Number of tokens in the input (prompt).
Number of tokens in the output (completion).
Total number of tokens (input + output).
OptionalcacheNumber of input tokens read from cache. This can reduce latency and cost.
OptionalcacheNumber of input tokens written to cache. These tokens can be reused in future requests.
Token usage statistics for a model invocation. Tracks input, output, and total tokens, plus cache-related metrics.