WeChat is developing its own model, intensifying the battle for the AI entry point among tech giants | Exclusive

Author | Sijia Xiao, Yuchan Wang
Editor | Qian Qiao, Xuan Yang

Tencent, which is always known for its steady style, has recently made a series of significant moves in AI: On March 6th, it held the "Free Lobster Installation" event; on the 9th, it launched the full-scenario AI agent WorkBuddy; on the 10th, it started the internal testing of the local AI assistant QClaw. But these are just the beginning.

36Kr exclusively learned that WeChat, under Tencent, is trying to develop an independent in-house AI model. Currently, the AI model has completed the construction of its basic capabilities and been given an internal code name, and it is expected to be launched publicly in 2026.

On the one hand, this model will be used to connect to WeChat's mini-program ecosystem to support the development of various AI agents; on the other hand, in the long run, WeChat also hopes to explore application scenarios deeply embedded in the social ecosystem based on the capabilities of its in-house model. For example, based on users' long-term behavior on WeChat, it will explore how to improve efficiency and user experience with AI within WeChat.

A year ago, when Deepseek emerged, Tencent was the quickest to respond. It integrated Yuanbao with Deepseek and drove up the daily active users in a short time. However, for a long time after that, Tencent made few splashes in the AI agent track. It wasn't until the daily active users of Doubao exceeded 100 million at the end of the year that Yuanbao was forced to use its "money power" to fight head-on.

But according to the QuestMobile report, after the Spring Festival event ended, compared with the popularity of Doubao and Qianwen, the daily active users of Yuanbao have almost returned to the level before the event, failing to sustain the growth momentum brought by the red envelope period.

For a long time, Yuanbao's performance has been underwhelming, which was once regarded by the outside world as Tencent falling behind in the AI competition. And internally, Yuanbao also needs to prove itself.
 

36Kr exclusively learned that during an internal executive meeting in December 2025, WeChat summarized several trends for its entry into the AI field:

First, WeChat must have built-in AI tools that do not rely on third-party systems.

Second, WeChat essentially consists of three parts: interpersonal social interaction, information acquisition, and efficiency tools. Among them, social relationships themselves cannot be replaced by AI, and the role of AI can only occur at the information and efficiency levels.

At the Q3 2025 Tencent earnings conference, President Liu Chiping also publicly stated that WeChat's ideal blueprint is to eventually launch an AI agent: "WeChat's ecosystem has a communication and social ecosystem that enables the agent to understand users' needs, intentions, and interests; it has a content ecosystem, including official accounts and video accounts; it has a mini-program ecosystem that basically covers most use cases on the Internet; it has a business ecosystem that allows people to buy goods, and a payment ecosystem that allows people to complete payments almost immediately. So, it is almost the ideal assistant for users, understanding their needs and being able to perform all tasks within this ecosystem. That's the ideal blueprint."

Different from Qianwen and Doubao, which spend billions or even tens of billions on promotion fees in an attempt to create an AI entry point, as the only national-level app with a monthly active user base exceeding 1.4 billion, WeChat's greatest advantage lies in the fact that it itself is the most stable and natural super entry point.

More importantly, there are two scarce assets behind this entry point: a vast social network and a mini-program ecosystem covering e-commerce, transportation, local life, and even government services. This means that if WeChat's AI agent can connect the entire social link and the mini-program service system, it can directly dispatch and use these capabilities based on the existing traffic. For other giants still burning money to acquire user growth, this could be the most terrifying competitor.

However, precisely because of this, the path for WeChat to build an AI agent is even more difficult.

The most core ability of an AI agent is to understand users' behavior and context. And the most important foundation for a social tool is trust. Users are willing to chat, discuss work, and exchange private information on WeChat because they default to the basic premise of content data security. Once AI starts to intervene, it's hard to guarantee that users won't have a resistance in terms of privacy. The aforementioned insider told 36Kr that how to make users feel that the AI is just a tool rather than a "peeper" when using data to empower AI has always been an important consideration for WeChat in the process of developing the agent. Therefore, every step must be taken with extreme caution.

Since the AI wave hit, the deepest anxiety within Tencent has always pointed to the same question: Will there be a super product that can subvert WeChat, just like WeChat replaced SMS and address books back then? After all, Tencent's core moat has long been built on social and gaming.

But at least for now, this subversive moment has not arrived. No matter how technology changes, the human needs for "connection" and "belonging" will never disappear. In other words, AI still cannot replace real interpersonal relationships. And in this regard, WeChat still holds an unshakable advantage.

WeChat's decision to layout its in-house model is both a defense and a reinforcement. To some extent, this also conforms to Tencent's consistent style - the first product launched may not be the final highlight. "Just like in the short-video era, people thought it was Weishi, but its real competitive product later was the video account," the aforementioned person explained to 36Kr.