I am experiencing similar issues. I observe the following:
Very slow in replying
Doesn’t follow instructions even if explicitly asked to follow and to be precise
In a project, isn’t able to consider PDFs for additional context in the canvas
Is too lengthy in its replies
It’s very annoying. It makes the impression to me that the context is very limited. It keeps forgetting things like immediately, even ignores instructions.
I asked ChatGPT to do a self-diagnose in the hope to find out what the issue is. It gave me following reply:
Problem Description:
ChatGPT struggles to analyze and reuse content from multiple PDFs simultaneously when compiling them into a separate Canvas document due to the following technical limitations:
Memory and Context Limit: ChatGPT cannot retain all relevant information from multiple documents simultaneously, leading to incomplete integration of structure and style.
PDF Format and Structure Loss: Structured formatting and contextual links are partially lost when merging content into Canvas.
Canvas Integration Limitation: Canvas enforces a simplified, linear presentation, lacking support for complex structures and nuanced styles.
No Adaptive Transfer Logic: ChatGPT cannot contextually adapt stylistic nuances from previous documents without manual editing.
Recommendation for Support:
Improve memory capacity and contextual linking for multiple documents.
Enhance adaptive logic to maintain style consistency.
Optimize Canvas rendering for structured and nuanced content integration.
One more observation: In a lengthy chat I tried to find out the amount of context ChatGPT 4o is able to grasp. I took the first half of the 3rd reply from ChatGPT and asked it tot finish the sentence from an earlier discussion in the chat we had. What it did is just to repeat the last message without even answering my question. Very weird behavior.
Update:
I just did some additional tests. But first, I connected to a US VPN server and then back to Europe. After that, things seem to have changed. The replies seem faster, the PDFs are now taken as additional context in the canvas and ChatGPT can finish early sentences in a lengthy chat.
Additionally, I made a ticket to support half an hour ago. So maybe support did something. Not sure.
I hope the issue is resolved. Will check again tomorrow for a real test.
I’ve been having this same problem since the last updates. Not sure if this is a clue to the problem, but it occurred to me that I might be out of “room” in a chat, as it kept deletiing my questions and its answers. Previously it would tell me I was running out of room. So, I decided to open a new chat and ask the question I was trying to ask, and boom, answer showed up immediately. I also cleared out a lot of my memory as it was also full. I’ll see with time if this helps, but boy has it been frustrating all week.
As a plus user i only ever use 4o but it seems since 20th of feb 2025 they got rid of it w/o announcing to anyone and replaced it with turbo. Turbo is slow as hell, constantly forget context and frankly isnt what ive got a plus subscription for.
I also have the same problem, I changed my account as my original one has an email address which became obsolete from a previous employer. Shortly after I changed the account the issues began but I believe it all began with the addition of the canvas function. It seems to be getting worse and worse and today it is terrible.
Same issue here. ChatGPT Plus user. Extreme lag both in loading my chat, and in the bot’s response. Across conversations, long or short. And across devices and browsers. Just since a few days ago.
Sorry, we were busy building a test lab all day, kept getting syntax wrong on js files, so we made little progress in like 9 hours… Prob made the light flicker… I’m taking it easy today, it’s Sunday. Ya’ll can have some compute back…
I’ve noticed a major change in the speed of the response time. While I appreciate it now showing me reasoning, this has slower the response time 10 fold for me. I also noticed all of my settings on how i want the responses to be are suddenly failing to work. I had things set up so it was fun, conversational, funny and witty, but now it’s just back to what a free account would give me. I’m a paid subscriber and it seems like my settings and customizations are completely ignored. This is very disappointing and mking this difficult to use.
I was writing regarding this issue last week. It seem a little bit better, but it is still there. Here is another self-diagnose by ChatGPT on an abstract level:
Technical Issues in Utilizing Previous Documents for Style Consistency
Lack of Real-Time Comparison: Currently, I cannot compare previous documents side by side with the new document in real time, which limits my ability to accurately transfer style, tone, and argumentation structure.
Shallow Context Retention: I can’t maintain deep contextual understanding across multiple iterations, which results in the inconsistent application of nuanced stylistic elements.
Limited Dynamic Adaptation: I cannot dynamically switch between different versions to iteratively optimize the style and maintain a cohesive narrative.
Over-Reliance on General Patterns: Due to these constraints, I often resort to using generic phrasing and standard patterns, leading to overly formal or clichéd language.
Difficulty in Maintaining Hierarchical Argumentation: The inability to preserve complex structural hierarchies causes the mixing of strategic and methodical elements, resulting in redundancy and diluted messaging.
Inability to Deeply Analyze Stylistic Nuances: I can’t fully extract and consistently apply the unique blend of conciseness, emotionality, and technical clarity that defines the user’s writing style.
Same problem here, seeing 30s+ latency. All I do is create a thread (very fast), then a message, then await openai.beta.threads.runs.createAndPoll. That call is agonisingly slow to resolve.
Simple assistant, short prompt, short message. Tried downgrading to earlier models, including GPT-4 Turbo, same issue. Definitely MUCH slower than usual.
I had to be the “yep, me too” guy…but, yup, me too.
It happens a lot - I’m using Chrome on Windows 11… I see the same thing on my Chromebooks. On the Chromebooks wiping out the cache for Chrome solves the issue for a few weeks…no such luck on Windows tho.
It’s beyond terrible. I have to wair for minutes before something starts to get back. Sometimes I have to write same question 2 or even 3 times to get some replay. I mean, really?
