To get full access to the Kademi Hub existing customers can login here, or new customers can register here.
Breakdown of the prompt token usage OpenAI can return for a request, distinguishing tokens served from the prompt cache, tokens written to it, and tokens spent on audio input from the total prompt tokens. Matches the shape of the prompt_tokens_details field of a chat completion and the input_tokens_details field of a response, which differ in which of these fields they populate.
Properties
| Property | Returns | Description |
|---|---|---|
| audioTokens | Integer | Number of prompt tokens that came from audio input. |
| cachedTokens | Integer | Number of prompt tokens that were served from OpenAI's prompt cache rather than freshly processed. |
| cacheWriteTokens | Integer | Number of prompt tokens written to the prompt cache for a later request to reuse. From GPT-5.6 on these are charged above the ordinary input rate, so they are counted separately when costing a call. |
Methods
getCachedTokens()
Returns: Integer
Number of prompt tokens that were served from OpenAI's prompt cache rather than freshly processed.
getCacheWriteTokens()
Returns: Integer
Number of prompt tokens written to the prompt cache for a later request to reuse. From GPT-5.6 on these are charged above the ordinary input rate, so they are counted separately when costing a call.
getAudioTokens()
Returns: Integer
Number of prompt tokens that came from audio input.
© 2025 Kademi Ltd