LLMs are intrinsically deaigned for token production, which is typically inversely related to optimization and efficoency.