88Z AI, 1 June — Tongyi Lab releases Qwen-VLA, built on Qwen3.5-4B and a DiT action decoder, integrating manipulation, navigation, and trajectory prediction to support 11 robot platforms.
At a glance
- Qwen-VLA integrates manipulation, navigation, and trajectory prediction into a single general model
- Supports 11 robot platforms and outperforms specialized models in benchmarks
The update
Tongyi Lab has officially launched Qwen-VLA, a unified action framework for general embodied intelligence. Built on the Qwen3.5-4B foundation and paired with a DiT action decoder, the model innovatively employs a unified action trajectory prediction framework, proprioceptive prompt conditioning, and text-to-action pre-training techniques. These core technologies enable Qwen-VLA to integrate robotic manipulation, navigation, and trajectory prediction into a single general embodied model. The model outperforms many specialized models in various benchmarks. Furthermore, Qwen-VLA demonstrates strong cross-embodiment control capabilities, supporting up to 11 different robot platforms, highlighting its broad applicability and versatility in the field of embodied AI.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 通义实验室Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.