← Leaderboard

RT-2

VLA
Google DeepMind · Jul 28, 2023

The model that introduced the term vision-language-action model: a web-scale VLM co-fine-tuned on robot trajectories, with actions emitted as text tokens. Results are self-reported from about 6,000 real-robot trials (roughly 2x RT-1’s generalization on unseen scenarios and about 3x on emergent-skill tests); weights were never released.

Type
VLA
Parameters
55B
Backbone
PaLI-X 55B (also a PaLM-E 12B variant)
Weights
Closed
License
–
Commercial use
Unclear
Released
Jul 28, 2023
Robots
Everyday Robots mobile manipulator (Google Robot)
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

Not reported on 13 benchmarks
Wrong or missing number? Report an error

Built on RT-2