# Pricing (/pricing)



**$0.035 per million input tokens**, for decisions and scores alike.

| What              | Endpoint                | Price                                                  |
| ----------------- | ----------------------- | ------------------------------------------------------ |
| A decision        | `POST /v1/decide`       | Its input tokens: about $0.0002 for a 5,000-token case |
| A score           | `POST /v1/score`        | Its input tokens: usually under $0.0001                |
| Create a score    | `POST /v1/scores`       | Free                                                   |
| Calibrate a score | `POST /v1/calibrate`    | Free                                                   |
| Check a job       | `GET /v1/jobs/{job_id}` | Free                                                   |

* **A decision's input tokens** are its data, context and questions, as the model reads them. Each response reports them in `usage.input_tokens`; up to 32,768 per request ([limits](/reference#limits)).
* **A score's input tokens** are the record and the score's own input, usually a few hundred to a couple of thousand tokens in all.
* **Each request is rounded up** to a whole micro-dollar ($0.000001). Usage in the Finic Portal shows every request's tokens and cost.
* **Only successful calls are charged.** Errors, including `429 rate_limited`, are free.

## Credit [#credit]

Usage is paid from your organization's prepaid credit. New organizations come with $25 of free credit; one created while connecting an AI agent starts with $5, and the rest is added once Finic approves the account. An admin adds credit in the Finic Portal under [Settings → Billing](https://portal.finic.ai/settings/billing), and Usage shows requests by key.

When credit runs out, decide and score calls fail with `402 insufficient_credit` until an admin adds more. Creating and calibrating scores still work.