Qwen/Qwen-Drive-1.0-4B · Hugging Face

Qwen has released Qwen-Drive-1.0, a finetuned version of their 3.5 4 model, specifically designed for autonomous driving. This vision-language foundation model integrates 3D perception, visual question answering, and motion planning within a unified framework. It features an external bird’s-eye-view (BEV) perception head for 3D object detection and map segmentation, and a Planning Expert for generating future ego trajectories. Experiments show strong 3D perception and competitive motion-planning performance while preserving general vision-language capabilities.