Breakdown of input tokens for a Responses API call.
Number of tokens read from the prompt cache.
Number of tokens written to the prompt cache.