WeChat has 1.4 billion monthly active users — a number no other app in China can match.
On June 2, the Financial Times reported that Tencent has moved a step closer to equipping all of them with an AI agent.
What the agent will do
This isn't another chatbot. Tencent is targeting the millions of mini-programs already inside WeChat.
Imagine telling WeChat, "Book a high-speed train to Shanghai for tomorrow, and arrange a ride to the hotel when I arrive." The AI would handle the rest — launching the ticketing mini-program to buy the ticket, then the ride-hailing mini-program to call a car — without you tapping a single button.
Ride-hailing, food delivery, ticket booking, hotel reservations — tasks that once required opening each mini-program individually — the agent would string together and execute automatically.
WeChat holds the densest mini-program ecosystem in China. Others building AI agents have to set up service interfaces from scratch; Tencent already has millions of ready-made services within its own platform.
Timeline
According to current reports:
- Beta testing: targeted for mid-2026, initially released to a small group of users
- Full rollout: planned for the third quarter
There are caveats — if the feature isn't mature enough, it will be delayed. For a product of this scale, Tencent won't rush.
The project has been kept under wraps for some time. Initial development can be traced back to the first half of 2025, and it has been a highly confidential project within Tencent.
Interesting model selection
The most noteworthy aspect: the WeChat team has yet to decide which model to use.
According to reports, they have tested several domestic large models — Zhipu, Alibaba, DeepSeek, plus WeChat's own small model. Tencent's own Hunyuan model has not yet been reported for use.
This is subtle. A company building an AI agent hesitating over whether to use its own model is an implicit admission that, in this specific scenario, its own model may not be the best choice. Tencent has indeed lagged in the large model race — the pace of model releases and benchmark presence have at times fallen behind DeepSeek, Alibaba, and Zhipu.
But Tencent's real strength has never been the model itself.
Tencent's true trump card
Domestic large models have been competing on parameters, benchmarks, and open-source releases — lively, but few have an entry point with 1.4 billion daily active users.
Tencent does.
If the model is lacking, it can be bought, fine-tuned, or swapped. But the attention of 1.4 billion users, plus millions of already-running mini-program services — that cannot be bought. It doesn't matter which model ends up inside WeChat's shell; what matters is the shell itself.
Turning WeChat from an app where you tap to act into an agent portal where you speak and it acts — that is the weight of Tencent's move. If successful, WeChat would become not just chat plus payments, but the central dispatcher for Chinese daily services.
Models can be caught up later. There is only one entry point, and Tencent has always held it.
Whether this plan can be realized will get its first answer from the mid-year beta test.
Sources: CocoLoop, Tencent moves closer to launching AI agent for WeChat's 1.4bln Chinese users (Financial Times); Tencent is said to be developing a top-secret AI agent project for WeChat (TechNode)