With OpenAI adding 4.1, I’m having to think a little too much about which models I should be using. I generally use o4-mini for coding, 4o for simple, fast questions. What now? I expect GPT-5 will help consolidate some of these choices.

@manton Have you tried Cursor instead for coding tasks, using Claude 3.7 Sonnet or Gemini 2.5 Pro? 4.1 is also okay at coding but ChatGPT is a poor interface for it

@apike Not yet. I should try Cursor, just hate to change text editors. Mostly use Nova and Xcode with ChatGPT’s “work with” feature so it can peek into the current file or selection. It’s clunky.

@manton There will soon be a model whose job is to only to ask it which model to use for your question. 🙈

@manton the Roo-code extension in VS Code using Claude is a remarkable experience. Especially with the “orchestrator” and “architect modes” and if you use a good rules file prompt.