Debunking the "mass resignation" rumors, the consecutive departures of Qianwen members have made Alibaba nervous

Listing | Damo Finance

Alibaba’s key member of the Qwen large model, Lin Junyang, officially announced his departure, sparking widespread market attention.

On March 5th, Alibaba issued a denial regarding online rumors of “Qianwen core team collectively leaving” and “open-source strategy adjustments.”

Alibaba stated that the Qianwen model team is currently stable, with no signs of “collective resignation,” and all products and services are operating normally. Additionally, Qianwen will continue its open-source strategy. The foundational model team has never been evaluated based on commercial KPIs such as DAU. The goal of Qwen’s large model is to continuously push the upper limits of model intelligence and achieve AGI. Alibaba also welcomes top AI talent worldwide and will continue to increase investment to support the Qianwen team.

The day before, Lin Junyang announced his resignation on social media, saying, “me stepping down. bye my beloved qwen.”

Following Lin Junyang’s official announcement, several other members related to Qianwen, including Hui Binyuan, scientist at Tongyi Laboratory, and Kaixin Li, a core contributor to Qianwen, also announced their departure.

Who is Lin Junyang? Public information shows that the 32-year-old Lin Junyang was cultivated by Alibaba as a technical backbone. After earning his master’s degree from Peking University in 2019, he joined Alibaba DAMO Academy to work on large model development. Before his resignation, Lin Junyang had become the technical lead of the Qianwen series large models and Alibaba’s youngest P10.

Moreover, Lin Junyang is regarded as the “spokesperson” for Qianwen in the international developer community—frequently sharing the latest updates on Qianwen through his personal social media accounts and engaging with international developers.

Lin Junyang has not explained his sudden departure, and speculation about his reasons for leaving continues to circulate.

According to Southern Metropolis Daily, an internal Alibaba source said, “The company has not changed its open-source strategy for Qianwen, nor has it evaluated the foundational model team based on commercial KPIs like daily active users. Lin Junyang’s departure is unrelated to these rumors. The actual reason is that as Qianwen evolved from a foundational model to a group-wide strategic project, the company believes it needs to recruit more top technical talent to increase the talent density of the foundational model team. During this process, there was a change in Lin Junyang’s responsibilities, which he did not accept, leading to his resignation.”

After Lin Junyang announced his departure, Tongyi Laboratory held an emergency all-hands meeting, where Wu Yongming, Alibaba’s Chief Talent Officer Jiang Fang, Alibaba Cloud CTO Zhou Jingren, and other executives addressed organizational and team restructuring issues. Alibaba executives stated that the Qianwen foundational model is currently the company’s most important project, and they hope to bring in talent to expand the team. However, the introduction of new personnel inevitably involves organizational changes, which may not have been communicated effectively.

On March 5th, Wu Yongming issued an internal email responding to Lin Junyang’s departure. Wu stated that the company has approved Lin Junyang’s resignation and thanked him for his contributions. Moving forward, Zhou Jingren will continue leading Tongyi Laboratory to advance subsequent work, and the company will establish a foundational model support team, jointly coordinated by Wu Yongming, Zhou Jingren, and Fan Yu to support the development of foundational models.

Wu also emphasized that developing large foundational models is a key strategic focus for the future. The company will continue to adhere to its open-source model strategy while increasing investment in AI research and development and attracting top talent.

Although Lin Junyang’s departure is now confirmed, discussions around Alibaba’s AI business continue unabated.

Alibaba’s AI Narrative Is Changing

Lin Junyang has contributed significantly to Alibaba’s AI development.

In February this year, Alibaba launched the new generation large model Qwen 3.5-Plus. Compared to the 10 trillion parameter Qwen 3-Max, Qwen 3.5-Plus reduces deployment memory usage by 60%, increases maximum inference throughput by up to 19 times, and offers API pricing as low as 0.8 yuan per million tokens. On March 2nd, Alibaba open-sourced the Qwen 3.5 small model, which Elon Musk praised for its “impressive intelligence density.”

According to the latest data from the open-source community Hugging Face, by January this year, downloads of Alibaba’s Qwen series models exceeded 1 billion, with over 200,000 derivative models, ranking among the top in open-source large models.

Under Lin Junyang’s leadership, the Qwen large models have already secured a favorable position in the international open-source large model competition.

However, solely considering open-source large models, they do not directly generate revenue for Alibaba. From a commercial perspective, Alibaba needs an AI-era traffic hub—a super gateway connecting its various businesses. To this end, in November last year, Alibaba upgraded the original Tongyi app to the Qianwen app, positioning it as “the future battlefield of the AI era.”

Previously, the Qianwen app’s presence on the consumer side was limited. Data shows that in January, the app’s monthly active users (MAU) were about 31 million, ranking outside the top 10 in global AI application charts. In comparison, ByteDance’s Doubao had 170 million MAUs.

For Alibaba, the Qianwen app has become a critical “entry-level” business. If it cannot gain traction on the consumer side, the company’s future AI commercialization capabilities may be constrained.

To boost traffic to the Qianwen app, Alibaba has been actively taking measures. In January, the app integrated with Alibaba’s ecosystem interfaces such as Taobao, Alipay, Fliggy, and Amap. In February, it launched a “Spring Festival 3 Billion Yuan Free Order” campaign, allowing users to place AI-assisted orders for milk tea with a single click.

In March, Alibaba unified the branding of its AI services, consolidating names like Qianwen, Tongyi Qianwen, and Qwen to avoid confusion and promote a single core brand: “Qianwen.”

These efforts have paid off. In February, the Qianwen app’s monthly active users surged to 203 million, a month-over-month increase of over 550%, setting a record for the fastest growth of an AI product in a single month.

Amid the rapid growth of the consumer-facing Qianwen business, the foundational large model business is also expected to adapt accordingly. According to Caixin, an Alibaba insider indicated that Alibaba Cloud is pushing for AI large model capabilities to be applied in vertical industries, requiring more application-oriented talent rather than just focusing on underlying models.

Notably, from the arrangements involving Wu Yongming, Zhou Jingren, and Fan Yu coordinating group resources for foundational model development, Alibaba still maintains a strong focus on AI large models.

Wu Yongming is one of Alibaba’s founders and its first programmer. Zhou Jingren, a prodigy from the University of Science and Technology of China, joined Alibaba in 2015 and has served as Alibaba Cloud’s Chief Scientist and CTO. There are reports that he was promoted to Alibaba Partner at the end of last year. Fan Yu (Wu Zeming) is also an Alibaba Partner, having served as Alibaba Group CTO, DAMO Academy Vice President, and currently chairman of Ele.me (Taobao Flash Purchase).

In summary, with leading open-source large models and rapidly growing consumer traffic channels, Alibaba remains highly competitive in AI. Whether it can leverage this momentum to break through traffic bottlenecks will depend on its ecosystem collaboration and resource endurance.