I wonder sometimes whether my query should be answered by ChatGPT 4o, o1-mini, or o1. OpenAI should have a mode where it decides which LLM should answer your query based on its complexity. Since 4o is so much faster than o1, you could have an agent successively try 4o, o1-mini, or o1 until a question is answered correctly.