Search
The Ali Qianwen team has proudly open-sourced the industry's pioneering visual language foundation model tailored explicitly for autonomous driving scenarios—Qwen-Drive-1.0-4B. Built upon the robust Qwen3.5-4B architecture, this innovative model seamlessly integrates 3D spatial perception and visual question-answering functionalities during its pre-training phase, while also incorporating a sophisticated motion planning module. Adopting a dual-expert system architecture, Qwen-Drive-1.0-4B showcases outstanding performance across a diverse range of autonomous driving tasks and complex scenarios. Moreover, its reinforcement learning iteration significantly boosts closed-loop safety while ensuring minimal trajectory displacement errors, setting a new benchmark in the field.
-
C114 Communication Network -
Communication Home
