Gpt-4 is practically unusable. What changed?

I paid for a subscription because I was impressed with gpt-4’s capabilities. After a couple days of use it suddenly became practically unusable. The response length is a fraction of what it was (just cutting off in the middle), nearly every submission generates an error message and requires repeated resubmission before it will finally respond. Pressing “regenerate response” uses up your quota, which is extremely frustrating. My experience with gpt-4 lately has just been trying to submit a single message, and by the time I have hit “regenerate response” enough times for it to actually reply instead of giving an error message, the quota is used up entirely. From what I have heard from other users, they are having the same problems, with the exception of new accounts who seem to have no issues. Is openAI throttling existing users in order to impress new accounts?


+1. It is way “smarter” than 3.5 but almost unusable. Essentially every response either fails entirely or outputs only partially, unless you cajole it into breaking its response up into many parts, which means more messages and more chances to time out before sending one of the parts…its a mess. Nearly unusable.


The best thing to do is unsubscribe until they fix it. When the revenue drops enough you can guarentee they will prioritise the issue.

Yes, I agree I’ve been using it to summarise data from tables by copy pasting the table in to the chat, now it is fobbing me off telling me to go and tabulate the data myself in google sheets.
This is after repeated prompting. also it is much worse at translation.

its not doing what i ask of it, simply refusing to behave like it did back in march

Why on earth did they pretty much nerf it?