Billing and usage

GenAI API bills you for what you use, measured from what you send and receive. You pay nothing upfront: there is no setup charge, reserved capacity, or long-term commitment.

How a call is billed

Each model is billed by a unit that suits its task, rather than a flat amount per request:

• A text or embedding model is billed on its input and output tokens rather than the number of requests. A single request can use many tokens.
• An image-generation model is billed on its inference steps.
• A speech model is billed on its output tokens or its audio duration.

Each model sets its own rate for its billable unit. A model with no billable unit set shows Pricing not available in place of a price. No model is labelled Free.

See your usage and pricing in the portal

Open a model’s page and select Usage to see its total tokens used, its input and output tokens separately, and the estimated cost for each. You can filter it by the current month or a custom date range. Check Pricing opens the model’s Pricing tab on the same page.

For current rates, see E2E Networks pricing.

To choose a model and how to call it, see Models and calling methods. For short answers on specific charges, see the TIR GenAI API FAQ.