SANA-Robot-AR
SANA-Robot-AR is a 4.5B-parameter world action model that predicts 32-step action chunks and future video. It combines windowed attention with Gated DeltaNet memory. The public pretrained weights are intended for fine-tuning on a target robot or benchmark.
View full model descriptionInitialization for further fine-tuning on target robot demonstrations or simulation benchmarks.
Papers and technical reports
Paper / report unconfirmed
Official introductory materials are recorded; no version-specific paper or technical report has been verified.
Last checked: 2026-11-22
How to access and use
- Model files
- Public access
Training code for this version is unconfirmed. Public code resources do not establish access to the exact training implementation. View verification notes
Training / fine-tuning code
SourceVerified on · 2026-11-22
Model files
| File | Access | Weights license | Commercial use |
|---|---|---|---|
| SANA-Robot-AR Pretrain | Access | Weights license | Commercial use Allowed under cited termsTerms Verified on · 2026-11-22 |
SourceVerified on · 2026-11-22
Checks before deployment
Direct robot control with the pretrained weights has not been validated. Load the text encoder and video VAE separately. Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.
Deployment and usage resources
- Runtime and frameworks
- PyTorch
- Fine-tuning requirements
- Target robot demonstrations, Gemma 2 2B IT and LTX-2.3 VAE
Architecture and inputs / outputs
- Model type
- Action policy
- Architecture
- WAM · Autoregressive
- Backbone
- Hybrid Diffusion Transformer (24 Gated DeltaNet layers and 8 softmax layers)
- Weight parameters
- 4500000000
- Action generation
- Flow matching
- Input
- Language / RGB images / Robot state
- Output
- 32 control steps per action chunk
- Released
- November 22, 2026
- Action generation · Backbone · Main category · Architecture: SANA-Robot project page
- Input modalities · Output action space · Main category · Architecture: SANA-Robot-AR Pretrain model card and files
Scope and training
Output
Actions / Video
Prediction output
Action chunk and future video at a frame stride of 4
Training
Pretraining
Derived models
SANA-Robot-AR RoboTwin 2.0 ↗Fine-tuned model
SANA-Robot-AR LIBERO ↗Fine-tuned model
- SANA-Robot-AR RoboTwin 2.0: SANA-Robot-AR RoboTwin 2.0 model card
- SANA-Robot-AR LIBERO: SANA-Robot-AR LIBERO
- Output · Prediction output · Training: SANA-Robot-AR Pretrain model card and files
Full parameters and verification notes (model card)
Each chunk predicts 32 control steps and encodes video at a frame stride of 4. Softmax attention reads the first frame and the previous chunk, while Gated DeltaNet summarizes preceding chunks in a fixed-size state.
External dependencies: Gemma 2 2B IT (Google terms acceptance) and LTX-2.3 video VAE (Diffusers layout).
Fine-tuning recipes: RoboDojo, RoboTwin 2.0 and LIBERO. Each has a separately released fine-tuned checkpoint, and the benchmark scores belong to those variants.