Alibaba's model is leading, but it lags far behind its competitors in the commercialization of AI. Its slogan is at the forefront, but few have truly implemented it, which is the most regrettable. Because the commercialization of Alibaba Cloud was not smooth, it is really outrageous to blame the Qwen team, a team responsible for developing large models, for the commercialization indicators. Let's see how bytes are made. It is the most thorough AI in the Internet 1. ByteDance's volcano engine ranks first in China in terms of token call volume, far surpassing Alibaba. The logic of ByteDance is to make all products AI based, supporting the entire group with a huge internal loop. The more tokens are used, the more obvious the group's efficiency will be, converting token consumption into real value growth. 2. Seed is responsible for (basic model research), which is consistent with the functionality of Tongyi Laboratory. But the core research is centralized and the application exploration is decentralized, not directly facing end users, but like a laboratory, providing the strongest, most cost-effective, and smartest foundation for the entire company. 3. The launch of doubao is far less than that of Qianwen and Yuanbao, and more depends on product growth. Even the probability of seeing Qianwen and Yuanbao advertisements in Tiktok is greater than that of doubao. The product features unique features such as voice calls, real-time videos, dialects, etc. 4. The AI functions of Yimeng and Jianying have also had a clear monetization path since their launch, and have never fallen behind in terms of functionality. Jianying and CapCut have already generated revenue of over 10 billion yuan, and the capabilities of AI have accelerated this process while maintaining competitiveness. 5. Feishu is almost a must-have tool for everyone in the AI industry, and it has become a recommended platform for users to access even though OpenClaw does not support Feishu. Feishu's intelligent agents, multidimensional tables, and Feishu Miaoji are almost all embedded into the entire workflow around the pain points in AI based office work. 6. There are also various AI products that have almost achieved top rankings in their respective fields, such as Trae for programming, Coze for Agent development platforms, Doubao mobile phones, and recording beans, all of which are well-known. On the other hand, how did Alibaba do it 1. Taobao/Tmall, patch based upgrades, and AI fitting rooms have not been able to solve the pain point of how to help users select the most suitable products. 2. Alibaba Cloud's AI transformation is not enough and very difficult to use. The coding plan follow-up is too slow, and Alibaba Cloud and Qwen teams are working together, burdened with commercialization KPIs. The model is intended for use by the group, but even the top-level experts have not figured out the specific application scenarios, resulting in Lin Junyang becoming the chief scapegoat recently. 3. DingTalk has indeed boosted the morale of the original founder Chen Hang (without any strategy). Feishu is knowledge automation (solving how to do it), while DingTalk is process automation (solving whether to do it or not). It's not about liberating productivity, Feishu can truly become the foundation of productivity, DingTalk is still a tool for managers, not about reducing the burden on employees and improving efficiency. 4. Gaode Map and Qwen have endowed Gaode with strong spatial semantic understanding, but they have applied these abilities to the street sweeping list. In order to help the group seize the local life market, they have not created real AI native navigation revenue.. 5. Qianwen, subsidies have brought about the short-term outbreak of DAU, but it is totally following the growth routine of the old Internet era. The growth methods lack retention logic, and although there has been an increase in data, it seems that no new revenue has been created Alibaba's regret is that for a long time, it had the best brain in China (Qwen), but was equipped with a torso that suffered from severe department walls and KPI anxiety. The slogan was loudly shouted, but the actual implementation of AI has not been greatly improved, and it has no ability to generate blood. If I were to say that Alibaba is better off splitting Qwen and developing independently, then in the capital market, minimax (252.6 billion yuan) and Zhipu (237.2 billion yuan) can still be worth one tenth of Alibaba's market value. Now, the risk of being harmed by the group's aging within Alibaba is even greater.