With current difference between max output for Sonnet and GPT it would be very cool to have ability to set specific settings for different models. Same goes for custom instructions and others.

Maybe I am just not aware how to do this? It surprises me that this is not already present. Sounds like reasonably easy feature to implement.