DEPLOYDatabase

Brain

RT-2 / RT-X

Google DeepMind's foundational 2023 research VLA. RT-2 is a transformer VLA trained on web text and images that directly outputs robot actions. Instantiations on PaLM-E and PaLI-X. RT-2-X is a 55B-parameter variant. Chain-of-thought reasoning for long-horizon planning. Superseded by Gemini Robotics for commercial paths, but conceptually ancestral to GR00T, Helix, pi0, and the broader dual-system descendants.

Research model · Maturity: Research · Closed

Machine-readable surfaces

Architecture

Transformer VLA trained on web text and images, directly outputs robot actions. Instantiations on PaLM-E and PaLI-X. RT-2-X is 55B parameters. Chain-of-thought reasoning for long-horizon planning. Trained on web data plus Open X-Embodiment.

Key facts

Significance
First-of-its-kind VLA; demonstrated VLMs can become VLAs
Ancestry
Conceptual ancestor of the GR00T / Helix / pi0 dual-system descendants
Powered platforms
Research robot platforms; not a commercial-deployment brain
Successor
Superseded by Gemini Robotics in DeepMind's commercial roadmap
RT-2 VLA Jul 2023 (primary DeepMind)
Google DeepMind (Jul 28 2023) introduced Robotic Transformer 2 (RT-2) as a vision-language-action model co-fine-tuned on web-scale VLM data and robot trajectories so actions are emitted as tokens for end-effector control, building on RT-1 kitchen demos (DeepMind)
Generalisation and CoT results (primary DeepMind)
The same DeepMind post reports more than 6,000 robotic evaluation trials, improved unseen-scenario success versus RT-1 (about 32% to 62% in their reported comparison), and chain-of-thought variants that plan then emit action tokens. Treat metrics as company-stated research results (DeepMind)
Office-kitchen demonstration scale
DeepMind says RT-2 was co-fine-tuned on RT-1 demonstration data collected with 13 robots over 17 months in an office kitchen (Google DeepMind).
Unseen-scenario success lift
DeepMind says that across more than 6,000 robotic trials, RT-2 raised success on scenarios the robot had not seen from RT-1's 32% to 62% (Google DeepMind).

Developed by (1)

Mentioned on (4)

Registry pages whose text names RT-2 / RT-X.

Common questions

What is RT-2 / RT-X?
Google DeepMind's foundational 2023 research VLA. RT-2 is a transformer VLA trained on web text and images that directly outputs robot actions. Instantiations on PaLM-E and PaLI-X. RT-2-X is a 55B-parameter variant. Chain-of-thought reasoning for long-horizon planning. Superseded by Gemini Robotics for commercial paths, but conceptually ancestral to GR00T, Helix, pi0, and the broader dual-system descendants.
Who developed RT-2 / RT-X?
RT-2 / RT-X is credited to Google DeepMind on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is RT-2 / RT-X open source?
No. RT-2 / RT-X is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.
What type of AI is RT-2 / RT-X?
RT-2 / RT-X is a research model, built on a Transformer VLA trained on web text and images, directly outputs robot actions. Instantiations on PaLM-E and PaLI-X. RT-2-X is 55B parameters. Chain-of-thought reasoning for long-horizon planning. Trained on web data plus Open X-Embodiment. architecture on the DEPLOY registry.
What is RT-2 / RT-X's maturity stage?
RT-2 / RT-X is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.
Which robots run on RT-2 / RT-X?
No robot models on the DEPLOY registry are recorded as running RT-2 / RT-X. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Sources (7)

  1. Google DeepMind Blog: RT-2 (primary) · https://deepmind.google/blog/rt-2-new-model-translates-vision-and-language-into-action/
  2. Google Blog: RT-2 coverage · https://blog.google/
  3. https://arxiv.org/abs/2307.15818
  4. https://deepmind.google/blog/scaling-up-learning-across-many-different-robot-types/
  5. https://arxiv.org/abs/2310.08864
  6. https://research.google/blog/rt-1-robotics-transformer-for-real-world-control-at-scale/
  7. https://arxiv.org/abs/2212.06817
Methodology: Verified · 7 sources (no primary) · last reviewed 2026-10-09

Verification posture

Verified

Low confidence

Review state

Stable

Last reviewed 2026-10-09

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

4
unclassified
Unclassified source
3
preprint
Preprint

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for RT-2 / RT-X.

Canonical ID 738c50ad-11cd-41d6-b5a4-1ac63e2acac2