Uses less quota (i.e., cheaper). For people who like to keep their contexts small, this is a no-brainer.
reply> k3 (1M) consumes about twice as much quota as k3-256k
replyCheaper?
k3-256 consumes half the quota, so... yes? (assuming the user isn't making use of the larger context window)
reply