1 hour ago
604
Tencent Holdings (HKEX: 0700) has started putting an AI helper inside WeChat, China’s largest social media platform.
The company said on Monday that it is testing Xiaowei inside Weixin, the China-facing version of WeChat, as it tries to keep pace with Alibaba Group (NYSE: BABA, HKEX: 9988), DeepSeek, Zhipu, and other AI players fighting for users, developers, and money in China’s packed AI market.
Tencent called Xiaowei a “native AI assistant” and said the test is running “on a small scale.” People in the trial can talk to the assistant by typing or speaking. They can also use it to contact friends and open mini-programs, the small built-in apps that run inside WeChat.
Tencent did not specify the model behind Xiaowei and also did not provide a complete feature list, which is frustrating but very common practice from Big Tech in order to get publicity without revealing everything about their product.
Tencent uses WeChat’s daily reach to test AI where users already spend time
WeChat is already being used by users for sending messages to their families, paying merchants, transferring money, booking restaurants, using various services, and completing minor tasks that would require multiple applications in a different setting.
That gives Tencent a serious distribution edge. If Xiaowei becomes useful, Tencent does not need to drag users into a fresh app store download. It can place the assistant inside a product people already open all day.
For a company trying to make AI pay, that matters. Investors have been watching Tencent’s WeChat AI plans since last year because the app could become a direct path to new AI services, paid tools, business features, ads, or other revenue streams.
However, Tencent’s leadership has been considering further integration of artificial intelligence within WeChat for several months already. AI agents have been offered all over the tech industry as assistants capable of performing multistep tasks within different applications and services.
There has not been any official information yet regarding the capability of Xiaowei to cope with some more challenging tasks. Earlier this year, Tencent has hired the chief AI scientist who previously worked at OpenAI.
Chinese AI labs are still pressuring U.S. rivals
The threat posed by China could no longer be ignored after January 2025 because of the launch of R1 by DeepSeek and scaring off the U.S. stock market. Almost $1 trillion has been wiped out from U.S. market value as a result of market sell-off. The market sell-off saw Nvidia (NASDAQ: NVDA) fall by 17%, while the Nasdaq Composite fell by 3.1% in one day.
Just last week, Beijing-based Zhipu, also known as Z.ai, announced GLM 5.2 and called it “a step closer to frontier intelligence for everyone.” The model was presented as the strongest Chinese-trained system so far, while running at less than one-tenth of the cost of Anthropic’s Fable 5.
Zhipu even made the model’s weights public. These are the hidden values that make the whole system work. This is how many Chinese AI scientists do things. In most cases, the model is downloadable and executable on your own computer, meaning it is not controlled by the initial laboratory anymore, if not also by the government.
At the same time, there is a completely different cost problem facing companies from America. The costs for AI have become so expensive that some firms now pay thousands of dollars per user. Some set token budgets since all prompts, outputs, and extended chats are costly.
As of June 12, the Trump administration blocked non-Americans from using Fable 5. As for Anthropic being private, it banned everyone from using it after June 12.
For those who are not Americans, the Chinese AI models such as GLM 5.2 can be more appealing because they are cheaper, open source, and not regulated by Washington.
If you're reading this, you’re already ahead. Stay there with our newsletter.
English (US)