Resignation of Lin Junyang: Collision of Technical Vision with Commercial Goals at Qwen

On the morning of March 4, 2026, Lin Junyang posted a short message on platform X saying: “Resigning from work. Farewell to my beloved team,” announcing his resignation from his position as Vice President of Technology at Alibaba and leader of the Qwen team. This message, which took only a few minutes to write, sparked a crisis in the AI world and raised questions about the direction Alibaba’s large model strategy will take.

Resignation with Impact: Events Leading Up to the Explosion

Facts revealed hours after the announcement showed that Lin Junyang’s resignation was not an voluntary decision but came under intense pressure. On the evening of March 3, Lin left a high-level internal meeting amid serious disagreements and immediately resigned from Alibaba. When he informed the Qwen team, several colleagues immediately broke down in tears.

Cheng Qing, a key contributor to the Qwen team, commented directly: “I am truly heartbroken. I know your resignation wasn’t your choice. Yesterday, we were still deploying the small Qwen3.5 model together.” This comment alone revealed that the departure was driven by administrative pressure rather than personal desire.

On the morning of March 4, two other core leaders of Qwen also announced their resignations: Yu Wenbo (subsequent training lead) and He Kaishen (main contributor to Qwen3.5). By noon, Lin posted on Douyin: “Sorry, friends, I won’t be replying to messages today. I really need rest. Qwen brothers, keep working according to the original plan.”

The Model of Conflict: Clash Between Technical Ambition and Business Logic

The question everyone asks: why did Lin Junyang resign? The answer lies in a deep conflict between two opposing visions on how to develop large models and manage the Qwen project.

First Disagreement: Development Structure

Tongyi Labs planned to dismantle the central model led by Lin Junyang. The plan was to split the integrated team—covering pretraining, fine-tuning, multimedia development, and infrastructure—into multiple independent horizontal units. This restructuring would reduce Lin’s direct authority and distribute responsibilities across other teams.

However, Lin believed that true competitiveness in large language model development comes from deep collaboration across all stages. Separating stages into separate production lines, in his view, would significantly impair development efficiency and innovation space.

Second Disagreement: Open Path vs. Commercial Goals

Under Lin’s leadership, Qwen adopted a comprehensive open-source strategy, making it a top global open-source large model. By January 2026, over 200,000 derivative models from the Qwen series had been uploaded to Hugging Face, with over a billion downloads. Qwen had become a global standard for exporting Chinese large models.

But Alibaba’s senior management shifted their evaluation from “building technical influence and open environment” to “direct commercial transformation.” Some senior executives described Qwen3.5 as “an unfinished product not yet refined,” while market pressure from unmet end-user application value (targeting $3 billion) forced them to seek immediate commercial results.

Third Disagreement: Talent Restructuring

Since 2025, Alibaba attracted prominent global AI experts. Hao Guo Hong (IEEE member) joined Tongyi Labs with responsibilities overlapping significantly with Qwen planning. In early 2026, Zhou Hao (former DeepMind researcher) joined directly under the lab director to take on fine-tuning.

This shift from Lin’s “single-core” model to a “multi-force parallel” structure inevitably weakened Lin’s influence and sent a clear message: the course is changing, priorities are shifting.

Effects of Resignation: Who Is Leading Qwen Now?

Lin Junyang’s departure, along with other leaders, leaves Alibaba facing immediate and long-term challenges.

Short-term: There is currently no one who can fully replace Lin. His managerial responsibilities will be distributed among multiple teams, which may directly impact the pace of Qwen’s development. Within just three months, early founding members of the core team—including the CTO, training lead, and coding lead—have resigned.

Long-term: The real impact could be more serious. Observers suggest the most likely scenario is that Lin will launch his own project or join a specialized large-model team, potentially creating a new major player in the global AI race.

As for Qwen itself, losing its core team may lead to decreased morale and innovation, especially since open-source models heavily rely on community trust and momentum.

Resignation as a Strategic Turning Point

What happened cannot simply be understood as an employee’s anger or personal conflict. Lin Junyang’s resignation marks a pivotal point in Alibaba’s AI strategy.

His departure signals that Alibaba is moving away from a phase focused on establishing a global technical benchmark and long-term scientific legacy, toward a new cycle centered on direct commercial transformation and immediate financial returns.

The real question remains: can Alibaba achieve this shift without losing the technical momentum and global influence built by Qwen? The coming days will tell.