Prompt caching

Paying once to have the model remember the beginning of your prompt, then paying a fraction of the rate each time you reuse it.

what you are billed for · the vocabulary

Agents and chat send the same long preamble on every call. Caching writes it into the model's working state once — usually at a premium — and reads it back at a tenth of the input rate or less for a few minutes. It changes the economics of anything with a fixed system prompt, and it is the reason cache read and cache write appear as their own rates on a card here.

Nearby