My general thoughts are that we want to slowly move towards removing this decision matrix from most users - in an ideal world they should be able to just talk to a robot and the engine underneath should figure out what the best llm is to serve the request.
It does add some overhead but also eliminates some decision paralysis non advanced users have.
What does “upgrade subscription” mean in this context, upgrade subscription to Discourse instance so you get higher limits, nudge admin to reconfigure llm?
There are a fair amount of assumptions built in here.