OpenAI announcedthat it has implement a new version of itsGPT-4olarge linguistic communication model to force itsChatGPT chatbot , but it has declined to specify exactly how the updated role model differs from its forerunner .
“ To be clean , this is an improvement to GPT-4o and not a raw frontier model , ” the companionship post on X ( formerly Twitter ) Monday .
there's a novel GPT-4o modeling out in ChatGPT since last workweek . hope you all are enjoy it and contain it out if you haven't ! we think you'll like it 😃
& mdash ; ChatGPT ( @ChatGPTapp)August 12 , 2024
“ We ’ve enclose an update to GPT-4o that we ’ve institute , through experimentation results and qualitative feedback , ChatGPT users incline to favor , ” the companywrote in its Model Release Notes . “ It ’s not a new frontier - class model . Although we ’d like to tell you exactly how the theoretical account responses are unlike , figuring out how to granularly benchmark and communicate manikin behavior advance is an on-going area of inquiry in itself ( which we ’re work on ! ) . ”
In the absence seizure of specific details from the company , many drug user have , unsurprisingly , begun speculating as to the nature of the changes and whether they represent new feature . X exploiter @misaligned_agitook to social medium to venture that the newfangled update had go through a multistep logical thinking method acting rather than an entirely new model .
OpenAI cursorily put the kibosh on that dividing line of thinking , witha spokesperson tell VentureBeatthat it was n’t in fact a new reasoning process and that the behavior that @misaligned_agi honor could have been triggered by the structure of their prompt .
Other users also vocalise their theories on societal media , arguing that GPT-4o recently began comport in subtly different and better ways , and that its image - generating quality had meliorate . “ For the first time in a tenacious time , it provided good ‘ vibe ’ on an yield than 3.5 Sonnet , ” observed X user@mattshumer _ .
After allowing user to take their better guesses at defining the new loop , which OpenAI is call ChatGPT-4o - latest , the companyadded a few short detailsonits Models pageon Wednesday .
Described as a “ active framework continuously updated to the current version of GPT-4o in ChatGPT , ” GPT-4o - latest has a knowledge cutoff of October 2023 , and can accommodate 128,000 tokens , or 96,000 words , per conversation , just as the previous GPT-40 version did . It can output 16,384 tokens , or 12,288 words , on par with the newer GPT-40 - mini model and or so quadruple what the older GPT-4o can do .
Unfortunately , hard stats like these do n’t offer brainstorm into what the unexampled model is actually capable of , and , seemingly , neither will its developers .
