> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aiohub.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Pricing

> Understand AIOHub billing, ratios, and usage records.

## Billing model

AIOHub uses prepaid usage billing:

1. Add funds to your account.
2. Each API request deducts cost according to the current model, endpoint, and group pricing.
3. If the balance is too low, the request is rejected.

## Price source

Model and group prices vary. Current prices follow the console pricing page and the `/api/pricing` response, which can include model prices, group ratios, available groups, and supported endpoint types.

<Note>Do not treat a static documentation page as a permanent price table. Billing and availability follow the console price data and the actual deduction at request time.</Note>

## Cost components

An API request can include these cost components:

* **Input tokens**: text you send to the model.
* **Output tokens**: text returned by the model.
* **Model price or ratio**: different models can use token-based ratios or task-based prices.
* **Group ratio**: the same model can have different discounts or premiums in different groups.
* **Endpoint-specific ratio**: some models distinguish cached tokens, images, or other endpoint-specific usage.

Image, video, audio, and task-based models may use different billing methods. When reconciling cost, rely on the console usage records and returned usage fields.

## View usage and cost

Use the console to view:

* [Current balance](https://api.aiohub.org/console/topup)
* [Payment history](https://api.aiohub.org/console/topup)
* [Usage records by time, model, and API key](https://api.aiohub.org/console/log)
* [Tokens and cost for individual requests](https://api.aiohub.org/console/log)

## Control cost

* Compare the same model across groups in the pricing page.
* For batch work, test with a small request first, then switch to economy or discounted groups.
* Limit context length for large requests.
* When streaming, remember that interrupted requests can still incur cost for generated output.