Cached input audio tokens.
Cached input image tokens.
Cached input text tokens.
Input tokens served from cache.
Input image tokens (for multimodal models).
Input text tokens.
Total input tokens.
The model name (e.g., 'gpt-4o', 'claude-3-5-sonnet').
Output audio tokens (for multimodal models).
Output text tokens.
Total output tokens.
The provider name (e.g., 'openai', 'anthropic').
Total session connection duration in milliseconds (for session-based billing like xAI).
Input audio tokens (for multimodal models).