There are many models available now, and they often give different answers. Some tasks are better suited to older or shorter models. It’s getting harder to predict which model will work best. I suggest adding a feature that lets users choose multiple models to answer the same question, which could help solve the problem of models giving incorrect or inconsistent answers.
For instance, I do have a task to change HTML structure, but maintains the TEXT (no alter texts). Claude 3.7 nails this, but gpt-4.1 or even o3, always edits my TEXTS, they try to make it grammatically right or whatever (i do not want it).
But, o3 and gpt-4.1 are way better in refactoring code (in many cases), just like Gemini. So… I think you see my point haha
you mention those in rules. go to cursor settings > rules > write it down.
Nah. I mention to ASK a question, but have Multiple models answer at the same time, instead of like: Ask to Claude > Rewind checkpoint > Ask to GPT4 > Rewind Checkpoint > Ask to Gemini… and so on