A major shakeup of core talent at Alibaba raises the question: Is Alibaba's grand model doomed?

Author: Amelia , DeniseI Biteye Content Team

Just after the Lantern Festival, the Tongyi Qianwen team experienced a major shake-up in its core personnel: Technical lead Lin Junyang resigned, along with three other key figures: Yu Bowen, head of Qwen post-training; Hui Binyuan, head of Qwen Code; and Li Kaixin, a core contributor to Qwen3.5, VL, and Coder.

This was not just a typical departure of a technical lead, but a systemic conflict concerning organizational structure, resource allocation, and open-source strategy. Biteye attempts to reconstruct the full picture of this personnel upheaval and to ask a more fundamental question: In the AI ​​era, how should large companies truly position their technological ideals?

I. Blood Loss Overnight: Core Staff Flee

Less than 24 hours after the release of the Qwen 3.5 mini-model, which had just been praised by Elon Musk for its "amazing intelligence density," Lin Junyang, the technical lead for Alibaba's Tongyi Qianwen platform, posted a brief farewell message on X in the early hours of the morning:

As of press time, the post has received over 11,000 likes and over 4.5 million views, with the comments section filled with heartbreak.

Lin Junyang, Alibaba's youngest P10-level technical expert at the age of 32, has passed away.

Lin Junyang's resume is a typical example of China's new generation of AI technology talents.

• Cross-disciplinary background: Born in 1993, he earned his undergraduate degree in Computer Science from Peking University, but chose Linguistics for his master's degree. Perhaps it is this experience, which is different from that of AI elites, that gives him an extraordinary intuition for multimodal and semantic understanding.

• Ali Feiyue: Joined DAMO Academy in 2019, leading the research and development of OFA and Chinese CLIP.

• In charge of Qwen: He became the head of Tongyi Qianwen in 2022, and in 2025, at the age of 32, he was promoted to the youngest P10 in Alibaba's history.

Three other people followed him. Yu Bowen, the training director of Qwen, also resigned. A few hours later, Hui Binyuan, the Qwen Code director, posted "me too" and changed his profile to "former Qwen".

A few hours later, Kaixin Li, a core contributor to Qwen3.5, VL, and Coder, also announced his resignation and changed his Twitter profile to "Pre-Qwen".

This once-star team, which has generated over 1 billion downloads globally, over 200,000 derivative models, and consistently ranked first among open-source large models, seems to be disintegrating at a visible rate.

II. The Root Causes and Dilemmas: The Game Between Individuals and Large Companies in the AI ​​Era

A tweet from Qwen team member @cherry_cc12 revealed just the tip of the iceberg of this turmoil. As information from internal meetings gradually leaked out, we pieced together the full picture of this mass exodus.

2.1 Organizational Dilemma: From Special Forces to Assembly Line

The author speculates that the original Qwen Lab was a crack team of tech geeks, each a special forces soldier, a jack-of-all-trades. Lin Junyang was like a reinforced company commander, leading everyone into battle. However, rumors circulate online that the Qwen team plans to split up, transforming from a "vertically integrated" system covering different training processes and modalities into separate horizontally specialized teams for pre-training, post-training, text, multimodal training, and other areas.

This is actually how traditional internet companies operate. I guess Alibaba thought this way: Qwen Labs was initially an internally incubated project, but things changed after the New Year; they wanted to start applying the incubated projects on a large scale. How to improve efficiency? By breaking down each step into Standard Operating Procedures (SOPs), improving the efficiency of each step, and thus improving overall efficiency.

This idea is definitely outdated. If you look at how big a fuss OpenClaw has made, you'll know that the way games are played in the AI ​​era has really changed.

2.2 The resource dilemma: Do we have resources or not?

On one hand, it's said that "Qwen is the most important thing for the group," while on the other hand, Wu Ma says, "Resources are scarce and everyone is not satisfied." This contradictory stance is reminiscent of a leader who makes empty promises but never delivers. What does it mean that "Qwen is the top priority"? What does it mean that "we've done our best as a Chinese CEO"? What does it mean that resource constraints are due to "a problem with the information transmission process"?

Who are they trying to fool? There are only two possibilities.

First: The higher-ups don't actually value Qwen that much; their investment in Qwen is just an AI FOMO (Fear of Missing Out).

Second: The top leadership is divided into two factions, one of which attaches importance to it and the other does not. The faction that does not attach importance to it starts to restrict and hinder various aspects of development.

In short, it's clear that some senior management only pay lip service to the importance of this product. As a result, even basic resource guarantees for what is supposedly the highest priority product line cannot be guaranteed.

2.3 The interplay between individuals and platforms: Who can rise above the organization?

The most heartbreaking part of the leaked information was the HR's statement: "We cannot elevate someone to a pedestal, and the company cannot accept irrational demands to retain them at any cost."

Are you right? AI companies are already fiercely competing for talent: In 2024, Zhou Chang, the former key technical figure at Qwen, left to start his own business and then quietly joined ByteDance's Seed team, receiving a "sky-high offer" of a 4-2 level position plus an eight-figure annual salary. In 2025, Meta offered a staggering $200 million compensation package to snatch Pang Ruoming from Apple, including not only high stock options but also milestone incentives directly linked to technological breakthroughs. Didn't this HR do any competitor research?

Do you think they're wrong? This statement seems to contain a philosophy of life that has been passed down in China for thousands of years: an individual cannot be above an organization.

2.4 Political Struggle: Whose Party Are You On?

Internally, it was said that "political factors were not considered at all," but then it was also said that "we needed to consider where to place Zhouhao for maximum efficiency." This is quite interesting. The implication seems to be that Zhouhao absolutely must be placed in this organization; the only question is where.

Anyone who's watched palace dramas knows that it's not who gets things done that matters, but who obeys. To put it bluntly, for most professional managers, whether someone can actually solve problems and whether that person threatens their position are two equally important things. In a startup, you can jump as high as you want; in a large company, the sense of security your superiors provide might be more important than your abilities.

Think about it carefully.

2.5 The Mismatch Between Open Source and Commercialization

A deeper tension stems from the misalignment between open-source and commercial paths. Qwen has built a tremendous reputation in the global open-source community – its download numbers, derivative models, and international recognition are all quite high.

But open source didn't bring users and revenue. Now that Qwen has grown big, the group naturally asks, "I've invested so much, shouldn't you give me something in return?"

III. Reflections: The AI ​​Dilemma of Large Companies

Actually, I'm not surprised this happened at Alibaba. You've all seen the skit "The Annual Meeting Can't Stop," right? It's a script based on Alibaba. There's a classic line in it: "If you can't solve the problem, solve the person who raised it."

Alibaba's logic should be: Qwen will continue to function even if someone is missing.

The statement, "What we're doing is very ambitious; 100 people are definitely not enough, we need to expand," makes it seem like not only Alibaba doesn't understand AI, but AI also can't understand Alibaba. Even Web3 next door was amused.

The internet age is about platforms empowering individuals and pursuing standardized, streamlined, and replicable organizational structures. Individuals depend on platforms, and platforms define the rules.

The AI ​​era is evolving into a time when super-individuals possess greater bargaining power and are even redefined as platforms. AI innovation relies on a "special forces" model characterized by small teams, high density, and rapid iteration.

When large companies attempt to manage the creativity of the AI ​​era using the organizational logic of the internet age, conflict is almost inevitable. Behind the chaotic organizational changes lies a collective confusion within these companies about how to manage geniuses.

When HR asks employees, "What price do you think you're willing to pay?" those who truly have the power to shape the future have already voted with their feet.