interface CompletionUsage {
    cacheCreationTokens?: number;
    completionTokens: number;
    promptCachedTokens: number;
    promptTokens: number;
    reasoningTokens?: number;
    serviceTier?: string;
    totalTokens: number;
}

Properties

cacheCreationTokens?: number

Tokens used to write to the prompt cache.

completionTokens: number
promptCachedTokens: number
promptTokens: number
reasoningTokens?: number

Completion tokens spent on hidden reasoning.

Already counted in completionTokens; do not add it to totals. Not all providers break reasoning out separately, and it is 0 when they don't.

serviceTier?: string

The service tier used for processing (e.g. 'default', 'priority', 'flex').

totalTokens: number