The researchers introduce Qwen-Drive-1.0, a vision-language foundation model designed to unify 3D perception, visual question answering, and motion planning for autonomous driving. Building upon a pretrained vision-language model (VLM) architecture, the system incorporates an external bird's-eye-view (BEV) perception head that simultaneously executes 3D object detection, semantic occupancy prediction, and BEV map segmentation. This BEV head functions as a probe to extract 3D scene structures from the shared representations. For navigation, a specialized Planning Expert utilizes these shared VLM representations to output future ego trajectories. To balance driving-specific competence with general-purpose intelligence, the authors employ a staged training strategy that integrates driving supervision with general vision-language data. This approach enables the model to maintain broad visual understanding and instruction-following abilities while gaining specialized driving skills. Experimental results indicate that Qwen-Drive-1.0 achieves robust 3D perception and scene understanding while preserving general VLM capabilities. Furthermore, the model demonstrates highly competitive motion-planning performance when evaluated across diverse testing environments, including open-loop, pseudo-closed-loop, and closed-loop scenarios.