Concept · Building with AI

Cache tokens

Also called: cached tokens, cache read tokens, cache hits, cached input, cache hit tokens

Input tokens the provider did not need to process again because they matched the start of a recent request in its cache. They appear as their own line on usage reports and cost much less than ordinary input tokens.

For example: A founder opens the usage dashboard and sees that most of the agent's input was cache tokens, which explains why the bill was lower than expected.

Connects to 6

Sources

Page last changed · Suggest a correction