Reference

Pricing & billing

Pay strictly per token: every model has a fixed input and output rate, and you pay for exactly what you use. No subscriptions, no minimums, no expiring credits.

How billing works

WhatHow it works
Unittokens (input + output)
Ratefixed per model, shown upfront in /v1/models
Minimum1¢ per request (micro-calls)
Creditsnever expire

Where to see prices

The price of every enabled model is shown on the pricing page, in GET /v1/models and in the dashboard (GET /v1/me, /v1/usage).

Payments are non-refundable. Credits are charged for a service actually rendered (a processed model request), deemed rendered at the moment of charge. See Terms §6 for the full grounds.

Telegram