USE CASE

Run ComfyUI on a cloud GPU

Build image and video workflows with ComfyUI on an RTX 4090 or RTX 5090, billed per second.

See RTX 4090 machines Checking live prices…

Billed per second from your USD balance. Stop or terminate the machine when you are done.

Recommended GPU

RTX 4090 · RTX 5090

ComfyUI workflows with SDXL, Flux or video models need 24 GB of VRAM or more to run without offloading.

PyTorch

NVIDIA CUDA + PyTorch development environment, ready with SSH and Jupyter.

1
Rent the machine
Open the catalog with RTX 4090 and the PyTorch template already selected and press Rent.
See RTX 4090 machines
2
Connect with SSH
Copy the SSH command from your machine's page and add the port forward for ComfyUI.
```
$ ssh -p <PORT> root@<HOST> -L 8188:localhost:8188
```
The SSH command is on your machine's page. The -L part forwards the port so the app opens in your own browser.

Run these commands once on the machine.

$ git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
pip install -r requirements.txt
python main.py --listen 0.0.0.0 --port 8188

4
Open it in your browser
Go to http://localhost:8188 on your computer and load your workflow.

Can I install custom nodes?

Yes. Clone them into ComfyUI/custom_nodes and restart, the same as on your own computer.

Which GPU should I pick for video models?

Start with an RTX 5090 (32 GB). For the largest video models pick an A100 with 80 GB.