While I'm sure some of the open weight providers do this as well, I think the comparison is frontier labs v local inference.

I'm not sure that is the comparison - the OP is planning to run an open weights model locally, the obvious comparison would be paying a hosted provider to run the exact same model