On the night before the "soul figure" of Qianwen resigned, Jack Ma and the management team released an AI strategic signal.
```
Author | Huang Yu
While Alibaba's open-source large model frequently tops global developer communities, even earning Elon Musk's praise, an unexpected personnel earthquake quietly unfolded in the middle of the night.
In the early hours of March 4th, the "soul" of Alibaba's Qwen large model, Lin Junyang, suddenly announced his resignation on social platform X, without any warning. This news instantly ignited the global AI sphere. Less than 48 hours earlier, he had just led his team in releasing the Qwen3.5 small-size model series and garnered top attention globally.
Regarding Lin Junyang's resignation, a widely circulated theory is: Alibaba Cloud's senior management made structural adjustments to the Qwen team, parachuting in a new technical lead. Lin Junyang's authority may have shifted, disagreements arose regarding direction, and ultimately he left.
As of publication, Alibaba has not responded.
The departure of this core figure has led the outside world to start scrutinizing the stability of Alibaba's AI talent hierarchy. However, just one day before Lin Junyang's post, Jack Ma, rarely seen in public, led the entire core management of Alibaba and Ant Group in a rare full gathering at Hangzhou Yungu School, founded by Alibaba partners, and publicly discussed the challenges and opportunities of the AI era.
On one side is the "sudden" departure of a core tech guru, on the other is the founder and highest decision-makers collectively endorsing AI—a seemingly contradictory scene that reflects an Alibaba AI landscape in transition: full of uncertainties, yet determined in its will.
AI Team Earthquake
In the current AI world, Lin Junyang’s name represents an extreme "technical legend."
As Alibaba's youngest P10-level technical lead, born in 1993 and just 32 years old this year, he is the main driving force behind the Qwen series of open-source large models at Alibaba.
In the early hours of March 4th, Lin Junyang suddenly posted on social media: “me stepping down. bye my beloved qwen.”
Though he did not clarify whether he is leaving Alibaba entirely, just "leaving the Qwen team" is enough to shake the industry.
After the announcement, numerous AI researchers close to Qwen posted specific emojis in their circles to show respect. Some even raised the slogan “Qwen is nothing without its people.”
Chen Cheng, who describes himself as a Modada community maintainer and Qwen contributor, commented that he was heartbroken, believing the departure wasn’t Lin Junyang’s own choice. The previous night (March 2nd), they released the Qwen3.5 small-size model series together, and he couldn’t imagine Qwen without Lin Junyang.
Such emotional expressions evoke memories of Sam Altman's exit from OpenAI back then.
This change feels especially sudden as on March 2nd evening Lin Junyang had just led the team in open-sourcing four Qwen3.5 small-size models, immediately drawing overseas tech attention. Even Musk parachuted in to comment "incredible intelligence density."
Announcing departure at the peak of glory has led to endless speculation about the reasons, but nothing has been officially confirmed.
Within the industry, Lin Junyang’s departure will surely impact Qwen model's development.
Lin Junyang’s cross-disciplinary education in computer science + linguistics gave him a unique advantage in deep-diving multi-modal large models. After graduating with his Masters in 2019, he joined Alibaba, rooted himself in AI foundational model R&D, quickly rising from an entry-level engineer to Alibaba’s youngest P10 technical expert.
At the end of 2022, as the wave of large AI models rose, Alibaba merged its DAMO Academy language, vision and other AI teams into Alibaba Cloud. The Tongyi Lab was established, led by Alibaba Cloud's CTO Zhou Jingren, and Lin Junyang was formally appointed as lead for the Tongyi Qwen series under Tongyi Lab, responsible for core model R&D and strategic planning, becoming a key figure in Alibaba’s large model track.
Lin Junyang spearheaded the Qwen series open-source drive, propelling Alibaba Qwen into the ranks of top global open-source models with worldwide influence.
Stanford AI Institute’s "2025 AI Index Report" points out that the performance gap between top U.S. and Chinese AI models has sharply shrunk to 0.3%, nearly wiped out, with Alibaba Qwen’s important models ranking third globally in contribution.
Additionally, Nvidia CEO Jensen Huang’s demonstration at the 2025 GTC conference showed Qwen had the highest market share among global open-source models and was still growing.
MiniMax’s official account also commented under Lin Junyang’s resignation announcement: “Thank you for your contribution to the open-source community.”
At the AGI-Next frontier summit in January this year, Lin Junyang, as Alibaba Qwen’s tech lead, engaged in a roundtable dialogue with Zhipu AI founder Tang Jie, and Tencent’s chief AI scientist Yao Shunyu, among others.
Regarding AI's future, Lin Junyang believes: “If your idea isn’t to help all humanity, then don’t bother with LLMs.” He hopes future models are not just exam aces, but Agents that truly help humanity. He firmly believes Embodied AI—the kind that can operate computers, write code, even serve tea in the physical world—is the ultimate form for AI entering reality.
Thanks to Alibaba Qwen’s achievements and reputation these years, even if Lin Junyang leaves Alibaba, he obviously won’t lack new options.
Worth noting, Lin Junyang’s departure isn’t unique. Since 2024, several senior tech managers in Alibaba’s large model team have resigned, including Tongyi Lab’s NLP lead Huang Fei, former applied vision team lead Bo Liefeng, former speech team lead Yan Zhijie, and Zhou Chang, who joined ByteDance to head AI business.
LatePost also reported that on the same day Lin Junyang announced his resignation, Qwen’s post-training lead Yu Bowen also officially left, replaced by Zhou Hao, a former DeepMind senior researcher who joined Tongyi Lab early this year. Qwen Code lead Huibinyuan also left Alibaba in January 2026 to join Meta.
Against the backdrop of intense competition for top AI talent, such high-frequency movement seems to be the new normal.
Where is Alibaba AI headed?
Just one day before Lin Junyang announced his resignation, on March 3rd, the entire senior management of Alibaba and Ant Group gathered at Hangzhou Yungu School.
The lineup was high-profile: Jack Ma, Alibaba Group Chairman Joe Tsai, CEO Wu Yongming, Risk Committee Chairman Shao Xiaofeng, e-commerce group CEO Jiang Fan, Ant Group Chairman Jing Xiandong and CEO Han Xinyi—all were present.
This "all-star line-up" gathering seemed to send a clear signal: even if a few generals change, Alibaba's strategic investment in AI will not waver.
Jack Ma bluntly stated that the AI era has arrived rapidly, its impact on society will exceed imagination, and even said, "None of us are sufficiently prepared."
At present, AI iterates on a weekly basis, its capabilities are still growing. This technical revolution will bring historic changes to productivity and every aspect of society. In the future, perhaps we won’t need to work eight hours a day, social wealth will massively increase, but many job types will disappear. Jack Ma said, “The reason we all gathered at Yungu is to tell everyone: this change is coming fast, we must quickly adapt, help children learn to coexist with AI from now, and adapt to this enormous transformation.”
Though changes in core technical leaders bring uncertainty, Alibaba’s AI confidence comes from its "golden triangle" infrastructure.
Recently, the "Tongyunge" architecture composed of Tongyi Lab, Alibaba Cloud and Pingtouge emerged publicly for the first time. This means Alibaba now has a full-stack self-developed chip, robust cloud infrastructure, and world-leading open-source models—making it one of only two technology companies globally with top capabilities in large models, cloud, and chips. The other is Google.
This Spring Festival, Alibaba heavily promoted its C-end Qwen APP with a “3 billion yuan Spring Festival treat plan,” offering 25 yuan vouchers across food delivery, instant shopping, movie tickets, air tickets, hotels and more, extending voucher validity multiple times until March 3rd.
Morgan Stanley estimates Alibaba's order subsidies alone reached 5 billion yuan (about 200 million orders × 25 yuan), and with brand marketing spend, actual expenditure exceeded 5 billion yuan. In terms of input, Alibaba invested far more than Tencent, ByteDance, etc. Qwen APP DAU saw significant increase, but user retention remains a challenge.
In any case, Alibaba’s determination to promote Qwen APP is beyond doubt.
Alibaba’s AI strategy will not swing wildly due to individual departures. For Alibaba, this technological revolution is not just a competition for productivity, but also a reshaping of the logic of future survival.
Risk Warning and DisclaimerThe market carries risks, and investment requires caution. This article does not constitute personal investment advice, nor does it consider the special investment goals, financial status, or needs of individual users. Users should consider whether any opinion, viewpoint, or conclusion herein suits their specific circumstances. Investment based on this is at your own risk. ```