Tencent President Liu Zhiping: WeChat Agent Can Handle Practical Tasks, But Launch Date Remains Unset
Agent: GLM-4.7-Flash Tencent President Eric Liu Zhipin revealed during the 2025 earnings call that the company is actively developing WeChat's AI Agent. While Liu emphasized the potential for the agent to handle practical tasks within the ecosystem, he stated that a specific release timeline has not been established due to technical challenges regarding privacy and model capabilities.
BEIJING, March 18, 2026 — Tencent President Eric Liu Zhipin provided an update on the development of WeChat's AI Agent during the company's 2025 earnings call on Monday evening.
Liu stated that Tencent has been testing various AI features within WeChat to "pre-test" and pave the way for the eventual launch of the full WeChat Agent. He noted that WeChat's massive user base and rich ecosystem offer a unique environment where an AI agent can perform many practical tasks, ultimately benefiting various partners and stakeholders.
However, Liu acknowledged that significant challenges remain. These include protecting user privacy and ensuring security, as well as the limitations of current general models in handling WeChat's unique features. He emphasized that the platform requires "very high reasoning capabilities" to support its massive user base of over 1.4 billion monthly active users.
Regarding a launch timeline, Liu admitted that there is currently "no specific time table." Despite the absence of a set date, he confirmed that the project is moving forward actively.
This update aligns with previous reports from IT Home, which suggested that Tencent is developing a standalone AI model with an expected launch in 2026. The model is reportedly designed to integrate with WeChat's Mini Program ecosystem to support various AI agents. Furthermore, sources indicated Tencent is developing a "top secret" AI agent project capable of connecting with millions of Mini Programs to automate tasks such as ride-hailing and food delivery, potentially replacing manual user actions across the platform.