This feature aims to reduce needlessly changing the models from the assistant when creating a new chat. Currently, the assistant appears to be using the model used from the most recently listed chat. This is a minor irritation when switching between models, since I might've used GPT-4o in the chat I just deleted, but a chat that I have pinned (which occurred far in the path) was using Claude 3.5 Sonnet. This will cause the preselected model (if opening assistant from another tab or device) to be Claude 3.5 Sonnet, rather than 4o, which is what I intended (as my most recent chat was done with 4o before being deleted).
I would like the preselected model to be saved based on the most recent chat, and not the chat the chat that appears at the top of the chatlist. In most cases, this is what the user wants and it saves a few clicks from changing the model in a new chat.
