Skip to content
Robotics

SWAM

scalable world action models for robotics

1 min readupdated

SWAM (Scaling World Action Models with Cross-Modal Alignment) is a family of checkpoints for robotics and action control tasks, utilizing cross-modal alignment. The repository contains pre-trained weights as well as versions fine-tuned on RoboTwin data. The model helps connect visual perception of the world with robot control actions and relies on Wan and UMT5 components. The checkpoints are suitable for researchers and engineers working in robotics and environmental simulation.

Key features

  • Cross-modal alignment: efficient linking of visual data and robot actions.
  • Two checkpoint variants: pre-trained weights and a version fine-tuned on RoboTwin Clean simulation data are available.
  • SafeTensors format: weights are split into shards (sharded Safetensors), with each checkpoint weighing approximately 27.98 GB.
  • Open ecosystem: full integration with the project's datasets and source code.

What you can do

  • Run simulations: use the checkpoints to evaluate and train robotic agents in the RoboTwin environment.
  • Fine-tune models: apply SWAM pre-trained weights for custom robotics tasks and hardware adaptation.

Example

Downloading a pre-trained SWAM checkpoint via the command line:

pip install -U huggingface_hub
hf download FreDERicia/SWAM --include "pretrain/*" --local-dir ./SWAM
Read moreDetails
Checkpoint size: ~27.98 GB each
File format: Sharded Safetensors
Architecture: Requires preliminary preparation of Wan and UMT5 components
Benchmark suite: RoboTwin Clean benchmark
Discuss Telegram 0

Will it run?

Runs in FP8 / INT8

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 26 GB · ready files29 GB
  • FP8 / INT8recommendedweights 13 GB15 GB
  • 4-bitweights 7.3 GB8.6 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 5080 · BF16 — from RTX 5090

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.

    # FreDERicia/SWAM
    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 FreDERicia/SWAM --local-dir ~/models/SWAM
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to SWAM
  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 SWAM (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

FreDERicia/SWAM
Author
FreDERicia
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments