The Visual Embodied Brain (VeBrain) is a new universal embodied intelligent brain framework proposed by the Shanghai Artificial Intelligence Laboratory and multiple teams in 2025.Visual Embodied Brain: Let Multimodal Large Language ModelsSee, Think, and Control in Spaces".

Traditional robot control usually involves complex sensor input, motion planning, dynamic modeling, etc., which are low-level or intermediate engineering control problems. VeBrain's innovation lies in: it transforms the original complex robot control problem into a "picture-speaking" task that multimodal large language models are good at, so that perception, reasoning and control can be completed under a unified framework, allowing robots to "see, think and act".