A lot of newer models are geared towards efficency and if you add the fact that more efficent models are trained on the output of less efficent (but more accura

A lot of newer models are geared towards efficency and if you add the fact that more efficent models are trained on the output of less efficent (but more accurate) models.... GPT4/3o might be the best we will ever have