vertically integrated" model into several horizontally specialized teams focusing on pre-training, post-training, text, and multi-modality, each operating independently and reporting directly upwards. This change shifted Lin's role from overseeing the entire chain to managing just one module.
Objectively, Lin, known for his exceptional talent, tended to prioritize his own perspective. The restructuring led to his discomfort. His recent post, "me stepping down. bye my beloved qwen," sparked significant discussion and was seen by some as a form of pressure using his resignation as leverage. In a large corporation like Alibaba, the essence is to replace individual heroism with organizational stability, eliminate functional overlaps, and align with group-level commercial needs.
The internal memo can be interpreted as providing a dignified farewell for Lin.
OpenAI has experienced the departure of up to 10 co-founders, including Ilya Sutskever and CTO Mira Murati, without significantly impacting its position at the forefront of AI. The company is now valued at over $800 billion and recently raised $110 billion from investors including SoftBank, Amazon, and Nvidia.
Recently, Alibaba's Qwen has been active, having just open-sourced four small-scale models in the Qwen3.5 series: Qwen3.5-0.8B/2B/4B/9B. These models inherit the technical capabilities of the Qwen3.5 family, feature native multi-modal training and the latest model architecture, and cater to various needs from extremely resource-constrained environments to high-performance lightweight applications.
On the same day, Alibaba's large model division announced the unification of its B2B and consumer application brands under the "Qwen" name, retiring the "Tongyi Qwen" designation.
In early 2026, the Qwen App announced it had surpassed 100 million monthly active users. It is now fully integrated across Alibaba's ecosystem, including Taobao, Alipay, Taobao Flash Sales, Fliggy, and Amap, enabling AI-powered functions like ordering food, shopping, and booking flights, and is open for testing to all users. The Qwen App stated its goal is to become an AI assistant capable of handling complex real-life tasks, aiming to push the AI industry from an era of "chatting" into an era of "getting things done."
Alibaba's management believes AI development will progress through three stages: "Learning from Humans," "Assisting Humans," and "Transcending Humans." They see large model capabilities as having entered the "Assisting Humans" stage of Agentic AI, deeming the time ripe for a major push into the consumer market.
AI is considered the only field in the next decade with the potential to create another business on the scale of "Taobao plus Alibaba Cloud" for Alibaba, which is a key reason for the company's all-out effort. Currently, the Qwen App is engaged in a battle of attrition against competitors like Doubao and Yuanbao. The era where model teams focused solely on research without market responsibilities is over; technical teams must now collaborate closely with product teams, which is the core reason behind the recent series of internal team adjustments at Qwen.
An industry insider commented that focusing solely on technology without attention to practical application renders even a 2 billion yuan advertising investment ineffective.
Naturally, Lin Junyang's departure is a loss for the Qwen team. Managing such highly talented technical individuals is notoriously challenging, and Alibaba has room for improvement in its approach to technical talent management. Yongming Wu stressed the need to continuously increase R&D investment in AI and enhance efforts to attract top talent, signaling that funding and computing power are sufficient, and technical staff can continue to leverage their strengths.
Wu also stated that technological progress brooks no stagnation, referring both to intense industry competition from rivals like DeepSeek and OpenAI, and the need to prevent internal "technical aristocracy" from resting on their laurels.
Comments