Leaderboard
Physical AI, measured on real robots.
8 models on 2 benchmarks. Each benchmark has its own tasks and its own score.
| Model |
DROID: Everyday tasks
DROID
144 tasks · 1061 runs
Reliability% of task done
|
Pick and Place
Pick and Place
4 objects · 594 runs
Reliability% items placed
|
|---|---|---|
Robot teleoperated by HumanTeleop
Reference
|
- | 100.0 |
NVIDIA Cosmos3 NanoCosmos3 Nano
|
70.1±6.9 | - |
Black Forest Labs FLUX 3 ActionFLUX 3 Action
|
62.6±7.3 | - |
Ai2 MolmoAct2MolmoAct2
|
49.5±6.8 | - |
Galaxea G0.5G0.5
|
38.5±7.3 | - |
|
NVIDIA GR00TGR00T
|
37.0±7.1 | 72.3±4.6 |
Physical Intelligence π₀.₅π₀.₅
|
14.4±4.7 | 75.1±4.8 |
Action Chunking TransformerACT
|
- | 63.7±7.0 |
Hugging Face SmolVLASmolVLA
|
- | 32.0±7.9 |
± is half the 95% range. Each benchmark's page explains its score.
Updated