I generally agree. Perhaps there's only a 5% chance that it would write better code or find a bug that it wouldn't have with a lesser model, but the economics of bugs is strong enough that preventing a single bug is worth hundreds of dollars.