Hy-Embodied-0.5-VLA
$ 0
Hy-Embodied-0.5-VLA is Tencent’s open robot action model. It turns images and instructions into precise two-handed movements and transfers to different robot arms.
Available on backorder
Brain Score
3
Specifications and details:
| Nationality | China |
|---|---|
| Website | https://www.tencent.com/ |
| Model type | Foundation Model |
| Manufacturer | Tencent Robotics X |
| Release date | June 15, 2026 |
Description
Hy-Embodied-0.5-VLA is a vision-language-action model from Tencent Robotics X and the Tencent Hy Team. It turns camera images and plain instructions directly into robot arm motions. Tencent built it on the HY-Embodied-0.5 backbone and added a small action module for smooth, continuous control. The team also trained it on more than 10,000 hours of two-arm demonstrations from a custom handheld capture rig. Because of this, Tencent presents it as a full robot learning stack rather than just a model.
The model handles careful two-handed chores, such as folding glasses, zipping a pen case and setting a table. It also moved to new robots, including the JAKA K1 and Astribot S1, after training only on the handheld data. Tencent says it already runs on a cosmetics production line with over 95% accuracy. Developers can download the code, two checkpoints and about 2,000 hours of training data under Apache 2.0. Meanwhile, Tencent lists support for robots such as Unitree G1 and AgiBot G1 on its Tairos platform. So the model acts as the hands of the Hy-Embodied family.
New! 2026 Humanoid
Robot Market Report
198 pages of exclusive insight from global robotics experts — uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
now Google DeepMind
Contact Humanoid.guide
Website: https://tairos.tencent.com/openSourceModels/hy-embodied-0.5-vla
