Ok, so hear me out… you can use gpt-4o image on the api and then send it to o1… how come on chatgpt we can’t do that?
it doesn’t make sense to me, like, sure, I can use my own api keys to do so (not sure if o1-pro is the same as o1 high, but ok) but why is it limited on chatgpt?
could anyone understand the reasoning behind blocking such functionality? its not like it would cost oai more since you can just have two tabs open, if anything it might save them server hosting costs… right? right?? anyone?
Hypothesis 1: those who don’t know what they are doing are going to use the api and waste their credits
why not hypothesis 1: because credits have half lives and anyone with brains might as well just use 2 tabs to do the same…
Hypothesis 2: (tech debt)there some nasty bug out there, either in sec, dev or ops, and adding such feature would break something along the way (and its getting worked on for future releases like o3)
Hypothesis 3: o1-image model is getting trained
Hypothesis 4: its not really a priority
Hypothesis 5: no one though about that for some reason
Hypothesis 6: it would be misleading to use image on o1-pro (or o1) that is really 4o, so someone using just image prompt to o1 might get the wrong idea or it would get benchmarked and competition would claim o1 and o1-pro do not perform well and over sensationalize it to skew public opinion
ok, I’m going with hypothesis 6 and I’m assuming there was some risk document staying that it was just not worth it