I use ChatGPT a lot for doing engineering research and technical analysis, and I’ve been fairly happy with the information and analysis o3 (and o3-pro) outputs, but I’m aware that many of my prompts could be answered/solved effectively and more efficiently by o4-mini or o4-mini-high without burning thru my usage limits for the higher-tier models, but I often forget that I have o3 selected, and accidentally burn a usage on a simple question. It would be really awesome if OpenAI came up with a "Dynamic Model Mode’ that somehow assessed the complexity of the request and ran it with the appropriate model, or prompted a recommended model, or ran it with the more efficient model, then gave you a prompt if the user would like to re-run with a more detailed model. This idea actually came from my colleague after I complained about hitting my usage limits surprisingly quickly this week haha. Thanks OpenAI for making some awesome tools for research and engineering!