You pay for what you run
Add credit, then pay exactly what a run metered, down to millionths of a dollar.
Example runmetered
inputprompt
gpt-image-2.5-flare$0.0403
1K tokens in and out
flux-dev$0.0288
one image
outputimage
Charged to balance$0.0000
How billing
works
You add credit, runs draw it down. Nothing is reserved, nothing is held.
Start with free credit
Every new organization gets a balance. No card, no trial.
Run anything in the catalog
Runs start while your balance is positive.
Pay when the run ends
One ledger entry, debited as the run finishes.
Model compute is the only metered line item
Priced per image, second, clip, token, or thousand characters, and charged to your balance as each run finishes. Top up $5–$500 at a time, one time payments, no card kept on file.
Everything else is included
- The node editor, and the agent that builds graphs for you
- The full node library
- API, SDK, MCP server, and CLI
- Published workflows, pinned and reproducible
- Team organizations and shared keys
- Run history and per run cost breakdowns
- How is a run priced?
- Each step reports what it consumed, and the run is charged the sum of its steps, at the same rates the engine bills with.
- Is there a subscription?
- Not today. You hold a prepaid balance and spend it on the runs you make.
- What happens when a run fails?
- You pay for the steps that ran. A step we were never billed for is priced at zero.
- Can these rates change?
- They can. A run pins its rate version when it starts, and is never repriced after it is charged.
- We want to run this at volume.
- Tell us what you are building and roughly how much you expect to run. Get in touch