Alibaba's Qwen releases open-source model for autonomous driving

Alibaba's Qwen team has released Qwen-Drive-1.0-4B, an open-source model that integrates driving scene comprehension with vehicle movement planning for autonomous driving.

The project was developed in collaboration with Huazhong University of Science and Technology. It utilizes the Qwen3.5-4B model as its foundation for visual and language understanding, adding separate components for 3D perception and generating driving trajectories.

Two versions of the planning component are available: one trained to imitate driving examples and another optimized further through reinforcement learning. Both use the same underlying vision-language model, which maintains its original architecture and ability to answer visual queries.

The project makes its code, model weights, and demo data available under the Apache 2.0 license. This release aims to contribute to the advancement of open-source development in the autonomous driving sector.