Tencent's big model faces a major personnel earthquake: Yao Shunyu takes the helm of the basic model. Recently, Tencent's Hunyuan multi-modal team has undergone a deep personnel reorganization and strategic adjustment.

Tencent's big model faces a major personnel earthquake: Yao Shunyu takes the helm of the basic model. Recently, Tencent's Hunyuan multi-modal team has undergone a deep personnel reorganization and strategic adjustment. Lin Xudong, the former head of xAI multi-modal understanding, has officially joined Tencent Hunyuan as the head of multi-modal content generation algorithms. This change is not only accompanied by a series of intensive adjustments such as the departure of Hu Han, the former head of multi-modal understanding, to start a business, and the joining of former OpenAI researcher Tian Yonglong, but also marks the ongoing restructuring of Tencent's multi-modal large model development route. Looking back at Tencent Hunyuan's past multi-modal development history, its technology landscape mainly revolves around several core products. Among them, the Hunyuan video model HunyuanVideo became the world's largest open source video generation model when it was first launched at the end of 2024. Later, in 2025, it intensively expanded its capabilities such as Hunyuan video, customized generation, and digital human driving. In terms of image generation, from the early HunyuanImage to version 2.1 that supports native 2K resolution and version 3.0 that supports 80 billion parameters, it has continued to evolve towards the industrial level. In addition, Hy3D, which focuses on single 3D models, is also widely used in e-commerce modeling and product design. In the field of spatial intelligence, Tencent has also launched the world's first open source, simulation-supported immersive 3D world generation model Hy World 1.0 and its subsequent versions, which is deeply in line with the research direction of spatial intelligence. However, as Tencent merged the large language model department and the multi-modal model department to form the "Basic Model Department" and Yao Shunyu fully took over, Tencent's multi-modal R&D ideas are facing profound changes. On the current technical roadmap, the industry has different views on the relationship between multi-modality and intelligent main lines. Li Feifei's World Labs and others believe that the world model is an important main line of AGI as part of spatial intelligence, while DeepSeek and others advocate using visual modality as a tool to serve language models and weakening the independent 3D or world model route. Judging from Yao Shunyu's recent strategic layout and the launch of his flagship product Hy3, he is more inclined to the latter technical route. Yao Shunyu has emphasized many times in public that the competition barrier for AI in the second half lies in context and reasoning capabilities, not simply the number of model parameters. By deeply integrating multi-modal understanding capabilities into the main line of basic models, optimizing tool call stability and long-term context acceptance, Tencent is committed to improving the understanding and execution efficiency of models in real office, GUI Agent and other products. The joining of Lin Xudong is to further solve pain points such as inconsistency and lack of spatial stability in the generated model by improving multi-modal understanding capabilities. This reorganization not only reflects the alternation of Tencent's internal technical routes, but also demonstrates its firm focus on the main line and productivity implementation in the second half of the AGI competition.