When I first signed up for GPT Plus so I could access GPT-4 about a month ago, the AI was a polished machine. Major errors were hard to come by and memory lapses were almost a complete non-issue. In the past week or so I’ve began to feel like it’s performance has regressed to be more reminiscent of GPT-3.5. I’ve begun to question the claim about 12 pages worth of token memory with GPT-4 as it now not only seems to have more frequent issues remembering prompts in their entirety past around 5-6 pages of tokens, but also disregards certain parts of a prompt that I even emphasize in its responses. Was wondering if anyone else was experiencing this too?

EDIT: Did not expect this big of a response so let me help address some of the most frequent responses I’ve seen in here.

-I don’t care that this has apparently been asked before. Not everyone is on Reddit all the time. If you’re on here so much that you constantly see this question I would recommend closing Reddit and going outside for a few hours as well as removing the stick from your ass.

-People are asking this question probably because they’ve been utilizing GPT-4 from their paid GPT Plus subscription since it’s release and therefore it wouldn’t be far-fetched that someone utilizing a tool on a daily basis would notice if it starts to become dull, especially if they keep using that tool for the same jobs. I have a feeling most of the people saying “no the hype has just worn off/the honeymoon period has ended” literally don’t even have GPT-4, much less have had it since it’s release.

-Nobody wants to comb through hundreds of old chat histories and draw up a screenshot comparing it to a similar recent conversation in a way that’s irrefutable just to prove a point to you. See above.