Robot AI model
RT-2
RT-2 co-trains pretrained vision-language models on web tasks and robot trajectories, representing robot actions as text tokens. The project evaluates PaLM-E and PaLI-X variants, so no single parameter count is assigned to the family.
Records last verified Oct 13, 2026 · Send a correction
Also known as Robotics Transformer 2
- Organization
- Google DeepMind
- Type
- VLA
- Released / introduced
- Jul 28, 2023
- Public weights
- Not established
- Licence / scope
- Not established
- Parameters
- Not disclosed
- Base model
- PaLM-E or PaLI-X (variant-dependent)
- Humanoid association
- None reported
- Inputs
- vision, language
- Outputs
- robot action tokens, natural-language reasoning
Release dates can mark an announcement or paper introduction. Public checkpoints, code licences and weight terms are distinct; follow the publisher’s links for the stated scope.
Disclosed training data
- RT-1 robot demonstration dataset
- Web vision-language data used by PaLM-E and PaLI-X
More robot AI models
Evidence
Sources
- [s1] RT-2 project: method and model variants ↗
Google DeepMind · official · accessed Oct 13, 2026
- [s2] RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control ↗
Google DeepMind · paper · published Jul 28, 2023 · accessed Oct 13, 2026