For 90% of my questions, a lightweight model like haiku or o4 mini would be sufficient and save time. I'd be much more willing to be economical about my model usage if I could easily switch to a beefier model without leaving the chat text input. I think this would likely save kagi money in the long term, as well as being nice for users.
The ideal would look something like:
1) User sets a default model. This is the model that is active whenever a chat is started.
2) User sets 2 or 3 "backup" models that they can toggle between using a shortcut
3) Hitting the shortcut brings up a dropdown menu, with the next model selected and further invocations continue to switch
