Luna is at a very compelling point on the price/performance curve.

I have found that sometimes a smaller model with max reasoning is actually more expensive than using the next tier model with a lower reasoning effort. It’s certainly faster.