Alibaba's Qwen open-sources Qwen-Drive-1.0-4B for autonomous driving under Apache 2.0
Alibaba's Qwen team, with Huazhong University of Science and Technology, released Qwen-Drive-1.0-4B — a 4B open-weights model that pairs 3D perception with trajectory planning on the Qwen3.5-4B vision-language base, shipping two planners, one trained by imitation and one tuned with reinforcement learning. Code, weights, and demo data sit on Hugging Face and ModelScope under Apache 2.0, putting a full perception-to-planning driving stack in the hands of any robotics team without a licensing deal — and extending Alibaba's open-weights strategy from chat models into autonomous driving.
Source: technode.com ↗
The project provides code, model weights and demo data under the Apache 2.0 license.
Alibaba Qwen team
Why this matters
- → Full autonomous-driving stack (perception + planning) now freely available to any robotics team
- → Extends Alibaba's open-weights strategy beyond chat into embodied AI
- → 4B model size makes it deployable on edge hardware without licensing friction
Open-source driving stack