Breakdown of the output tokens counted in ResponseUsage.outputTokens.
Number of chain-of-thought tokens the model produced in thinking mode.