Overview
The Estimates API lets you preview token counts and costs before sending a request to a model. This is useful for budgeting, comparing models, and validating prompt sizes.Base URL
Endpoints
Authentication
All requests require a virtual key in theAuthorization header. See Authentication.
Estimate tokens
Request body
Example request
Example response
Estimate cost
Request body
Example request
Example response
Estimation methodology
- Token counts use a heuristic of approximately 4 characters per token plus a per-message overhead.
- Cost is computed from the gateway pricing cache. If a model is not in the cache, the gateway falls back to conservative default rates.
- Pricing is expressed in cents and is normalized per 1 million tokens.
Error codes
Related pages
- Chat Completions — request shape used for estimates
- Models — model registry and pricing
- Token Budgets — budgeting guidance