I have always suspected that the o1-Pro is some kind of workflow on the o1 model. Is it possible that it dispatches to say 8 instances of o1 then do some type of aggregation over the results?