You bring up a great point. If it doesn’t show the the per prompt cost, then its hard to understand which model produces the best value for particular needs.
As for the 2nd part, the in IDE percentage lags behind sometimes up to an hour or more, so it could be that it had not fully updated since recent requests when you exited the IDE.
The solution would be at least doing what they have been doing, which is show per request costs. I think this change is only affecting certain users or team users. Because I still am seeing per request costs (Pro plan)
Also you can try this