We have been using the API for 2 months (3.5-turbo) and basically, it generates sections of an article and compiles them. It began outputting more tokens as of today, 6/28 when the prompts haven't changed at all. It ends up being 1,500 more per article. Did something change causing this? It adds up to a bit more of a cost every month at the volume we're doing, and it seems to be out of nowhere
#API Outputting More Tokens Than Usual on 6/28/23
stable seal
stable seal
I see the system changed the 3.5 turbo models yesterday. That is likely the cause...