open weight models wouldn't be processing more tokens if that were true

many companies are now using both, it's only been a few months since it really picked up