Alibaba's models are leading, but in AI commercialization, they are far behind competitors. The slogans are there, but very few have actually been implemented, which is the biggest regret. Because Alibaba Cloud's commercialization has been unsuccessful, they blame the Qwen team, a team responsible for large model development, for the commercial results—it's truly absurd.

Let's see how ByteDance is doing; they are the most thoroughly AI-ized in the internet industry:
1. ByteDance's Volcano Engine consistently ranks first domestically in token call volume, far surpassing Alibaba. ByteDance's logic is to AI-ize all products, supporting the entire group with a large internal cycle. The more tokens used, the more efficient the group becomes, turning token consumption into real value growth.
2. Seed is responsible for ( foundational model research ), with functions similar to Tongyi Laboratory. However, core research is centralized, while application exploration is decentralized. They do not directly face end users but act like a laboratory providing the strongest, most cost-effective, and smartest foundation for the entire company.
3. Doubao's deployment is far less than Qianwen and Yuanbao, relying more on product growth. In fact, the probability of seeing ads for Qianwen and Yuanbao is higher than for Doubao. Features like voice calls, real-time video, and dialect support are unique.
4. AI features in Ji Meng and Jian Ying have had clear monetization paths from day one, and their functionalities have never been neglected. Jian Ying and CapCut's revenue has long exceeded 10 billion, and AI features have accelerated this process while maintaining competitiveness.
5. Lark (Feishu) is almost an essential tool in the AI circle. Even without support from openclaw, it has become a recommended platform for integration. Lark's intelligent agents, multi-dimensional spreadsheets, and Lark Memos are all embedded into the workflow to address pain points in AI-augmented office work.
6. Various AI products are also among the top in their respective niches: programming tool Trae, agent development platform Coze, Doubao mobile, and Recording Dou are all highly acclaimed.

In contrast, how does Alibaba do it?
1. Taobao/Tmall: patchwork upgrades, like Ask and AI Fitting Room, failed to solve the pain point of helping users find the most suitable products.
2. Alibaba Cloud's AI transformation is insufficient and very difficult to use. The coding plan follow-up is too slow. Alibaba Cloud and the Qwen team are jointly responsible for commercial KPIs. Models are meant for group use, but top-level leaders haven't even figured out specific application scenarios. Recently, Lin Junyang became the scapegoat.
3. The original founder Chen Hang (Wu Zhao) indeed boosted morale. Feishu is about knowledge automation (solving how to do), but also process automation (solving whether it was done). It’s not about productivity liberation; Feishu can truly become a productivity foundation or a management tool, not just reducing employee workload or improving efficiency.
4. Amap (Gaode Map): Qwen endowed Amap with strong spatial semantic understanding, but these capabilities were used mainly for street sweeping rankings. To help the group compete in the local life market, it did not create real AI-native navigation revenue.
5. Qianwen: subsidies led to a short-term spike in daily active users (DAU), but it followed the old internet growth model. The growth methods lack retention logic. Data has indeed improved, but it seems no new revenue has been generated.

Alibaba's regret is that, for a long time, it possessed the best brain in China (Qwen), but it was paired with a body suffering from severe departmental silos and KPI anxiety. The slogans are loud, but very few AI implementations have actually taken root, and it has no real blood-making capability.

In my opinion, Alibaba should spin off Qwen for independent development. In the capital market, Minimax (2526 billion ) and Zhipu (2372 billion ) could be worth a tenth of Alibaba's market cap. Currently, the risk of being harmed by internal group politics and KPI anxiety within Alibaba is even greater.