Tencent merges multimodal and LLM teams into new base model unit

Tencent is merging its multimodal and large language model teams into a new base model unit, led by chief AI scientist Vince Yao. The goal is to boost R&D efficiency and push the limits of full-modality AI. Tencent's Hunyuan now covers image, video, voice, and 3D generation. The Hy3 LLM launched in July saw call volume jump over 68 times in its first week, with 60% of WorkBuddy users choosing it. The Hunyuan 3D World Model 2.0, open-sourced in April, generates 3D worlds from text, images, or video, with over 3 million community downloads. The image model 3.0 ranks first among domestic open-source models on LMArena. Yao, a former OpenAI researcher, created the ReAct reasoning-action paradigm.