ComfyUI on a rented GPU
Run ComfyUI on a cloud GPU
Pick a GPU and it starts with ComfyUI installed, running and behind its own private link. Flux, SDXL, Wan, Z-Image: download the models onto the machine with one button and queue your workflow. Billed per minute, at the machine's price.
- H200 NVL 140 GBfrom $3.50/h
- RTX A6000 48 GBfrom $0.40/h
- RTX 5090 32 GBfrom $0.63/h
- RTX 4090 24 GBfrom $0.45/h
- RTX 4070 Ti SUPER 16 GBfrom $0.25/h
Measured, not quoted
On a rented RTX 4090
Measured on an RTX 4090 rented on ParalonCloud. The install and every model download go at the machine's own network speed: the Network column on the GPU list tells you which machines are quick.
What you get
Installed, running, yours
Installed, pinned, tested
A ComfyUI version we installed and ran before offering it, with ComfyUI-Manager on.
Download to this GPU
One button in ComfyUI's top bar fetches every model a workflow names, onto the rented machine, each into its folder.
Its own private link
ComfyUI opens at its own address with a token in it. No port is open on the provider's network.
A full machine underneath
JupyterLab or SSH as well: wget, huggingface-cli, nvidia-smi, your own custom nodes.
Which GPU
What fits on which card
| What you run | 16 GB | 24 GB (RTX 4090) | 32 GB (RTX 5090) |
|---|---|---|---|
| SDXL, SD 1.5 | Comfortable | Comfortable | Comfortable |
| Flux schnell / dev (FP8) | Works | Comfortable | Comfortable |
| Z-Image Turbo | Works | 21 GB measured | Comfortable |
| Wan 2.2 5B video | Tight | Works | Comfortable |
| Wan 2.2 14B video | No | With offloading | With offloading |
From the model sizes and community reports, except where a cell says measured. More in Rent a GPU for ComfyUI.
How it works
Three clicks to ComfyUI
- 01
Pick a GPU that runs it
The list filtered for ComfyUI shows only machines with 16 GB of VRAM or more and room on disk.
- 02
Rent with ComfyUI
The machine answers in about a minute, billing starts; ComfyUI sets up on top and the start screen shows each step.
- 03
Open ComfyUI
Load a workflow, press Download to this GPU for its models, and queue it.
FAQ
ComfyUI in the cloud
How much does it cost to run ComfyUI in the cloud?
The price of the machine, per minute, with nothing added for ComfyUI. The live prices are above; an RTX 4090 is usually around $0.45 an hour.
Which GPU do I need for ComfyUI?
16 GB of VRAM or more: SDXL and Flux in FP8 run on it. A 24 GB RTX 4090 is the safe middle for Flux and Z-Image; a 32 GB RTX 5090 has room for video models.
How do I get models onto the GPU?
Open a workflow and press Download to this GPU in ComfyUI's top bar: every model it names comes down onto the machine, into the right folder. ComfyUI-Manager and the terminal work too.
Are my models and outputs kept after the rental?
No. Everything on the machine is deleted when the rental ends. Download your outputs from /workspace/ComfyUI/output first.
Can I install custom nodes?
Yes, with ComfyUI-Manager or git in the terminal. It is your machine for the length of the rental.
Can I start it from code or an AI agent?
Yes: "template": "comfyui" on POST /api/v1/rentals, or ask an agent with the ParalonCloud MCP server. The rental reports the install and gives ComfyUI's link.
ComfyUI, ready in minutes
From $0.25 an hour, billed per minute. Stop whenever you like.