AIRoA-0
$ 0
AIRoA-0 is an open robot foundation model from Japan’s AI Robot Association. It learned two-arm skills from human recordings instead of robot data, and it runs on many different robots.
Available on backorder
Brain Score
3
Specifications and details:
| Nationality | Japan |
|---|---|
| Website | https://a-0.airoa.io/ |
| Model type | Foundation Model |
| Manufacturer | AIRoA |
| Release date | November 6, 2026 |
Description
AIRoA-0 is a robot foundation model from AIRoA, the AI Robot Association, a Tokyo-based group backed by Japan’s Ministry of Economy, Trade and Industry. The association released it in November 2026 as Japan’s open answer to leading robot models. It’s a vision-language-action model that turns what a robot sees and hears into two-arm movements. However, AIRoA-0 learned in an unusual way. The team collected no robot data at all. Instead, people recorded more than 16,000 hours of everyday tasks with YUBI, a handheld gripper that Toyota helped design.
This approach makes data collection much cheaper and faster than operating real robots. As a result, AIRoA argues that it is the key to scaling robot learning. The model also works across different robots, because it controls the gripper rather than specific joints. For example, the team ran it on two-arm Franka, UR and OpenArm setups, plus the TX-G2 semi-humanoid. In tests, AIRoA-0 folded towels, sorted small parts and adjusted when someone moved a box mid-task. Moreover, it averaged a 78% success rate across ten tasks, far ahead of other models trained on the same data. AIRoA has shared the model and code openly for research and development.
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://a-0.airoa.io
