Hi Folks, is it just for me or is ChatGPT 4 and 4o exceptionally slow? No one seems to mention it but the Mac app and web are almost unable for me. (Paid user)
Any ideas?
16 Likes
For me it may be worse, and I am actively trying to use it for things I am trying to do. I have had to simply abort a bunch of times even to the point of giving up, shutting down and logging back in.
It is also behaving erratically. I asked it to provide a prompt to create an image for an article I was using it to develop. I asked it to do that and then use the prompt to generate an appropriate image. It generated an image that was almost unrelated and afterward, instead of giving me a rational prompt it proceeded to reprint the article, so slowly I wondered if it was broken.
1 Like
I thought it was just me but this is horribly slow. Itās smart but turtle-like in rendering the response
5 Likes
same, and itās even occasionally come with error, which wasted the remaining usage I have to use 4o, it is frustrating really
Hello, I am glad that I was not the only one experiencing this. As a daily user, I can say that its performance has degraded. I am a free user as I use it just to bounce off ideas, but nowadays, even doing that is hard. The default model runs smoothly for me though. But 4o takes much more time to generate a single response.
2 Likes
Same here, it does not happen always though. I currently have around 1.5 to 2 tokens per seconds. What is going on? (Chatgpt App in mac, Model 4o)
1 Like
same for those who paid (windows dell)
4 Likes
I just mentioned this in another post, but I think itās important enough to bring up again here. Since the release of ChatGPT version 4o, thereās been a noticeable lag in response timesāit sometimes takes over five seconds just to send a prompt after pressing āEnter.ā This delay is something I wouldnāt expect as a Premium subscriber, where quick and efficient service should be a given.
Beyond just the increased wait times, Iāve also observed a dip in the quality of responses. This combination of issues has me seriously considering whether to continue my subscription. It seems as though the system is overwhelmed by the volume of new users, and this is impacting not just speed but also the quality of interactions. Itās disappointing, particularly because one of the main reasons I opted for Premium was to avoid these kinds of issues.
I believe itās reasonable to expect that part of our subscription fees should ensure a certain amount of dedicated bandwidth for paying customers. If these problems persist, it might just be the push I need to look for alternatives that can offer consistent quality and responsiveness.
2 Likes
Perhaps the issue lies with the Mac app rather than GPT-4o. It generates responses quickly on the web, but its performance on the Mac app may give the impression of slower generation. When the conversation gets lengthy, the smoothness of the Mac app significantly decreases, and the app occupies an entire CPU core. This might require the OpenAI team to optimize their Mac app.
1 Like
From my experience, this is most definitely not a Mac issue. My activity monitor shows low cpu and memory usage while client is outputting code at speeds approximate to 2 or 3 words per second.
Yes, the ChatGPT client is very slow. Iāve noticed a difference between the Mac client and the web browser version. Specifically, the Mac client never shows the ācontinueā button like it does on the browser when outputting long posts. Perhaps this is affecting the speed and causing the general sluggish output? Itās just a guess.
This not only remains an issue, it is getting worse. Particularly annoying is that it is so slow it seems to hang and when it eventually does some of what was requested it requires repeated restarts. At one point it got completely confused and just generated garbage.
Getting the wrong answer quickly is generally not a good thing but itās at least preferable to getting the wrong answer after a long wait and then having to start over. Note that I am typing this in while waiting on a response.
It seems to me that the system is overloaded despite paying for the access. It appears that in order to lighten the load on the backend OpenAI has made ChatGPT issue a discussion as to how it might do what you have asked without doing it or by emitting documents that have placeholders for the things you have asked it to generate. It also stops midstream to require pressing a button to continue doing as asked. This, I am sure, is not too bad if you are not actually trying to get real things done. It is quite frustrating when you are.
I appreciate that it is challenging to manage an enormous system like this, but as a user it is disappointing to have spend what is literally now going on an hour attempting to complete something that should have taken a minute or so. Perhaps if someone from OpenAI reads this they can comment as to why things are like this and what is being done to correct it.
I am going to close this comment now, but for the record, during all this time I have been waiting on ChatGPT to complete something and it is still frozen with no indication as to whether it will continue or not. Perhaps the interface could be reworked so that it (a) Worked until it actually completed the task. (b) Gave some indication as to how long it will take. (c) Allow you to batch submit it and return to collect the output when it is done. (d) Give clear indications of progress and ways to inspect progress on demand. (e) Allow reset when ChatGPT has gone off the rails and is misbehaving.
1 Like
I had the same problem. Closing the app completely helped, now i get the answers as fast as the web app.
For me 4o seems to take forever to process the prompt without giving any visual indications if the prompt is even accepted. The arrow just stays grey like nothing every got through. And I canāt continue. Usually I just copy paste the whole prompt (since nothing happens after pressing enter or the arrow) to a new chat window and sometimes that works. Iām not uploading any files or images just text.
I thought, it might only be happening with unpaid users, I was only facing this issue in IST evening time, but now i am facing it 24 hours. they should atleast have different servers for different regional users.
1 Like
Same here. It is almost impossible to work with it. Whatās the point on paying if I can barely use it?
Is there any way we can contact OpenAI with this issue?
1 Like
I agree. It is almost useless. It took 5 minutes for it to respond this morning. I timed it. Anyone know how to fix it?
1 Like
Interestingly, on my phone works perfectly. Not sure why on my computer does not.
This is so frustrating. Iām paying for Teams accounts and didnāt expect this long wait for each response.
Incredibly frustrating. The paid account serves as my digital assistant and I can barely do anything now. Since the release of 4o, getting anything done is almost impossible. Painfully slow, repeated network errors. Tried all possible suggestions but nothing helped. Wonder when this will be addressed?
1 Like