Cost of each request

Every request is priced from the tokens the provider reports and the offer's prices per million tokens:

• input tokens at the input price;
• cached input tokens at the cached input price, when the provider reuses your prompt from its cache;
• cache writes at the cache write price, where the provider charges for them;
• output tokens, including reasoning, at the output price.

Prices are listed per model and provider on the models page. On credits, this is the amount taken from your balance. On your own provider key it's an estimate at the same prices: the provider bills you, and DigitSpace charges nothing.

Where to see it

• Activity: every request, with its cost, tokens, duration, model and provider. Open one for its cost breakdown. See Activity.
• Usage and Analytics: spend over time, grouped by model, by API key or by member. See Usage and analytics.
• Model usage: requests, tokens, cost and share of spend per model.
• Org analytics: spend by project, model, key or member across the organisation.
• GET /v1/key: the calling key's total spend, from code.

The API response itself doesn't carry the cost. Its x-request-id header matches the request ID shown in Activity.

Tag requests to break cost down

Use a separate API key for each app, feature or customer: Analytics and Model usage group spend by key. Sending X-Source (or X-Title) with the name of the app also records it as the request's Source in Activity. See Request headers.

Keep cost down

• Use auto:economy for simple work; see auto.
• Set max_tokens: it caps the answer, and on credits the balance needs to cover it before the request is sent.
• Choose the Cheapest routing strategy where speed matters less.
• Keep long, repeated instructions at the start of the prompt so providers can cache them.
• Add your own provider keys for heavy use: those requests are free on DigitSpace's side.