My current optimum workflow is using two models simultaneously as seperate tabs like this:
1 low cost model on ‘Ask’ mode, for general enquiries, clarifying questions, and especially for prompt optimisations to feed the second agentic model.
AND
1 premium/thinking mode model (personally Sonnet 3.7) that executes these polished prompts as an agent.
I like this method because it saves precious context window space by first using the cheaper model to optimise your prompt, so you can feed the higher thinking model more specific and concise instructions.
The feature I’m requesting to be implemented is the ability to select a unique model for each conversation tab in the AI sidebar menu. At the present, one model is selected globally for all tabs, but this is sub-optimal.
Thank you for considering!