Luckily inference is cheap, and other providers offer efficient models.

It’s only going to get better in the future.