Is the Replicate API free? Free runs, credit and tokens

Only briefly. Replicate says "You can run select models on Replicate for free, but after a bit you'll be asked to set up billing", and after that you buy prepaid credit up front. The API token costs nothing: Replicate creates a default token when you create your account.

The Replicate facts are quoted from its billing, prepaid credit, rate limits and API tokens pages, read on 2026-09-29. Replicate doesn't name which models run free or for how long, so this post doesn't either. Sume facts come from the generation admission docs.

How does Replicate's prepaid credit work?

"To use Replicate, you must first purchase credit upfront." Usage is then deducted from the balance as you go.

• To buy credit, add a payment method in billing settings, then choose an amount. The card is charged immediately.
• Purchased credit is valid for 1 year from the purchase date and is not refundable.
• Auto reload is optional: the minimum threshold is $5 and the minimum reload balance is $15.
• When the balance hits zero, Replicate stops new work from starting. In rare cases a prediction runs over the balance, and the default payment method is charged at the end of the month.
• Some features are only available to customers with billing set up.

What limits apply before I add a payment method?

The rate limits page says: "If you have been granted credit and don't have a payment method on file", you're limited to 1 request per second with a maximum of 6 requests per minute, and going over returns status 429. Replicate API rate limits covers the normal limits and how to raise them.

What is free once billing is set up?

A few things still cost nothing. On a public model you pay only for the time it's active processing your requests; "Setup and idle time for the model is free." And "For all models, if a run fails, we don't charge you", with exceptions for private models and deployments. Each model page shows its run time and cost; fal vs Replicate covers how the rest is billed.

Is a Replicate API token free?

Yes. Tokens authenticate requests to the HTTP API; they are 40-character strings that start with r8_. Replicate creates a default token with your account, and you can create more, for example one each for development, staging and production. The token is free; the model runs it authorizes are what you pay for. Treat tokens like passwords.

How do other media APIs handle failed jobs?

If you are comparing hosts, check what each one does with a failed job. On Sume, for example, the generation admission docs say the estimated cost is reserved when a request is accepted, success captures it, and "Failed jobs and failed queue admission release or refund the reservation where applicable." Some Sume checks are unbilled; Free AI video API calls on Sume lists them.