Tencent is investing unprecedentedly in its large model strategy. Today, the company announced a deep restructuring of its AI R&D system, establishing new departments: AI Infra, AI Data, and the Data Computing Platform Department, to comprehensively strengthen the full-stack capabilities from computing infrastructure, data systems, to model development. This move marks the official upgrade of Tencent's AI strategy from "product-driven" to a systematic approach of "infrastructure + engineering + scenario implementation."

More importantly, Vince Yao, a former researcher at OpenAI, has officially taken office as Chief AI Scientist in the CEO/President's Office, and also oversees the AI Infra Department and the Large Language Model Department. According to multiple sources, Yao graduated from Tsinghua University and Princeton University, and was one of the first contributors to OpenAI's intelligent product Operator and the Deep Research project. He will report directly to President Liu Zhiping and President of the Technical Engineering Business Group Lu Shan, highlighting his core position in Tencent's AI landscape.

The goal of this structural upgrade is clear:

- AI Infra Department: Focus on large model distributed training, high-performance inference services, and other underlying technologies to build an autonomous and controllable AI infrastructure;

- AI Data Department (Headed by Liu Yuhong): Responsible for building and evaluating high-quality data for large models;

- Data Computing Platform Department (Headed by Chen Peng): Build an intelligent computing platform integrating big data and machine learning.

At the same time, Tencent's Huan Yuan large model is iterating at an astonishing speed. Huan Yuan 2.0 was just released on December 5th, and today, Tencent launched the world's first open experience real-time world model, "Huan Yuan World Model 1.5." Its flagship model, TurboS, is the industry's first ultra-large-scale MoE architecture based on a hybrid linear attention mechanism, maintaining an intense iteration pace of one update per month.

Talent competition is also accelerating. It is reported that the Huan Yuan team has recently been actively recruiting key personnel from major companies and the so-called "AI Six Tigers," with the core purpose of supporting intensive technical R&D.

Tencent is not alone. The AI competition among domestic tech giants has become extremely intense:

- ByteDance will hold the FORCE Original Power Conference tomorrow, expected to launch a new product of the Doubao large model. Its "Doubao phone" has already attracted widespread attention;

- Alibaba has established the "Qianwen C-end Business Group," integrating resources from Quark, UC, Shuqi, and AI hardware, aiming to make Tongyi Qianwen the "first entry point for users in the AI era."

However, a senior executive from an A-share AI company told Caigeli said: "Eventually, foundational models and computing power will become common, and the real competitive advantage lies in 'scenario deepening.' Who can deeply integrate into industry workflows and solve real business pain points, will be able to build an insurmountable moat — which is precisely the opportunity for independent third-party vendors."

At this critical turning point where large models are moving from a "technology showcase" to "industrial application," Tencent's structural reorganization is both defense and offense. And Vince Yao, a technical leader from OpenAI's "agent origin," may determine whether Tencent can evolve from a follower to a pioneer in the next generation of AI competition.