How to quote AI API capacity

Separate throughput from billable usage, spell out available models and limits, and quote the unit price for each kind of token.
Name the platform, model and access channel
List the exact model name or version and identify its access channel. A model accessed directly from a provider can have different limits from the same model accessed through a cloud platform. If region or availability differs by model, give each model its own line.
Separate throughput from total usage
TPM means tokens per minute and RPM means requests per minute. Both describe throughput limits. State how much is available to the buyer, the per-period usage you can supply, and any concurrency or daily limits.
- Available TPM and RPM, where known.
- Any shared limits, per-request limits or caps on daily use.
- Volume per delivery and delivery frequency for ongoing supply.
Give the price with the billing unit
Quote the currency and price per million input and output tokens separately. Specify rates for cached tokens or other usage if those charges apply. If you quote a flat monthly fee, state the included usage and the rate for any excess.
Put the limits and price together in the offer
An example quote might say: Claude Sonnet, 100,000 available TPM, monthly delivery, USD 2 per million input tokens and USD 8 per million output tokens. Specify the effective time, any daily cap and whether those limits are shared. The figures are illustrative.