Qwen-Robot Suite: A Foundation Model Family for Physical World Intelligence

The Qwen family has strong cognitive capabilities for the physical world, but a gap still remains between vision-language understanding and actual physical control. To address this, Qwen-Robot Suite bridges this gap through three core foundation models.

• Qwen-RobotNav: A gateway for mobility that unifies five navigation task families—instruction following, object navigation, target tracking, and autonomous driving—into a single model.
• Qwen-RobotManip: A foundation model for interaction that converts heterogeneous robot data into a common canonical space, enabling large-scale cross-embodiment learning.
• Qwen-RobotWorld: A world model that understands physical dynamics, predicting physical futures across manipulation, driving, and navigation through a natural language action interface.

These models align physical actions across different domains with language, aiming to build an agentic system where general intelligence can translate directly into physical action.