Skip to content
Robotics

SANA-Robot-AR-RoboDojo: Efficient hybrid-memory robot control model

1 min read

SANA-Robot-AR is a world action model created by Efficient-Large-Model for robotics. It takes current multi-view camera observations, proprioception (robot body position data), and a language instruction, then predicts a chunk of 32 control steps and future video. The model uses a hybrid memory (combining a softmax attention mechanism and Gated DeltaNet layers), which keeps query costs from growing with episode length. This checkpoint is fine-tuned on the RoboDojo benchmark for the ARX-X5 dual-arm robot.

Key features

  • Hybrid memory: Processes episode history without increasing computational complexity over time.
  • Robot control: Outputs a 32-step control chunk for both arms of the ARX-X5 robot.
  • Multi-view support: Uses footage from head and dual wrist cameras.
  • Strong results: Achieves an average success score of 27.09% on the official RoboDojo leaderboard.

What you can do

  • Simulation testing: Run the model to evaluate manipulator control policies in the RoboDojo environment.
  • Robotics research: Explore the application of world dynamics models to manipulation tasks.
  • Fine-tuning: Use the checkpoint for adaptation to new tasks and robotic setups.

Example

Downloading model weights for further use in code:

hf download Efficient-Large-Model/SANA-Robot-AR-RoboDojo --local-dir checkpoints/sana_robot_ar_robodojo
Read moreDetails
Policy: SANA-Robot-AR, 4.5B parameters.
Text encoder: Gemma 2 2B (not included).
Video VAE: LTX-2.3 VAE (not included).
License: Apache 2.0.
Discuss Telegram 0 20

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 17 GB · ready files19 GB
  • FP8 / INT8recommendedweights 8.3 GB9.7 GB
  • 4-bitweights 4.7 GB5.7 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 3080 · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Efficient-Large-Model/SANA-Robot-AR-RoboDojo
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Efficient-Large-Model/SANA-Robot-AR-RoboDojo --local-dir ~/models/SANA-Robot-AR-RoboDojo
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to SANA-Robot-AR-RoboDojo
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of SANA-Robot-AR-RoboDojo (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Efficient-Large-Model/SANA-Robot-AR-RoboDojo
Author
Efficient-Large-Model
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

200

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Other

SANA-RWM

SANA-RWM