Deploy guide
Deploy ComfyUI on Vast.ai
Serve a ComfyUI instance on a Vast.ai GPU behind one URL, so image and video workflows can be called from your app instead of from a desktop. Cheapest card that fits SDXL and Flux workflows: RTX A4000 (16 GB) at $0.09/hr — about $66/month if it ran 24/7.
Quickstart
| Image | your-comfyui-image:latest |
|---|---|
| Container port | 8188 |
| Health endpoint | / |
| Minimum VRAM | 16 GB |
- Create a deployment with the your-comfyui-image:latest image. In the dashboard choose New deployment, paste your-comfyui-image:latest, and set the container port to 8188.
- Pick the GPU pool. Select SKUs with at least 16 GB of VRAM — the cheapest right now is RTX A4000 (16 GB) at $0.09/hr.
- Set the health endpoint and start. Use / as the health endpoint so a worker only receives traffic once the container answers. Start the deployment and open the public URL it gets.
Things that bite
- ComfyUI serves its UI and API on 8188. Bring your own image, or build one from the ComfyUI repository's Docker instructions — the deployment only needs the port and a health path.
- Workflows are submitted over the API, so the container must expose the API, not just the web UI; make sure the image starts with --listen on all interfaces.
- SDXL fits a 16 GB card; FLUX and video workflows want 24 GB or more. Model checkpoints are large — raise the disk size above the default.
What it costs
GPU time is Vast.ai's own price, billed per second to your Vast.ai account; CUDAnetes adds a flat subscription (free plan or $19/mo on Pro) and takes no markup. A fleet running 24/7 for a month costs $66 per card at the cheapest live offer, before the subscription.
See the live GPU price index for current rates, or the Vast.ai vs RunPod comparison for the fixed-price alternative.
Questions
Do I need to rebuild ComfyUI for CUDAnetes?
No. Any image that serves on 8188, listens on all interfaces and answers a health path works as-is.
How is a ComfyUI deployment billed?
You pay Vast.ai per second for the rented GPU plus a flat CUDAnetes subscription (free or $19/mo) — there is no per-image markup.
Can the deployment scale to zero?
Yes, set Min GPUs to 0: an idle fleet stops renting, and the first request after that counts as demand to start a worker (the request itself gets a 503 while the GPU boots).
Deploy it now
Free plan: one deployment with up to five GPUs, no credit card. Any Docker image, one public URL.