Alibaba Ships Qwen-Drive-1.0-4B for Self-Driving Under Apache 2.0
TL;DR
- Alibaba's Qwen team released Qwen-Drive-1.0-4B, a driving-scene and trajectory-planning model built on the Qwen3.5-4B vision-language base.
- Two planning variants ship — one trained on driving examples, another further optimized through reinforcement learning.
- Code, weights and demo data are on GitHub under Apache 2.0; the work was co-developed with Huazhong University of Science and Technology.
Alibaba's Qwen team has released Qwen-Drive-1.0-4B, "an open-source model that combines driving-scene understanding with planning a vehicle's movements," per TechNode, built on the Qwen3.5-4B vision-language base and co-developed with Huazhong University of Science and Technology.
Under the hood, a BEV perception head handles 3D object detection, semantic occupancy prediction and BEV map segmentation, while a planning expert conditions on the shared VLM representations to generate future ego trajectories. Two planning variants ship, "one trained on driving examples and another optimized through reinforcement learning," and the base VLM keeps its visual question-answering ability intact.
Code, model weights and demo data are on GitHub under Apache 2.0. It lands in the middle of a steady run of Qwen open-source releases we have tracked this quarter, one of 74 Alibaba items in our last 90 days.
Originally reported by technode.com
Read the original article →Original headline: Alibaba Ships Qwen-Drive-1.0, an Apache-2.0 4B Vision-Language Model for Autonomous Driving