I've paid for a month of OpenAI's top-tier plan to give their most powerful model a try, and I'm underwhelemed. My gripes are mainly as follows: - It wasn't noticeably better than o1-preview, but it was much, much slower. - The ChatGPT UI is a terrible way to have it write code, but that's one of it's most hyped use cases I used it exclusively for coding, so if you have other uses in mind YMMV. However, it just wasn't that good in my limited testing. I asked it to write a small program for me and after successfully scaffolding the project I used Cursor + Sonnet to clean it up. This is a mix of both points in one: I needed to make some modifications to the program after initial creation, so it was not perfect. Once i went from generating all new code to modifying existing code the ChatGPT UI really gets in the way. I don't want to copy and paste code back and forth. Manually shuffling code between two digital systems is just ridiculous. As a result I didn't use o1 much for coding despite its touted benefits. I've usually been able to get the results I want using Claude or o1-preview integrated with Cursor. I would certainly like the option to fallback on the more powerful model when lesser models fail, but for now it's not worth $200/mo in my opinion.