Entity dossier

Model model--gemini-robotics-on-device-2

Gemini Robotics On-Device 2

On-device VLA for standing bi-arm manipulation, distributed to select Trusted Testers.

Reviewed relationships 1

  1. Verified

    Google DeepMind Organization publishes Gemini Robotics On-Device 2

    Scope model card

    Boundary Access is limited to select Trusted Testers.

    First observed · reviewed · [M27] · Show in graph

Canonical evidence claims 7

Claims are projected from authored tables. A dash remains an explicit missing value; it is never filled from a neighboring product or name match.

Robot developer ecosystems, SDKs, and simulation Robot Learning Runtime 7 rows

MetricValue and evidence
Artifact kindvision-language-action model based on on-device Gemma models [M27]
Published data pathimages, text, robot sensor data, and action data; dataset details are not published [M27]
Distribution / accessselect Trusted Testers only [M27]
Execution topologygeneral-purpose manipulation runs on local devices [M27]
Input / output contracttext, images, and numerical proprioception → numerical robot actions [M27]
Published boundaryhigh-DoF generalization is limited; mobile-platform and whole-body risks are outside the evaluated scope [M27]
Stated scopestanding bi-arm manipulation across several robot types [M27]

Source ledger 1

  1. [M27]ManualGoogle DeepMind — Gemini Robotics On-Device 2 model card. Type: official model card. Publication date: 2026-07-30. Retrieved: 2026-09-04.Retrieved

What is not inferred

No fuzzy entity matching, adjacent-column borrowing, universal compatibility, commercial maturity, or checkpoint-to-product identity. Absence of a relation means unreviewed or unsupported here—not proven incompatibility.