QuickStart - Deploying ComfyUI 1

Note

Recommended: 1 × NVIDIA RTX 5090. Storage is optional — the ComfyUI template provisions its own model folders.

Tip

This page deploys ComfyUI empty and you supply models afterwards. If you would rather have the weights and a ready-to-run workflow arrive with the Pod, attach an official model set instead — see Bundled ComfyUI workflows — or use Quick Deploy, which attaches one for you.

1. Select Workspace

   Create a Workspace and select it.

   

2. Deploy Pod — Select Template

   Click Pod in the sidebar, then click the Deploy button.

   

   Select the Generative box and click ComfyUI in the Image section.

   

   Select the image tag — base-torch2.8.0-cu128.

   

3. Deploy Pod — Select Resources

   Enter your Pod name and select VRAM 32 GB, then select RTX 5090.

   

   (Optional) Select CPU and RAM.

   

   Storage (optional)

   The ComfyUI template is fully managed — the folders ComfyUI reads models from are provisioned with the Pod and grow as you add models, so there is no volume you have to mount before deploying.

   Add a volume only if you want files to survive the Pod itself. Everything the managed folders hold is released when the Pod is deleted (stopping keeps it).

4. Connect ComfyUI and Use

   Click Accept & Deploy.

   Caution

   After Pod creation, wait until ComfyUI is fully running.

   

   Once fully running, click the Connect button to access the GUI.

   

Next

• Run ComfyUI and generate an image or video