Alibaba's Qwen AI team has been a powerhouse in the world of machine learning, consistently delivering cutting-edge models that have garnered international acclaim. However, a recent development has left many questioning the future of this innovative team.
Just a day after releasing the open-source Qwen3.5 small model series, which received praise from none other than Elon Musk, the project's technical architect and several key members have departed Alibaba under mysterious circumstances. This sudden exit has sparked concerns and raised eyebrows worldwide, especially given the team's focus on open-source initiatives.
Among those who left are Junyang "Justin" Lin, the technical lead who transformed Qwen from a lab project into a global phenomenon with over 600 million downloads, and his colleagues, Binyuan Hui and Kaixin Li. Their departure marks a significant turning point for Alibaba Cloud and its role as a leader in international open-source AI.
While the company celebrates a technical achievement, the absence of its core leadership suggests a growing divide between the researchers and the corporate hierarchy, which seems to be prioritizing aggressive monetization strategies.
The departing researchers' final contribution, the Qwen3.5 small model series, is a testament to their expertise in "intelligence density." These models, ranging from 0.8B to 9B parameters, utilize a unique Gated DeltaNet hybrid architecture, allowing them to rival much larger systems in terms of reasoning capabilities. By employing a clever attention mechanism, the models maintain an impressive context window while remaining efficient enough to run on standard devices, even web browsers.
Lin, a graduate in humanities and a polyglot, has long advocated for this "algorithm-hardware co-design" approach, aiming to bypass computational constraints. His philosophy, detailed at the Tsinghua AI Summit, has now left a lasting impact on the developer community.
For enterprises, the leadership vacuum created by these departures is a cause for concern. Many companies chose Qwen for its unique offering - the performance of a proprietary US model with the transparency of open weights. Alibaba's recent consolidation of its AI efforts into the "Qwen C-end Business Group" further highlights its intention to transition Qwen into the operating system for AI-integrated wearables.
However, the appointment of Hao Zhou, a veteran from Google DeepMind's Gemini team, to lead the Qwen project suggests a shift towards a more metric-driven leadership style. Industry analysts warn that as Alibaba strives to meet investor expectations, the openness of Qwen's models may take a backseat, similar to what happened with Meta's Llama AI model release.
Enterprises relying on Qwen's Apache 2.0-licensed models now face the possibility of future flagships being locked behind paid, proprietary APIs. The takeaway? If you value Qwen's open-source efforts, download and preserve the models now.
The internal tensions at Alibaba mirror those seen at OpenAI and Google, where the soul of the machine often clashes with the scale of the business. The departure of Junyang Lin, who served as a bridge between China's engineering talent and the Western open-source ecosystem, is symbolic of this struggle. There are fears that without his advocacy, Qwen may retreat into a "walled garden" strategy, similar to its Western rivals.
The heartbreak of the Qwen team's departure is palpable on social media, with sentiments of mourning rather than celebration. One contributor, Chen Cheng, alluded to a forced departure, expressing their sadness and imagining a Qwen without Lin.
Internal reports from Alibaba's "Tongyi Conference" on March 4 suggest an atmosphere of organizational tension. Executives defended the departures as a result of fundamental disagreements over AI development. The primary catalyst appears to be a shift away from the vertically integrated R&D model championed by Lin, towards a more horizontally managed approach.
Leadership, including CEO Wu Yongming, defended their decision, stating that while Lin's centralized efficiency was undeniable, the project's scale required a different governance structure. The Chief HR Officer reportedly struck a firm tone, stating that the company cannot accept irrational demands to retain Lin, signaling a shift towards a more traditional corporate culture.
As Alibaba prepares for its fiscal Q3 earnings report, the focus will likely be on efficiency and commercial scale. While enterprises may benefit from the cost reductions promised by Qwen3.5, the larger AI community may lose the most vibrant open-source lab in the East.
The world now watches to see if Qwen will continue to be a model for the world or become just another component in Alibaba's corporate strategy. What do you think the future holds for Qwen's open-source AI efforts? Share your thoughts in the comments!
