Dozens of people around the world using ChatGPT Plus experience the same issues, apparently since Aug 20th, with ChatGPT’s reasoning (amongst quite a few other issues). After 25 minutes it stops, tells the user that it couldn’t finish the task, then you have to prompt “continue” for it to continue the reasoning, sometimes continue doing this a few times. Does anyone here have further information on this? I let ChatGPT research this issue and it found numerous bug reports on Reddit and openAI Developer Community with the exact same problem, so I know this is definitely something serious on openAI’s end.
For the past ten days I have also had issues and the response I got from OpenAI support was so vague abd corporate as you see in the image and the other image shows that it accepts my prompt but it’s like something in there servers is just broken , Because I been going through the same issues for ten days straight check this post out I listed all the issues I been experiencing myself Why is GPT-5.6 failing long-context tasks?
Hopefully you have better luck than me 
Like you said too corporate/bureaucratic of a reply to know for sure, same with 5.5 Mini
I did run into it recently once after getting the “our systems taking time” message, it does seem like a hidden “nerf”, probably to push long tasks to Work which is usage based. But yeah if you type continue it should be ok, only concern is durability over multiple prompts if it keeps happening frequently, you might want to break down and split larger task into smaller parts. But yeah my only concern if there is a forced finish and the reply it gives can it like be trusted, does it still remember initial task, or does it get like a forced stop and hallucinate/assume the reply, so yeah safest bet seems trying to split into smaller tasks
Yup its been frustrating the issues me and many users have been dealing with over the past week
I pay for pro because its supposed to be able to handle long complex tasks but over the past week it feels like im not getting what im paying for i dont know what happened behind the scenes everything was working fine before August 19th and August 20th but after August 19th and August 20th it just broke and OpenAI either changed something or refuse to acknowledge that something in there infrastructure really broke support really hasn’t helped even with all the evidence I have sent them so its been frustrating
I just hope it’s because of a positive reason: Maybe they are preparing their infrastructure for the next model Astra …?
@MyBrainIsNotAI yeah it’s a valid point with astra whatnot however I still think given the introduction of Work this is a planned move to force hard expensive tasks to usage based metrics
@mcmillanjoseph18 yeah you’re right I think 5.5 mini and the drop from 100 minutes to 25 happened in the same time frame. 100 to 25 is probably an intended change and 5.5 mini which doesn’t even officially exist not sure maybe it was a trial but quite harsh to try counting usage not only off 3000 messages but perhaps usage based and when you exceed allotted reasoning tokens you get moved to a worse model however we already have 5.6 Instant not sure if Luna or what but yeah 5.5 is more expensive to run not sure
Yeah I don’t know what been going on the past ten days I don’t know if OpenAI is doing something behind the scenes wirh there infrastructure, changing something, preparing for there next model i dont know but the past ten days normal chats have been broken ,5.6 has faced serious regression issues, now even work and projects are being affected i feel like 5.6 was just a test and that the next model is the one they wanted to actually release i dont know but I have not been able to do long complex tasks and im a pro member i wish OpenAI would come out and just say something
Maybe this is why 5.6 is having so many issues and there have been so many issues in the past ten days First outputs from GPT-6 "Astra" model from OpenAI maybe OpenAI is preparing for the next model but if that is the case they need to fix all the current issues before releasing the next model and fix any bugs that chatgpt has been dealing with in the past ten days