I mean, it’s fascinating OpenAI, considering what it does, it can’t make a chat the loads only the last 10 replies dynamically with lazy load instead of dumping the entire chat on the browser. It helps no one.
err they did that before … maybe someone had some complains about that?
I actually slightly fixed this (for coding projects) by using nativefier build and an injection script that deletes all but the last 20 articles in the main. It seems to come down to large (600+) articles – an article being one message in the chat bogging down the browser DOM.
I tried doing a quick and dirty VirtualList for the chat but tbh its too much work and I’m busy in a project.
Even after deleting the old chats and orphaned nodes submitting a reponse starts to slow down. I know react shadow Dom should slow with large load, but I think it does something like it.
I’ll go back to tweaking the cleaner script and the virtualizer script as a hobby but I really shouldn’t have to. The desktop app and the ios app seems unaffected but I’m on a 2019 mac intel and they don’t have a desktop app for me.
Hey fellow makers!
I just wanted to share a tool I think is really cool. It’s called Json2Media.com and it lets you create custom AI assistants without being tied to a single LLM provider.
With this platform, you can:
- Build and customize your own AI assistants
- Define models that fit your needs
- Connect your assistants to your models with ease
- Get access to free tools to help you get started
I used it and it’s still a work in progress, but I think it has a lot of potential. Give it a try I think few of you may like it.
How is this still not fixed but has become much worse to the point of unusability?
Here’s how I convinced Gemini Advanced that OpenAI programmers are idiots, compared to Google programmers, because a typical Gemini tab uses less than 1% of the memory a typical ChatGPT tab uses.
I have the same issue: 1.6 GB of RAM usage in the single Chrome tab that is ruunning the GPT. When I use the Native app on my Windows it’s a similar situation if not worse
I can dare telling you all for sure that, since the increase of memory use seems not always happening, that they use your browser to process first someone else prompts to distribute the load of their own servers.
I found a solution, I asked chatgpt to create a plan to avoid memory saturation, it made the detailed plan, gave me the plan and a key phrase. I tested this and it works wonderfully. I’m on a long project and it got too saturated
Same here. 32gb of RAM and fans at max speed and max CPU usage and PC destroying itself. The moment I close chatgpt 99% reduction. I believe it’s because unlike other AI chatbot websites, it ALWAYS renders EVERY SINGLE message up to the top in the DOM… This has been a solved problem long before LLMs. Messaging websites and social media websites have infinite scroll algorithms. But chatgpt doesn’t making it nearly unusable after half an hour into a conversation.
My best workaround is prompting it to summarize the entire conversation for me so I can then move to a brand new one and it won’t “forget” everything.
So, OpenAI: the reason you need to fix this is that it is causing users to make multiple duplicate submissions to your service to get as single result on screen. I’ve observed this, repeatedly. This could be costing you more, transactionally, than older models. If fixing it for users is not compelling, fix it for your overhead. Thank you.
Still an issue and for me personally this has gotten to be much worse since GPT-5 - 2-3x a day the browser tab in Chrome gets killed with an OOM / unresponsive message .. I only have Toby and EatThisCookie and a password manager as plugins, nothing else, and I haven’t changed anything with those in many months.
I am on a workstation with 10+ cores and 64GB RAM and 33% RAM left free - not even cached, free
It was validating to find this thread and also concerning that people have been reporting this since May 2024 - have any of you seen improvements in behavior since May (for those who reported it that long ago) - or have you just developed workarounds?
All it takes to fix it is to virtualise the message list but they won’t do that
Still seeing this issue as of 10/24/25 I cannot perform tech stack migrations using ChatGPT or Codex without it creating a 6GB RAM tab consumption and slowing the UX down. This is unacceptable. Good thing I am not paying for Pro if the issue is there as well. Have GPT Plus and Business. This should be at the top of the priority list for OpenAI. This negates the whole purpose of memory. I am tempted to go to Claude if this issue persists.
its 5/12/2025 now, and they still not fix this bug yet
**!!! VERY dangerous eploits and hacks detected in trusted chrome plugins in the chrome store (with green badge, and irst search reuslt included with 600.000 installs, green badge. Ths is VERY dangerous and has ben collecting all (company) ai chats in localstorage and sended it in batches to very similar servers. This is very likely also the cause for some of the problems here for so,e people. its also already found in different vpn plugins nd privacy plugins that are well known, trusted and still working but injected with collection tools of ALL data in your browser. this should be pinned here maybe (and also at your work for every employee to check ;)) at my software company already three persons had the malicious variant installed.
this explains it very clearly in a cyber security report (those plugins are still in the store by theway and under review now by google):**
Ssorry openai for still pasting links but this is very very important tfor evryone and VERY dangerous. (replace dot with .)
www dot x dot security/blog/malicious-chrome-extensions-steal-chatgpt-deepseek-conversations/
de dot .to/manupranay_adupa_c1dd6247/malicious-chrome-extensions-steal-ai-chats-how-to-protect-your-conversations-in-2026-2f6p*\*
EVERYONE please directly put this under attention as this could have caused the extreme slowness or at least worsened it a lot:**
I have added the cybersecurity urls with extended details at the end of this post (you should IMMEDIATELY notify your work with this, as 99% of employees had such a plugin intalled and laso devs.
It mined chats (so also designs, company secrets, company document, company code, requests containing sensitive information, passwords and everything else thats obatinable in the chrome browser without noticin because it wer eliteral copies of the real apps, just working as normal but with injected all those things.
Here a shrot summary:
What Happened in the 900K AI Chat Theft Campaign?
Two malicious Chrome extensions pretended to be legitimate AI assistant tools and quietly stole users’ AI conversations and browsing data. They targeted popular AI platforms, including ChatGPT and DeepSeek, and even appeared trustworthy inside the Chrome Web Store. (including the verified badge and 600.000+ installs)
The campaign was discovered by OX Security’s research team in late 2025.
The fake extensions mimicked multiple well known chrome plugins for chatgpt while remaaining functional, but added interceptors for all fetch calls and XHR calls being done in the brwser, and immediately collects every chatgpt chat the user has, the complete history by putting it in the localstorage and then periodically sending it to their own servers.
^^ This was a BIG part of the performance problem discussed here. I would sugest everyone STRONGly to immediatelhy check for the plugins and delete them.
Once installed, they scraped AI chats directly from the browser and sent the data to attacker-controlled servers every 30 minutes.
again sorry openai for still pasting links but this is very very important tfor evryone and VERY dangerous. (replace dot with .)
www dot ox dot security/blog/malicious-chrome-extensions-steal-chatgpt-deepseek-conversations/
Well, I would debug that but I think it is against the terms of usage to analyze how it works.
The issue is still present Chrome Version 146.0.7680.80 on MacOS, I had this tab open and put my computer to sleep and when I came back next day I found the tab crashed like this, this is clearly because of my usage as it happened when my mac was idle, this is probably a memory leak you guys have somewhere, and this isn’t the first time it happens, actually it almost happens everytime I leave ChatGPT open for too long.
Hey everyone, Huge apologies for this experience. Can someone please confirm that this is still the issue? Also providing a recent screenshot with high ram usage would help. Thank you!