AIRoA · robot AI model
AIRoA-0
Japan · 2026
Sources retrieved: 9 November 2026 (A, B, C)
- Model type
- Vision-language-action (VLA)Source and note: Model typeB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “AIRoA-0 builds on the pi0.5 vision-language-action model and the openpi code of Physical Intelligence”. The arXiv abstract compares A-0 with “four other vision-language-action models post-trained on the same task demonstrations”. The model card tags the repository vision-language-action.
- What it controls
- Arms and handsSource and note: What it controlsA arxiv.org/html/2611.03815v1 · Retrieved 9 November 2026Note. Technical report: “every action in a chunk is the pose a gripper should reach”. Actions are defined at the grippers of a bimanual setup.
- Inputs
- Camera Language Robot stateSource and note: InputsA arxiv.org/html/2611.03815v1 · Retrieved 9 November 2026Note. Technical report: “The model observes two wrist images at 224 × 224 pixels, an instruction and a proprioceptive state”. The cameras are wrist cameras.
- Access
- Open weights (download)Source and note: AccessB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “This repository is publicly accessible, but you have to accept the conditions to access its files and content.” “Requests are processed immediately.”
- Parameters
- 3.36BSource and note: ParametersB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “Memory: 3.36 B parameters”.
- Base model
- pi0.5Source and note: Base modelB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “AIRoA-0 builds on the pi0.5 vision-language-action model and the openpi code of Physical Intelligence”. “Its vision-language backbone is” Google’s PaliGemma.
- Licence document
- Gemma Terms of UseSource and note: Licence documentB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Weights: “The weights are a Model Derivative of Gemma and are distributed under the Gemma Terms of Use”; “use is subject to the Gemma Prohibited Use Policy”. Code: “The AIRoA-0 code is Apache-2.0”.
Specifications
11 of 14 stated 3 not stated
2 fields hidden 1 not-stated field with a note stays ·
How to read the figuresShow guideClose
- not statedNo source
- yesStated by a source
- 0Reported as zero
- ASource
Select a source marker for source details and notes.
Release1/1 disclosed
Released2 November 2026
Source and note: ReleasedC arxiv.org/abs/2611.03815 · Retrieved 9 November 2026Note. [A] arXiv: “[Submitted on 2 Nov 2026]”.
Input and output3/4 disclosed
Model typeVision-language-action (VLA)
Source and note: Model typeB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “AIRoA-0 builds on the pi0.5 vision-language-action model and the openpi code of Physical Intelligence”. The arXiv abstract compares A-0 with “four other vision-language-action models post-trained on the same task demonstrations”. The model card tags the repository vision-language-action.
InputsCamera Language Robot state
Source and note: InputsA arxiv.org/html/2611.03815v1 · Retrieved 9 November 2026Note. Technical report: “The model observes two wrist images at 224 × 224 pixels, an instruction and a proprioceptive state”. The cameras are wrist cameras.
What it controlsArms and hands
Source and note: What it controlsA arxiv.org/html/2611.03815v1 · Retrieved 9 November 2026Note. Technical report: “every action in a chunk is the pose a gripper should reach”. Actions are defined at the grippers of a bimanual setup.
Degrees of freedom controlledNot stated
Size and speed2/2 disclosed
Parameters3.36B
Source and note: ParametersB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “Memory: 3.36 B parameters”.
Control rate20Hz
Source and note: Control rateA arxiv.org/html/2611.03815v1 · Retrieved 9 November 2026Note. Action output grid of the model: “in pre-training it outputs a chunk of 32 future actions on a 20 Hz grid”; post-training recipe: “Action chunk 20 anchored actions at 20 Hz”.
Where it runs and what it is built on1/2 disclosed
On robot or cloudNot stated
Base modelpi0.5
Source and note: Base modelB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “AIRoA-0 builds on the pi0.5 vision-language-action model and the openpi code of Physical Intelligence”. “Its vision-language backbone is” Google’s PaliGemma.
Openness2/3 disclosed
AccessOpen weights (download)
Source and note: AccessB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “This repository is publicly accessible, but you have to accept the conditions to access its files and content.” “Requests are processed immediately.”
Licence documentGemma Terms of Use
Source and note: Licence documentB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Weights: “The weights are a Model Derivative of Gemma and are distributed under the Gemma Terms of Use”; “use is subject to the Gemma Prohibited Use Policy”. Code: “The AIRoA-0 code is Apache-2.0”.
Licence typeNot stated
Source and note: Licence typeB huggingface.co/airoa-org/AIRoA-0 · Retrieved 9 November 2026Note. Model card: “The weights are a Model Derivative of Gemma and are distributed under the Gemma Terms of Use”; “use is subject to the Gemma Prohibited Use Policy”.
The developer’s own claims2/2 disclosed
Learns new tasksNeeds retraining
Source and note: Learns new tasksC arxiv.org/abs/2611.03815 · Retrieved 9 November 2026Note. Developer claim: post-training per task, “Anyone can build the gripper, record a task of their own and post-train the model on it in under four hours on a single node of eight H200 GPUs.”
Works across robotsYes
Source and note: Works across robotsC arxiv.org/abs/2611.03815 · Retrieved 9 November 2026Note. Developer claim: “A-0 is embodiment-agnostic: it runs on robots that mount the YUBI gripper, such as a bimanual Franka, TX-G2, OpenArm and UR.”Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.
Notes
AIRoA describes A-0 as a bimanual manipulation system learned from handheld demonstrations, with no robot data collected.
The Hugging Face model card says AIRoA-0 builds on the pi0.5 vision-language-action model.
AIRoA’s arXiv abstract says A-0 runs on robots that mount its YUBI gripper, such as a bimanual Franka, TX-G2, OpenArm and UR.
The model card gives AIRoA as the AI Robot Association.
AIRoA’s website lists its address in Tokyo.
Sources
One line per source, not per figure: a source has one retrieval date.
- Aarxiv.org/html/2611.03815v1Retrieved 9 November 2026
- Bhuggingface.co/airoa-org/AIRoA-0Main source. Retrieved 9 November 2026
- Carxiv.org/abs/2611.03815Retrieved 9 November 2026
