Where to run it
If it fits your card, run it at home: an RTX 4090 draws about $0.09 of power an hour. If not, an RTX 4090 rents from $0.33 an hour, Comfy Cloud bills $4.54 an hour of generation on a 96 GB card, and RunPod sells a 720p Wan 2.2 clip for $0.30. Prices read 9 to 19 October 2026.
Your own GPU
$0.09 an hour in power
An RTX 4090 at its full 450 W, at $0.20 a kWh. When it fits: no meter, and your files stay at home.
Computed
Comfy Cloud
$4.54 an hour
Of generation only, on a 96 GB RTX PRO 6000. Plans from $20 a month; a curated set of nodes.
Read 16 October 2026
A GPU by the hour
$0.33 an hour
An RTX 4090 on SaladCloud, $0.89 on RunPod, which runs its own data centres. Any node; it bills while it idles.
Read 19 October 2026
A per-clip API
$0.30 a clip
Wan 2.2 at 1280×720, 5 seconds, on RunPod's model API: their model and settings, not your workflow.
Read 9 October 2026
On-demand prices before tax, as each provider lists them. Power is the card's full board power at $0.20 a kWh, before the rest of the PC.
What you pay for
Your own GPU
Power only, a little while idle
Comfy Cloud
Only while it generates
A GPU by the hour
The whole session, idle included
A per-clip API
Each clip it makes
Which one, when
The short answer
Check your own card first; pay for what it cannot do.
- It fits your card
- Run it at home No meter, and your files stay with you. A smaller file often closes a gap of a few GB.
- A ComfyUI workflow whose nodes Comfy Cloud has
- Comfy Cloud It bills only while a workflow runs, on a 96 GB card. You cannot add custom nodes yourself.
- Custom nodes, your own models, long sessions
- A GPU by the hour Any setup you like. You set it up, and you pay while the machine is on.
- One model, called from your code
- An API The simplest bill per image or clip, on their model and settings rather than your workflow.
- An open LLM to chat with
- Your card, or Ollama Cloud Ollama Cloud sells monthly credits for the open models it hosts; the guide says which of them fit a machine at home.
Compared in depth
Every price with the date we read it, and what the hourly rate leaves out.
Rent a GPU: cloud GPU prices compared On-demand prices from 13 providers for every card from the RTX 4090 to the B200, the card your model needs, the price per token, and what the hourly rate leaves out. 15 min read ComfyUI in the cloud, compared Comfy Cloud, RunComfy, ThinkDiffusion, MimicPC, Floyo, RunPod and Vast.ai: price per hour by card, what idle time costs, nodes, files and privacy. 12 min read ComfyUI API: your options compared Comfy API, RunComfy, ViewComfy, fal, Replicate, RunPod Serverless and Modal: what a GPU hour and one job cost, custom nodes, cold starts, agent spend limits and privacy. 12 min read Ollama Cloud: pricing, models and privacy What a month of credits buys on each cloud model, which ones fit a machine at home, and what Ollama says about your prompts. 11 min read