Good realease but the annoying part is they're very unclear about which types of models they are comparing. They provide benchmark comparisons for the base mod

Good realease but the annoying part is they're very unclear about which types of models they are comparing. They provide benchmark comparisons for the base models only and arena comparisons for instruct only? Was that intentional? Why would you ever do that? This makes things unnecessary complicated imo and the only payoff is a short term win for google on paper. Guess I'll just fully test it for my own tasks to know for sure