“Google's Gemini models process 22 billion API tokens per minute.”