Pending benchmarksTracking since Nov 4, 2026
Is AIRoA-0 (A-0) state of the art?Pending benchmarks.Announced Nov 2, 2026; no independent results yet.
AIRoA's robot foundation model pre-trained only on 16,151 hours of handheld human demonstrations recorded with the YUBI device, with no robot data.
Announced facts
The details so far
- Lab
- AIRoA
- Announced
- Nov 2, 2026
- API
- Not yet
- Open weights
- Announced, not yet published · Lab says dataset, pre-trained weights and recipe are released; no weights repo located on Hugging Face as of Nov 4, 2026
- Pre-training data
- 16,151 hours of YUBI handheld demonstrations; no robot data
- Embodiments
- Bimanual Franka, TX-G2, OpenArm and UR, with two wrist cameras
- Post-training
- Under 4 hours on one 8×H200 node (lab-stated)
Tags marked unsure are not stated on the lab's own pages.
Next steps
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
Self-reported
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: 78% average success across ten real-world tasks, versus 27% for the strongest of four other VLAs tested.
Receipts
Sources
PAPERAIRoA-0 (arXiv 2611.03815)arxiv.org/abs/2611.03815 · Nov 2, 2026LABA-0 project pagea-0.airoa.ioTracking ends Feb 2, 2027 unless the API or weights become available.