Model Releases

Qwen-VLA: from understanding the world to taking action

Original title:Qwen-VLA: From Understanding the World to Acting in It

AI summary

Qwen has released Qwen-VLA, a general vision-language-action model built on the Qwen multimodal backbone that extends visual perception, language understanding and spatial reasoning to continuous action generation and trajectory prediction. Training has four stages: text-to-action pretraining (T2A), continued pretraining (CPT), supervised fine-tuning (SFT) and reinforcement learning (RL). It reaches 97.9% on LIBERO, 73.7% on Simpler-WidowX and 86.1%/87.2% on RoboTwin-Easy/Hard, matching or surpassing specialized models. The data covers over 10,000 hours of public robot trajectories, 1,000+ hours of internal real-world trajectories and 8 million+ synthetic simulation trajectories.

Key Takeaways

  • Qwen has released Qwen-VLA,a general vision-language-action model built on the Qwen multimodal…
  • Training has four stages: text-to-action pretraining (T2A),continued pretraining (CPT),supervised…
  • It reaches 97.
  • 9% on LIBERO,73.
  • 7% on Simpler-WidowX and 86.

Quick Facts

Published
Updated
Source
Qwen:Blog Retrieval(API)
Category
Model Releases
AI score
66
Read the original Read the full article at the original source
Why it matters

Qwen-VLA 把机器人操作、导航和跨实体控制统一进一个模型,在多个基准上打平甚至超越专用模型,这是通用具身智能的一个重要信号,但离实际可用还有距离。

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Qwen-VLA: from understanding the world to taking action. May 29, 2026. https://new.nat6.net/en/news/cmpr2b1l40aa6slnoit9b7p92/

View all →