Comment by arshxyz

Cost per Token is not representative of the actual cost incurred because a cheaper model can waste a lot of tokens to accomplish the same task. Also this does not consider cost for cached tokens which can often make a "cheaper" model more expensive even when total tokens spent are constant