Without latency, throughput, and benchmark filters I don't see the value. Of course the dumbest and slowest model is going to be cheapest.

[flagged]