this is what I have been struggling to understand too. the accolades speak about the size and other input related capabilities of the model, but the output - not so much