Gasgoo Munich- As the automotive sector accelerates its shift toward intelligence and electrification, embodied AI is emerging as the defining variable for the next generation of smart terminals. Alibaba has officially launched the Qwen-Robot series — the first complete embodied AI model family within its Qwen large model ecosystem, according to Gasgoo. Featuring a full-stack technology matrix covering "hands, feet, and brain," the series provides robust technical support for scaling up intelligent agents, including in-car robots.
Image credit: Qwen Large Model
The Qwen-Robot series comprises three core models: Qwen-RobotManip, a vision-language-action (VLA) model focused on fine manipulation; Qwen-RobotNav, a vision-language-navigation (VLN) model responsible for spatial understanding and path planning; and Qwen-RobotWorld, a world model for environmental cognition. This combination effectively equips robots with "dexterous hands," "path-finding feet," and a "thinking brain."
Architecturally, Qwen-RobotManip addresses the weak task generalization inherent in traditional robot control. As a VLA model, it interprets natural language instructions and directly outputs action trajectories, significantly raising the operational ceiling for robotic arms in unstructured environments. Meanwhile, Qwen-RobotNav strengthens the spatial perception and decision-making loop of mobile robots — capabilities critical for logistics AGVs in factory zones and future mobile spaces with autonomous driving features.
Of particular note is Qwen-RobotWorld. By modeling and simulating the physical world, it empowers robots to predict environmental shifts, serving as key infrastructure for high-level autonomous driving simulation tests and autonomous robot learning.
