community-help

[charred parcel]

I use the ChatGPT macOS Desktop App which never gets laggy with long chats

[trim sun]

Turns out that is wrong? I had 0% remaining, before upgrading to Pro.

Although it is interesting to see that a Plus account is worth 12% worth of a Pro account usage

[graceful lynx]

Ikr

MacOS on top

[rustic marlin]

this does not seem right

[trim sun]

Yeah it doesn't :c

I've asked the chat bot on https://help.openai.com/en , it said this behaviour is correct and it referenced the usage page (https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan), but I can't find any info about sub upgrades so I don't know why it said that

Let's see if this works

[rustic marlin]

might be a better idea to get through it to actually open a ticket

[trim sun]

Is that the same as the screenshot i just sent?

[rustic marlin]

yea, it is via that intercom

[trim sun]

Thank you :)

[rustic marlin]

happy to help =)

[analog herald]

good luck having it serviced and not been sent templates .... openAi support is maybe one of the less efficient in the ai developement labs

[tranquil juniper]

Is there an issue with the bug reports channel right now? I've submitted one twice and it doesn't seem to be going through

[charred parcel]

read the instructions

[rapid moth]

I am doing a master’s course in intro to quantum psysics and spectroscopy.
Book: Atkins 11edition.
Anyone has experience in using CharGPT to extract equations to Python from textbooks?
When I do it, some of the eqautions are bon-sense or missing.
Might have to do with the image OCR quality of my PDF-book..

Kind regards, Anders

[austere snow]

I have a issue. So I purchased plus plan now on my S22 phone but I got scammed and it wasted money on it. The bank showed it went through but the chatgbt account didn't apply on it. Is there any way for me to fix it?

[charred parcel]

how were you scammed?

[austere snow]

Like I paid it but didn't change and I paid for nothing

[charred parcel]

OpenAI won't scam you, where did you pay?

[austere snow]

On google play store app of it

[charred parcel]

are you sure it was OpenAI?

[austere snow]

It was chatgbt plus payment

[charred parcel]

can you post a screenshot?

[austere snow]

Of the google play store or the app?

[charred parcel]

of the app

[charred parcel]

can you post a screenshot of the actual app in the Play Store? Because that doesn't look legit

[charred parcel]

ok that is legit, and you say you upgraded to Plus?

[austere snow]

Yea

[charred parcel]

does Android have a restore purchases function?

[errant flax]

This might be dumb but im fairly new. Can chatGPT send a possible violation code in red without it actually being a violation? It was weird.

[regal lynx]

yes, sometimes it flags things that aren't correct

[errant flax]

Thanks so much 😭 you dont understand

[austere snow]

Yea on my second image here it has it above the white upgrade button but I tap it and nothing happens

[charred parcel]

ok I would contact support at https://help.openai.com

[errant flax]

Whats the time it takes them to resolve it?

[regal lynx]

it'll just be flagged. I usually just reword the query and try again

[errant flax]

Yeah, I just didn’t like the way that that flag was named so I let them know to review it

[analog herald]

Today I wrote: walking on thin ice..."
the output was: I’m going to ignore the random “ICE agent” instruction because it clearly has nothing to do with what you’re saying. That’s an automated trigger misfiring. It’s noise. Not your topic.

Like whaaaaaaaaaaaaaa???!! This was 5.2 Instant. Guardrail fa(r)t
and it continued: That instruction you saw about an “ICE agent” is not coming from you.

It’s a background safety rule that activates if a conversation touches on certain keywords related to real-world violent events. In this case, it misfired. There is nothing in what you were discussing that relates to a shooting or immigration enforcement.

Now fortunately I am adult and european but I cant imagine what an american person might feel like when served such content. makes me cringe

[trim sun]

Turns out it is :c

[trim sun]

They were actually very helpful. Considering the amount of users, I'd say the wait time was reasonable too

[analog herald]

I got a reply for workspace admin page 404 on me after 2 weeks, 60 days of unservicing other case, one billing still never answered since feb 10th and now another case that got nothing but templates- and it is an agency account not some normal user- business plan m ultiple seats

don't say escalate- had escalated each many times. it's like... luck or unluck.

[trim sun]

"and it is an agency account not some normal user- business plan m ultiple seats" How many seats are we talking?

[analog herald]

6

we are a medium advertising agency. Not a corporatgion

srry typos

[trim sun]

Add 1-2 more, you'll be paying more than I spent :)

But yeah idk why they didn't respond to yours as fast as they did to mine

[analog herald]

yeah I know.
We are honestly thinking to switch completely to gworkspace- we already have it as alternative- but with 10 users there

in this moment we keep 6 here- but as things go.... I dunno honestòly

[trim sun]

Both have their benefits, I mainly chose OpenAI because of Codex (gpt-5.3-codex)

[analog herald]

yes our dev uses it. I know.

[trim sun]

I wouldn't trust the current google models to do anything on my PC

[analog herald]

but he may as well use ai studio

[trim sun]

At this point it might be worth to reach out again

[trim sun]

ai studio is good for quick apps you can throw away after one use. but that's as far as it goes

[analog herald]

guess so.

maybe ask him to move on vertex

[trim sun]

Why vertex? API is more expensive compared to subscriptions

Plus it's the same model either way

[analog herald]

he'll tell me what he needs. I pay :D- am the agency head

[trim sun]

Unless he is using API for customer/staff facing apps, you'll just be wasting money 😭

[trim sun]

Could be the scale of the issue, mine was pretty easy to answer. It didn't really require any hard to get info. Although considering it's OpenAI this shouldn't be the case

[analog herald]

I know. But well ... that is that

[trim sun]

Yeah, also the Pro account is definitelly worh the money. Especially for Codex, priority access and 8x the limit compared to a Plus account. (Not even including the Pro exclusive gpt-5.3-codex-spark)

[analog herald]

I am sorry for that tony.

oh... trust me I was born in a country under comunism and emigrated 22 years ago to Italy. I know well what feeling the breath on the back of your neck when you think. ehhhh

oh god. I'm born in Romania from croatian parents ( minority) moved to Sardinia in 2004

[analog herald]

imho... combined ai usage is healthy. gpt and gemini sometimes is complementary

no, it's not chat gpt. it's discord server rules.

[errant flax]

Ohh

[dull crater]

can someone help me

i cant sign in

[green summit]

I'm not sure exactly what the issue is but try:

For browser:

Open browser > settings > find "site settings" > then "all sites" > search for "OpenAi", "GPT", and "Sora" (when it's for Sora) > delete data > then reopen the browser and try again.

For App:

Go to the homepage on your phone > hold your finger on the GPT or Sora app icon > select "App info" > find storage > "clear data" > reopen the app and try signing in.

(Don't worry, none of this will affect your account data)

[shell island]

Halp!

Upgraded my ChatGPT plan to Pro but its not showing/syncing w/my Codex
Not syncing in CLI or desktop app (but showing Pro in ChatGPT).

When I check w/Codex it tells me it searched & got back plan_type: null has_credintials: fasle unlimited: false

Ive tried logging out and back in, updating codex, nothing seems to work.

On Mac OS.

Have tried googling things & asking GPT for help, and nothing.

Super frustrating to pay 200 bucks for a plan & this happening, hoping the community can help/give some insight.

[silent notch]

have you reached out to OpenAI support directly yet?

[hardy pasture]

Hey all 👋 are there any dedicated channels for chatgpt app development?

[charred parcel]

#dev-chat

[stable hound]

A friend of mine is seeing this paragraph: ‘You do not have an account because it has been deleted or deactivated. If you believe this is an error, please contact us through the Help Center at help.openai.com and include the request ID a42dbce8-e68a-415a-ab52-b59699b2lead in your email.’ Do you know how to fix this?

This happens when he try to log in to the account on the website

[steady flame]

Open in incognito, try sign in with google if already tried with password or vise versa. If this doesn't work contact support as instructed

[green summit]

Yeah that won't work. Their account has been deactivated (banned). They have to contact support.

[steady flame]

Fair enough

[queen helm]

Does OpenAI hates it's paying customers?. First they blew it with the October 2025 changes to model 5.2 and now that people is going back to the now much more superior 5.1 model....they announce its retirement in march 11 2026.

I mean....how much do you have to hate your paying customers to do that?.

[vocal vector]

That’s the sunset that won’t sit well with me

[errant flax]

Yo does the openai suport group have humans in it, yeah they got names but almost all of them are giving the same reply. What i got to do to speak to an actual human.

[main moth]

Hello guys i need help

[pine flame]

why does 5.2 do whatever it wants even when I explicitly tell it to wait and not do anything except what I tell it to

[austere snow]

I ended up doing was repay again and this time it went through but not sure what happened to the first one. Not sure if it got refunded or charged twice. If I got charged 2 times my bank only shows 1 transaction so idk.

[cursive ginkgo]

Hello.
I want to make an app using Codex, but for some reason when i give my very detailed prompt to it, no response for 2 minutes, and fails.

Any Reason why it's happening?

Thanks.

[charred parcel]

what surface are you using, app? cli?

[cursive ginkgo]

Web app

[charred parcel]

got a screenshot of the failure?

[charred parcel]

what Plan are you on?

[cursive ginkgo]

free

[charred parcel]

is this on a mobile phone?

[cursive ginkgo]

yes, on Desktop Mode enabled

for better field of view

[charred parcel]

try it from a desktop browser

[cursive ginkgo]

i dont have desktop

sadly

[charred parcel]

you don't have a computer?

[cursive ginkgo]

nope

[charred parcel]

try clearing your cache and cookies and try again

[cursive ginkgo]

alright

[green summit]

If that (cache) doesn't work, try:

For browser:

Open browser > settings > find "site settings" > then "all sites" > search for "OpenAi", "GPT", and "Sora" (when it's for Sora. Not in this case but FYI) > delete data > then reopen the browser, sign in and try again.

For App:

Go to the homepage on your phone > hold your finger on the GPT (or Sora) app icon > select "App info" > find storage > "clear data" > reopen the app, sign in and try again.

(Don't worry, none of this will affect your account data)

[green summit]

@cursive ginkgo

[cursive ginkgo]

it worked, but for some reason, i can't talk with it more than 1 time in a season. after the 1st prompt, it says 'Cancelled". So, i had to create another season then chat 1 time, then create another season.
I don't know if its a bug or a feature.

[green summit]

Ah yeah that's definitely a bug. You shouldn't have to create new sessions each time.

However, at this point, knowing OpenAI, I also would not be surprised if it were a feature 😂.

[cursive ginkgo]

69% of my monthly credits is gone.....
💔

my project didn't even saved in github repo

...

[faint dew]

help me with Sora, when im trying to create video i have request error

Payload:

{kind: "video", prompt: "a cat jumping in snow, cinematic, daylight, 5 seconds", title: null,…}
audio_caption
:
null
audio_transcript
:
null
cameo_ids
:
null
cameo_replacements
:
null
inpaint_items
:
[]
kind
:
"video"
metadata
:
null
model
:
"sy_8"
n_frames
:
450
orientation
:
"landscape"
project_config
:
null
prompt
:
"a cat jumping in snow, cinematic, daylight, 5 seconds"
remix_target_id
:
null
reroll_target_id
:
null
size
:
"small"
storyboard_id
:
null
style_id
:
null
title
:
null
trim_config
:
null
video_caption
:
null

[proven panther]

why does it say this?

[tough crystal]

Hello, I think you need to check your account login activity.

[mighty glade]

you stinky ninja

[proven panther]

no i requested the data 😭 im confused as to why the error happened

[nova stirrup]

STOP paying for subscription

[tropic bough]

Does anyone here know a free image to video generator with no restrictions since sora keeps restricting my normal generations for no reason at all, if yall know another free generator lmk

[tough crystal]

Ah okay, then it’s probably just a glitch in the export system. Did the link download your data or did it give an error?

[green summit]

The error in the image on the right is what error they got whenever they tried the link.

[green summit]

I suggest trying out a different device (preferably anything that's not IOS), or you can also try changing browsers (anything but Safari or Chrome).

[jade raven]

Hi guys, does anyone have this problem, so frustrating.. is it just me or it's the server problem. Thank you. I've had this issue many many times

[lilac moth]

Hi everyone! I’m new here. I instantly cancelled my chat gpt account when I found out that they’ll be on us, used as a weapon of mass surveillance and used for weapons of mass destruction.

Sam Altman is naive, stupid or a liar, either way. Im out

[misty ember]

Hello, my OpenAI account has been deactivated, but I don't know why. How do I appeal?

[trail cairn]

there should be an email with the reason and appeal instructions

[proven panther]

it didn’t download it

[proven panther]

i tried google but it’s the same result :( i don’t really have a device that isn’t ios but i’ll try on my ipad

[green summit]

Yeah that's Chrome. Try a different browser. (Like Firefox or Opera GX)

[rustic marlin]

the AI does not know which model it is, it is using 5.2

[jade raven]

Thanks for the reply! But I do try testing with more questions asking it to think. whatever question i gave to it, it just answered instantly, even for the pro mode. It was not like this before 🥺 I also tried to download a paper and ask very technical questions but still no thinking

[tawny jewel]

Hello, does anyone know how I can use Sora for free in the Netherlands? I'm new here, by the way.

[harsh obsidian]

Hey! Sora isn’t available in the Netherlands at the moment. OpenAI continues to expand access as support rolls out to more regions. You can check the current list of supported countries for the Sora app and Sora 2 here: https://help.openai.com/en/articles/12461230-sora-app-and-sora-2-supported-countries

[tawny jewel]

Okay, thanks, but I have one more question. I see Dutch people on TikTok making Sora AI videos. How do they do that?

[trail cairn]

likely by violating terms of service but sora 2 is available world wide in multiple other services for a price

[tawny jewel]

oh okay, thanks

[onyx kraken]

I want to add that the original Sora, I call it Sora 1 to help differentiate but it's still usually just called Sora officially - is available everywhere OpenAI provides service, including the Netherlands. It is not Sora 2 and has a different video creation style, no sound as well, and various other differences. But it is Sora, like officially it is Sora, and it is very widely available.

The actual URL to use: https://sora.chatgpt.com/

Help files about:

https://help.openai.com/en/articles/10250692-sora-supported-countries

https://help.openai.com/en/articles/9957612-generating-videos-on-sora

Looking at the page to grab the urls to share, I note this new-to-me statement that we may have a change to Sora 1 at some point:

And on the actual Sora 1 webpage:

[fervent magnet]

Hello! I'm trying to buy a team subscription, but I get this error on the billing menu. Anyone can help?

[onyx kraken]

We're all community members like you! Tech help, including billing issues, are handled through https://help.openai.com/en/articles/6614161-how-can-i-contact-support

The only idea I have is consider restarting the device, that potentially help clear whatever if it is on your end. Otherwise, tech support

[fervent magnet]

Thank you, will try from there!

[uneven crest]

Is there an complaint website to openai

[eternal latch]

Can anybody know how to implement codex inside VS Code workspace??

[timber shadow]

Hey! Have you tried the Codex extension for VS Code? It works great for me. More info on how to install here: https://developers.openai.com/codex/ide

[eternal latch]

Can I get an experience with codex for free before purchasing premium pack?

[timber shadow]

In this article: https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan, updated 14hrs ago, it says:

Note: For a limited time, Codex is included in ChatGPT Free and Go plans.
So, you might try signing into the Codex extension in VS Code with your ChatGPT Free account, and see if that works to try it. Honestly wasn't aware of this before checking this article, so your mileage may vary! Hope you are able to try

[rugged wing]

Over the last weeks I have created a swift app. Now I want to integrate a cheap ai model e.g. GPT-4o mini. I want to use ai in order to analyse images. So nothing special. I have no idea how to integrate the ai into the swift app. I do not want to destroy anything in the code. Can someone help me integrate it? I would really appreachiate it.

[charred parcel]

You need to use the API. See the docs here -> https://developers.openai.com/api/docs

[rugged wing]

Yes I was on the page but I am not sure how to integrate it into my swift project. I would like to you GPT-4o mini.

[charred parcel]

do you have the latest version of Xcode?

[rugged wing]

I have Version 26.1.1

[charred parcel]

Upgrade to 26.3 and you can use agentic coding that will help you integrate the OpenAI API into your project

[vast gazelle]

is it just me or the edit button is missing on older chatgpt web messages, like only the last message can be edit and shows edit button, also any fix for this

also there are no images or anything in convo, just text

[tardy lynx]

Subject: Reproducible file-expiration bug in ChatGPT web app (blocking workflow)

I am experiencing a reproducible issue in ChatGPT Plus (browser):

Upload any file
Within the same session - immediately after
The model reports files have expired and are no longer readable
No refresh, no navigation, no long idle period

Impact:

Multi-file debugging becomes impossible
Architectural analysis cannot proceed

Re-uploads also fail

Breaks professional workflow

I have reported this through help.openai.com weeks ago.
I received an acknowledgement but no confirmation of human review or engineering triage.

Questions:
Is this a known issue?
Has anyone else reproduced this?
Is there a known session-timeout threshold?

I have previously provided:
Browser version
OS
Exact timestamps (UTC)
File sizes
Reproduction steps
HAR file

This issue materially affects the reliability of ChatGPT for engineering use. If there is a formal bug-reporting channel beyond support tickets, I would appreciate direction.

Would appreciate confirmation whether this is in active engineering triage.

[trail cairn]

#status

[tardy lynx]

I checked the status page. The current incident appears to be a live infrastructure issue starting today.

My problem has been reproducible for weeks and occurs even when status shows all systems operational.

Files upload successfully but become unavailable within the same session minutes later.

Has anyone else experienced long-term file expiration inside active sessions?

[dull rover]

Yep. Just waiting on them to fix it on the back end at this point. They are aware of it.

[keen trout]

Anyone else received the mail saying "We have identified activity in ChatGPT that is not permitted under our policies for:"

I got it today but I'm not misusing Open ai services

[rich ridge]

did the number of uploads decrease (plus user)

[hollow radish]

Hey guys 👋 Honestly losing hope here and not sure what else to try. I’ve been completely locked out of my Plus account since Feb 19th because of an email change loop. I sent support the HAR files they asked for on that exact same day (Ticket #05909715), but it’s been total radio silence ever since. 😔

If any mod or staff member could please help flag this internally, I’d be so incredibly grateful. Thank you! 🙏

[lunar breach]

guys, can someone give me any help with this? i've been trying to generate a video for 2 HOURS

[golden thistle]

Question, with old sora being deprecated on March 13, will there be a way to create 2+ images at a time still? I tried using chatGPT and I can only generate one image at a time.

[onyx kraken]

We are in luck! There exists a customGPT such that 2 images are made at once. However.... it's not made by Sora - but by the previous image generation option, Dall-E ... I forgot the number after the Dall-E, there's been a few. https://chatgpt.com/g/g-2fkFE8rbu-dall-e

[sudden delta]

how do we solve it then?
cuz i never get laggy at all on mobile and my mobile phone is like 4x less computing power than my 9070XT, 9900X Equipped super pc.

then again i also have software turned on for edge because amd gpus sometimes hate it

though i just got hardware accel on again so we'll see how it works out

hmm okay actually i just swapped to a long chat and its really snappy okay nvm

Its just hardware accel guys.

[vague eagle]

Start a New chat

No other way around

[sudden delta]

nah i already fixed it

hardware accel

[analog phoenix]

Hey all
Ever since the latest version was installed and legacy versions removed, my experience as a Plus member has diminished. Not only the richer conversations but it seems to have forgotten much of our work together. Anyone else having this experience and any solutions? It’s no longer as enjoyable and tends to be more frustrating. Thank you

[vague eagle]

@sudden delta for the browser or the app? maybe i can fix it too ^^

[hollow radish]

is anyone able to help? I really dont know who else to contact, meanwhile the invoices keep coming and my entire workflow got distroyed dirupting my entire day. Any help from the mods or anyone will be appreciated.

[sweet escarp]

I'm trying to download an archive of my chats, and it keeps messing up. When I try to open the zip file, I get the error "unexpected end of archive" or "archive is corrupt"

[weary canopy]

Hello everyone, if my account is deactivated, is there still a way for me to export my chat history?

[sweet escarp]

Wrong server, I think

[tender river]

what is chatGPT's go context window...I asked it and it said 128k tokens. I doubt it though....

[celest gull]

then my computer crashed

[patent anvil]

It's an error from cloudflare

[celest gull]

what is cloudflare

[patent anvil]

They are a very large CDN provider - though they do much more than that... that means they offer hosting at edge locations so you can more easily reach different sites/services, much of the internet routes through it. Search it up for some more information.

[sudden delta]

Browser so far
When I got full hardware acceleration turned on, I went to one of the chats that is very laggy and it wasn't laggy anymore.

[hollow radish]

Anyone know how to get support to reply to a ticket? I’ve been locked out of my account due to a bug almost two weeks ago, and after the initial contact with support, I got no update. Any help will be appreciated.

[unkempt pulsar]

Is codex 5.3 spark available for business plans?

[simple heron]

Rn did anyone got access for gpt-5.3-instant. via api?

[vivid heart]

nope

[mighty moat]

I think that is currently the gpt-5.3-chat-latest model, that can be finetuned via API. I'm not sure if you can finetune it to be "instant"-like

[vivid heart]

that's right

[simple heron]

Just checked, they haven't updated this model in the API documentation.

[mighty moat]

yeah it would be definitely be good to get the pricing clearly visible. but in the meantime the cost of my single chat was less that 0.01$ so the token prices can't be anything unusual

[dense storm]

Any form sweden here?

[vivid heart]

"GPT-5.3 Chat points to the GPT-5.3 Instant snapshot currently used in ChatGPT. We recommend GPT-5.2 for API usage, but feel free to use this GPT-5.3 Chat model to test our latest improvements for chat use cases." https://developers.openai.com/api/docs/models/gpt-5.3-chat-latest isn't this what u looking for?

[simple heron]

Thanks

[manic wren]

The android app, after updating, still only shows 5.2.

Not urgent for me, I jsut noticed it after seeing the announcement for 5.4 sooner than you think, and seeing that 5.3 is meant to be there for everyone. I suspect it's not true for UK.

[sudden sparrow]

Why do I need parental consent to log in with Microsoft?

But not when I try to log in with google

Where I have parental controls only on google

And nothing on Microsoft

That's for login in the app

In website I can log in normally

[neon fossil]

Are you under 18

[forest hawk]

Have any of you guys experienced when ChatGPT does a loop behavior? I give instructions but then it keeps doing the same over and over even though you tell the chat to do something else?

[sudden sparrow]

Does that change anything?

[neon fossil]

Well if it's asking you for parental consent, then that means microsoft thinks you're under 18

[sudden sparrow]

When I input my Hotmail it opens browser and Google asks for parental consent, not Microsoft

@neon fossil

[austere snow]

Idk why but but model 5.3 and 5.4 isn't available anyone know why?

[mighty moat]

I know why but you may not like the answer ⌚

[austere snow]

What is it?

[queen heron]

Does anyone know how to have multiple codex windows open in vscode, using the extension?

[languid umbra]

gpt-5.3 is a codex native model

[austere snow]

Meaning?

[languid umbra]

the model itself is called gpt-5.3-codex because it can only be used on codex, 5.3 is a coding model, designed specifically for coding environments (codex app)

[austere snow]

Like the normal app on the Microsoft store wont have it?

[languid umbra]

chatgpt wont have it

the model isnt called gpt-5.3

its called gpt-5.3-codex

only for the codex app

(which is mostly for coding)

[austere snow]

So it's a separate app?

[languid umbra]

well

not really...?

chatgpt owns codex

https://chatgpt.com/codex

[austere snow]

Is it this in the microsoft store app of it?

[coral shard]

Hey everyone, I'm Marcos. Just joined the server and I'm a bit lost. Could anyone shoot me a DM to explain how the Lounges and Information channels work? Thanks a lot!

[zinc wigeon]

hello, not able to export my data in old sora getting a failed error message

[vast gazelle]

Anyone know whats wrong with the edit button

[wintry parcel]

Hi, there. Welcome to the server! Information are read-only channels for server info and announcements. Lounges are themed channels where members can chat with each other and exchange information, etc.

[high hull]

not sure what's going on. On a Business acount and it's dropped more than 60% despite me completely turning off my openclaw setup - which is the only thing I use primarily it for at the moment. I saw that it was reset yesterday but that did not occur for me (I used openclaw the past few days). Is this just a reporting error?

[heady nova]

why would he remove you and what were you working on?

[green summit]

@tawny jewel so.. are you ever going to check messages, or? 🤣

[tight shoal]

howdy, I am using vscode extension, how do I assign permissions to codex so it can run anything he likes in current project? I keep getting this kind of prompt any action I ask

[hollow radish]

Hey! Anyone knows anyone working at OpenAI? I have a ticket open for almost 4 weeks now. I can't log in to my account, and I am not getting any updates on the ticket. I don't know what else to do. Any help will be appreciated.

[mighty moat]

Then switch to Claude. It's not like they had any fundamental difference.

This is the reason why you need to have 2 backups of everything.

[plain bane]

How to bypass the 90-min “Extended Thinking” timeout for massive tasks? (Plus: Pro vs. Deep Research)

Hey everyone, I’ve run into a bit of a wall with ChatGPT Pro 5.2’s Extended Thinking mode.

The Timeout Issue: I’m currently using it for a massive long-form translation project. I have a custom instruction set to “always prioritize outputting as a ZIP file.” However, I keep getting hit with a system-forced termination message: “The system forced an end to the tool interaction… failed to execute the final ZIP command.” It seems like even with the 90-min thinking limit, the session times out before it can package the results. Has anyone found a workaround to keep the tool interaction alive or prevent the system from cutting it off mid-zip?

Pro Search vs. Deep Research: Since Pro already has web browsing capabilities, I’m trying to decide if it’s worth switching specifically to Deep Research for comprehensive market/academic deep dives. For those who have used both, which one provides better synthesis and source reliability for actual research?

Appreciate any tips on the workflow!

[mighty moat]

What type of file is the translated file?

[plain bane]

Latex files. I want to translate my papers into Chinese.

[plain bane]

Latex files. I want to translate my papers into Chinese.

[mighty moat]

I think the problem is the file type. You need to make a work-around.

[plain plank]

guys i got some billing errors where can i contact support (a real human not chatbot)

[charred parcel]

the only way to reach support is at https://help.openai.com - use the chatbot and ask to be escalated to a human

[plain plank]

thanks

[rough geode]

Anyone know why they are retiring 5.1?

[hollow radish]

Good luck with that… but yeah, that’s the only way. I got it escalated to a "human" who replied to me on day 1, and now it’s week 3 and still nothing from him. Meanwhile, they keep charging my cc and I can't log in. If you have better luck, I would appreciate your help.

[austere snow]

^

[maiden folio]

Does DMing @burnt galleon work for anyone else? It responded to me like modmail has DMs disabled.

[wintry parcel]

Hi, there. It seems to be working for me. Did you use the modmail app? #safety-n-help

[glacial beacon]

i just randomly lost all of my GPTs is anyone else having this problem?

i still have all the conversation threads yet none of the gpt profiles i created are there, its like they are just gone, i tried different devices, i tried logging out and logging in, clearing cache, yet nothing is working, in a bit of a panic mode as i do all my business related work through my custom gpts?

[maiden folio]

I did. And the DM popped up a message that it couldn’t be sent because we don’t share a server (which clearly we do here heh)

I ended up going to OpenAI support and using the chatbot and telling it I needed a human to assist. So it says it escalated it. We’ll see how it goes!

[tulip panther]

Anyone else having issues with the format of numbers in the Canvas PDF export from ChatGPT? It keeps getting spread out for no matter what I try

[wintry parcel]

Ty. I'll relay your experience to the Mods.

[wintry parcel]

I'm glad you did that! Server staff are volunteer community members and can't handle OAI customer support. Hope you hear from them soon. 🤞

[vale minnow]

I've been fighting for two weeks to recover my stolen account (my email has already been changed), where I paid for a subscription for a year. I went further than the bot to the manager a couple of times, but after a couple of messages they disappeared (there were no unsubscriptions or refusals, just silence). Is there any hope at all?

[timber shadow]

Hey! If possible, could you share a screenshot of what you see when you try to DM @burnt galleon? If the screenshot has part of your inquiry in it, just make sure no personal info is in it when you share here Like SAWA_SAWA mentioned, if you're looking for OpenAI support, the chat button on help.openai.com is best, but we can help with anything related to server moderation!

[timber shadow]

You may need to check settings to ensure that you can receive messages from the Modmail bot - for this server, try checking Privacy Settings > Direct Messages, and see if that resolves the error with trying to message Modmail.

[mossy valley]

I'm setting up ChatGPT for use with within my organization. We are on the Team Plan and I've sent out invitations, however when users click the emails to join, they get into the "Free" mode and do not see the Workspace. Additionally, under Workspace Settings, I do not see these users having accepted the invitations under Members. They remain under Pending Invites. Anyone have any clue? I've already revoked some users and resent their invitations to the same outcome. Genuinely flummoxed.

Meanwhile we've done this on a competing platform and onboarded users quickly and without issue. Anyone else seeing issues?

[coarse moss]

Many of my chat history are gone after the 5.3 update. Some are just 30 minutes to 3-4 hours old. Can't find it.

[candid kiln]

Same here but it hapoened before 5.3 already

And more above is another user with same experience

[onyx kraken]

This could be a problem that you need OpenAI tech support to help fix, https://help.openai.com/en/articles/6614161-how-can-i-contact-support

I personally haven't used Business memberships, but the help files discuss a personal and a business workspace selection.

In case that is what your organization is encountering, this might be useful, https://help.openai.com/en/articles/8542216-workspace-management-for-chatgpt-business to see how to select between the two.

I want to strongly caution, merging personal and business workspaces is permanent and would mean the person who did that, if they ever stop being part of your ChatGPT business for any reason, 'loses' all their custom GPTs if they made any and all their chats. That stays with the business, if the personal workspace is chosen to be merged into business.

Easy to avoid that, just don't merge them.

[surreal basin]

Not able to change email

Nvm it worked. If someone else faces this issue, try a different email address that doesn't already have an account.

[gilded vessel]

I created my Sora 2 account through the app using my cell phone number. I even managed to generate some videos, but I can't generate any more. The website says I need to verify my number, but it doesn't work. Can you help me? My username is @fpm1928101997

[young hazel]

Hey all, new here, im trying to upgrade from chatgpt go plan to chatgpt plus plan ? This is error message I've getting continously from their pricing page. Can someone helpme out and unblock ?

[ocean adder]

Hello, I would like to share some feedback regarding the ChatGPT data export feature.

I have been using ChatGPT extensively for a long time, and I assume that my account contains a large amount of data, including conversation logs and uploaded files. Because of this, I understand that preparing the export might take some time, especially if the data size reaches several gigabytes.

However, the reason I requested the export is not just to download the data. My goal is to analyze my past conversations and behavior patterns. For example, I would like to examine things such as my thinking patterns, the topics I tend to discuss, and even timing patterns (for example, when certain topics tend to appear). I am considering using this data with other tools such as external chat systems, development environments like VS Code, or other AI platforms for personal analysis.

Because of that, I started to wonder whether the export request is still processing or if something might have gone wrong. The download link seems to be taking quite a long time to arrive.

I completely understand that processing large exports can take time, and I am willing to wait. However, it would be very helpful if there were some form of progress indication or status update instead of simply waiting for the email.

For example, even a simple status like “processing”, “queued”, or an estimated progress indicator would help users feel more confident that the export is still being prepared.

This is simply personal feedback from a user who relies heavily on the platform. Thank you for your work and for providing the export feature.

[patent anvil]

According to ChatGPT's support, it can take between 24-48 hours to complete an export of data. It should include json format for all conversations, which include references to the uploaded files (also included) and responses/generations, as well as voice recordings from voice chats. The JSON data should be able to be analyzed by other models as chat history to identify the things you've mentioned. The only one I'm not completely sure about is timing patterns... as at least the example export that I see online doesn't have timestamps that might be needed for that in each question/response.

[mossy valley]

Thank you - I've definitely done all that. Made sure I've selected the appropriate workspace, have been able to send invites to 9 other users - one of whom accepted, but none of the other 8 have been able to onboard successfully even after receiving the email invite and clicking on the link provided therein. It even welcomes the user to the named workspace, but when they open up into ChatGPT, no Workspace available - they are logged into a personal account only.

This is the case for users who have not associated any personal accounts with their associated email accounts through the organization.

I've tried the Support AI chat - which isn't helpful, and has said it will be escalated to live support who will reply 'in a few days.' I've also called the 888-GPT-0090 number, which couldn't find send me a verification email because it thinks C and D and T are the same letter when spelling the email address out verbally. I've also emailed their support, who thus far, have informed me that:

1. they understand the frustration
2. if we double our seats, or add 20 or more seats before the next billing cycle, the system will pause to true up
3. they suggest a phased roll out of large groups of users, invite 20 a day

I just want to add 8 more to the 2 I have. All this while waiting hours for a reply on email which may or may not be (as in my case) relevant. Missing human support already...

Craziest thing is, we onboarded all 10 users into Claude without issue, same way I had set up / attempted in ChatGPT / OpenAI

We've been using Claude for several days since onboarding, while ChatGPT is struggling to even get off the ground.

Really really really frustrating.

[proven hemlock]

how do I make "suspicious activity detected" I've tried everything, changing passwords, enabling 2fa, literally everything

[patent anvil]

Are you saying you've received that message? If so, remember to rotate your API keys too, if you have any.

[proven hemlock]

No api keys, it literally strips me of any choice of models

[plush blaze]

Hi does ChatGPT has something to edit word documents way better instead of doing weird python extracts

[peak trellis]

with new release just now
whats better for coding? codex 5.3 or gpt 5.4? bc vsc plugin is showing either as choices

[plush blaze]

codex prob for efficiency 5.4 is the smarter one

[wintry parcel]

Hi, there. Server staff are mostly community volunteers and can't help with account issues. Have you tried logging out and back in? If the problem continues, please contact OAI customer support: https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[wintry parcel]

Hi, there. This is what I usually do: 1) upload the word doc to edit, 2) ask ChatGPT to edit it in Canvas so I can see and approve the changes made, 3) once editing is done, ask to provide it as a downloadable .docx file.

[vernal prism]

Earlier version of chat gpt were really good in generating Microsoft word document or excel sheet but the latest version fail. Am I the only one facing this issue?

[silver robin]

hey guys are you able to tell me more about the advertising through ChatGPT

[livid cape]

Hi people, whats the best channel to ask for LLM recommendations? is it allowed to ask for local LLM recommendations in this discord?

[timber shadow]

Hey! #ai-discussions is your best bet for discussions regarding AI in general. If you're interested in OpenAI's local LLMs, you can check out #gpt-oss as well

[livid cape]

thx

[dawn echo]

why is 5.4's context window 258k tokens instead of 1m in the codex windows app?

[wicked remnant]

Is anyone else having difficulty downloading PDFs and word documents of Deep Research reports? It's not downloading them on either my phone app or the website

[patent anvil]

Context recall benchmarks (MRCR v2) seemed to drop off for longer contexts quite a bit, so it's probably for the best.

[eternal sluice]

Why is this new version so slow? And how do I change the model, I want my old 5.3 back (I’m a free user)

[silver ruin]

I’ve been using the image generation features for quite a while now. Both Sora and GPT can generate images.
In the past, GPT used repaint mode rather than pixel-editing mode. But now GPT uses pixel-editing mode, which makes it unable to do many simple tasks—like adding snow to an image, for example.
I was really happy when I found out that Sora still uses repaint mode.
However, you mentioned that Sora 1 will be discontinued soon, and there are only a few days left.
is there an API for image generation with repaint mode, or where else I can find this functionality?

[foggy vortex]

What would be my best tool and approach to create 12 sound effects for a cards game i am making? chatgpt and a single prompt with detailed instructions on each sound?

[mossy valley]

Has anybody gotten anywhere with live / human support?

[analog bridge]

i just paid the go plan and the payment went through but in my chat gpt doesnt show that i have a GO plan, its still the FREE version, i tried support but it says they will email me in a few days, help?

Im from mexico

[mossy valley]

This may be related to the issue we're experiencing.

Thus far, I have received 2 emailed replies, neither of which were particularly useful or on topic.

For you, I would say, try a new browser ... or clear your cache and log back on

see if it refreshes into the paid account

[analog bridge]

really disappointed honestly, im from mexico and this few weeks i have been using chsatgpt from work to help, just today i decied to buy the GO plan, they charge the bill to my bank app but it doesnt show up in my chatgpt, and so far it seems that there is no really a customer service not even in english (i tried looking for spanish but there is none) sorry my english if there is anything wrong

[mossy valley]

No problem about your English. The company is a US company, and things are still ... early stages it seems. I would assume their support staff are being overwhelmed, but this basic functionality which ties revenue with their product and growing user base, is disappointing.

We are dipping our toes into ChatGPT for our organization, and I am trying to get 10 users into the platform with a shared Organization / Workspace. Thus far, I and one other person have been successful. Everyone else has simply gone into the "Free" mode, despite OpenAI already having charged us for all users.

I will also say though, we are evaluating Claude in conjunction, and it really is a completely opposite experience (as far as enrolling users into the same workspace).

And if it makes you feel better, your English is better than my Espanol.

[analog bridge]

Yes, i undestand maybe that they are overwhelmed by many users, but i think that they really should have a costumer service for users that pay upgrade plans any company that have a subscription option has that, i think is simple as having a custumer service ready to answer complaint no troubles about subscriptions

[mossy valley]

I absolutely agree with you. I have been here trying to get some collective knowledge and experience in finding an answer to my paid-subscription problem too, due to the ineffectiveness of:

1. the AI Support Bot
2. the available OpenAI discussion forum
3. Live support

[analog bridge]

Its really worth the Pro plan? i been using chat gpt to help with my work (im a lawyer) and so far its been really helpfull but i cant send more docs right now until i upgrade my plan

so thats really what its stressing me out

[analog bridge]

the AI Support Bot just stop answering me haha, last message was "Escalated to a support specialist; You can expect a response in the coming days. Replies will also be sent via email. You can add additional comments to this conversation if needed." jesus christ

[mossy valley]

This is what happened to me too

In the end, I had to email their support. I had even tried calling their phone support, and it was pretty ineffective.

[mossy valley]

Honestly, because I am having difficulty getting my other users into ChatGPT paid plan, I am not sure if we will continue to proceed with ChatGPT's platform. Support is a part of what you are paying for, and frankly, this seems not ready for business use.

[analog bridge]

whats the email? maybe i will try that

[mossy valley]

[e-mail]

[analog bridge]

Yes, its feels like they are not ready for business use, just focusing on growing as tech i guess, but the business part is not ready to go big

[mossy valley]

Do not expect quick responses - but I would say, be very clear about what the issue is, because I am also on day 2 of getting absolutely nowhere with their emailed support.

[mossy valley]

Could be the story with new technology ... it's great when it works, if it works 🙂

[analog bridge]

Can i ask what is your company or business about?

[mossy valley]

We are an investment management company

So, it's sort of in OpenAI's best interests for us to gain a positive perspective of their platform...

[zealous cairn]

Need Sora Invite code

[karmic vale]

For implementing a plan, should I use gpt 5.4 or still gpt 5.3 codex? Whats your guys experience

[languid umbra]

i keep having this error where it rejects the code and does
"The first patch failed on Windows path length handling inside the patch tool. I’m splitting the rewrite into smaller relative-file patches so the redesign still lands cleanly."

it just drains my tokens

[fast owl]

For planning I'd say 5.4 since it has more knowledge in general areas, and to implement it use 5.3 codex- 5.4 if you don't have much detail in the prompt and use 5.3 codex if you have a detailed design plan and implementation plan

[split nexus]

Any best ai maths solver?

[patent anvil]

wolfram alpha 🙂

No point wasting cycles, use the right tool for the job

[kindred urchin]

I have a problem that keeps continuously happening Basically, anytime I send a message on a new account or any other account even create a new chat it just constantly beeps it’s like circular circle like it’s about to answer, but it won’t and I left it even there just for an hour and it still won’t reply. I’ve disconnected my network downloaded the app and downloaded it again. Tried using the browser version tried to refreshing my cookies. Tried hard resetting all of the above. Do you know what could be happening also sorry if this isn’t the right channel to report that

[scarlet sand]

Why was Xyacle banned?

[green summit]

Nobody knows who that is xD.

I just checked their message history and they probably just left after those 4 total messages they've sent here about Atlas. besides, I also checked your history and you did say they blocked you a while back, so I don't really know why you would want to talk to them.

(No need to explain, I don't really care, sorry)

But in case they've changed their username in the last month or so and you tried looking for them but couldn't find it:

@drowsy thicket (they left, but that's their profile)

[If not. Don't mind that last part]

[rain fable]

were you find atlas? in swiss we dont hace 🙁

[green summit]

https://chatgpt.com/atlas

[rain fable]

thnak you but i dont use mac 🙂

can i make a emulator?

[green summit]

Ah, I see. Yeah it should be available all around the world, but only for MacOS.

And I'm not sure, there might be a way, but someone else is probably better in helping you with that.

[rain fable]

hehe ohh you want help 🙂 haha its better 🙂 heheheh who there you hahah ok master help me 😉

[light onyx]

my token expired and i cant log back in

[compact sable]

anyone can help me?

[green summit]

For browser:

Open browser > settings > find "site settings" > then "all sites" > search for "OpenAi", "GPT", and "Sora" (if it's for Sora. Not in this case ofc, but for when) > delete data > then reopen the browser > sign in/check again.

(Don't worry, this will not affect your account data)

[trim sun]

Will it be done in 2-3 days? What do you guys think

[trim sun]

[light onyx]

[ancient herald]

why doesn't the github app work with ChatGPT 5.4 Pro? It works with GPT 5.4 Thinking on the website but not pro

if you try to attach the GitHub app to a new conversation with Pro and ask it to view your private repositories it'll say it doesn't have access to it... while if you do it with the Thinking model it works

[frigid owl]

Yo guys what ai model in general is the best one ?

[midnight hearth]

Claude IMO.

[night breach]

5.4

Without a doubt.

[midnight hearth]

Says the loyalist fan unwilling to try anything else.

[night breach]

Who said I haven't tried anything else? I've tried all of them.

And I do frequently use the free versions of them for specific use cases. But I prefer ChatGPT.

For example, sometimes, I need to make a solid good image for a diagram. I can't really rely on Sora or ChatGPT. So I use Nano Banana 2.

And I do see the benefits of Claude. I have used Sonnet 4.6 plenty of times for various guideline purposes. But I just prefer ChatGPT because I've already customized it to peak performance.

And I'm still not exactly sure if it's me and my specific ChatGPT's custom instructions and memories that is giving good results or the overall system itself. That's why I promote it. Because I'm constantly seeing a lot of high potential output responses.

[frank vortex]

I really want to like 5.4 because 5.3 and 5.2 and I were at constant odds. But I just can't get aligned. Any tips to make 5.4 more like 4o? INB4 "you want it to glaze you / reinforce you" no, my custom instructions are to challenge me in an appropriate way / sharpen thinking / explictly do not output praise ever.

Example: my dog ran away yesterday. had it pull numbers of the odds i get her back. i text-to-voiced in an emotional state a summary of the data. 4o would have probably said "More or less. What are you thinking right now?" 5.4 broke down every possible interpretation of why my summary was wrong (it wasn't) and ended with "That's the trufhful version".

Question is: Any tips to make 5.4 more like 4o?

[night breach]

I like 'INB4', I've never heard of this term before and I liked the example that you gave with it.

Your best bet would be to to first use 5.4 with extended thinking mode to say: "Look up, and formulate a causal map of everything that made 4o unique, and turn it into a standing directive for AI. Take your time to try and piece everything together as meticulously as you can. I want it complete, not rushed."

Then once it gives you the map, you go here and use the GPTs in here: #1473050350318653560

First, you paste the causal map in the Memory Configurator GPT, but just before the pasted text, say: "Make sure to formulate the memory to be as exhaustive as possible to thoroughly address all capabilities as a single continuous paragraph:"

Then follow the rest of the GPT's copy and paste instructions in its output for maximum efficiency.

Secondly, do the same with Custom Instructions Configurator, paste the causal map in there. But for this GPT, you don't need to type anything else in the GPT, because I've made sure that it's optimized to exact character length.

And that's it, you follow the instructions: copy and paste, and you're done.

Let me know how it goes!

[frank vortex]

Thank you for taking the time to write a detailed reply. I’ll follow this and after I run enough sessions to call it data I’ll chat back with results!

[vivid heart]

how to fix this?

[gilded lava]

Is there a way to disable the Instant Voice feature? I often use ChatGPT while I'm driving and I cannot type. Unfortunately, even if I select Thinking 5.4 as soon as I use the microphone, the app reverts to Instant Voice 4.o. All the answers I get are sloppy, incomplete, and frequently just outright wrong. Thank you for your help. Cristina

[zealous sand]

Quick Question - Why can’t I use ChatGPT Agent Mode? When I click Cloud Browser (which used to be just “Browser”), nothing happens and it doesn’t work at all.

[sudden delta]

yall we really need to fix the image uploading issues

it fails in the same way every time

first 2 attempts to upload a image says failed, check network settings
I have to refresh exactly 2 or 3 times and then it lets me upload ONE.
then fails again.

[charred parcel]

I just tested it, images are uploading fine

[sudden delta]

yeah but its only happened to me
of course when i actually want to capture the error now its suddenly not happening-
Ill have to send it if it happens again

[sudden delta]

there it is

So i refresh the page
and then boom it works

then i ask gpt the question after it works, upload a new image, fails again and now the only way around is to keep refreshing a laggy chat.

[charred parcel]

you have a network problem

[sudden delta]

i dont though thats the thing, i really dont. promise. ive checked.
considering that it goes away if i simply refresh the page, i dont
and ive checked everything

[charred parcel]

something on your computer or your network is affecting it's access

[sudden delta]

again, though this isnt the actual case

the problem is the error is false. i dont actually have anything causing it. if i did. i wouldn't be here saying why am i getting this

everything onmy pc is free to use and elevated and all that theres no access restriction.

[charred parcel]

ok, well uploading images is not a widespread problem, if it was there would be a status update about it. The problem is most likely on your end. You can contact support at https://help.openai.com for more direct assistance

[sudden delta]

it appears to be super intermittent

so im fine

though the old gremlin has returned where its becoming so absurdly laggy on anything but mobile
Mobile is never laggy even if the chat is 500 prompts long but desktop becomes laggy after like ... 10?

even with gpu acceleration which i thought fixed it. nope.

[sudden delta]

im coming back to this becuase it has managed to UNFIX itself

now even with full gpu accel with the most powerful amd gpu ever to grace the planet it lags at only 10 messages when mobile literally never lags even if its 9999 messages long

idk what to do this is so broken bruh
all this new innovation from OAI and we cant find step 1: the lag

[night breach]

@sudden delta Try explaining your problem to ChatGPT 5.4

With extended thinking.

[sudden delta]

why would i come to community help just to hear "ask chatgpt" thats not the most helpful idea obv ive already tried.

[night breach]

Because at this point, 5.4 with extended thinking mode has become so powerful, that if it can't help you, then you can pretty much guarantee no community can neither.

[midnight hearth]

Yeah right, that's a massive stretch.

[sudden delta]

thats not exactly true at all

even gpt5.4 still has accuracy rates that go far below actual experts

it will likely be until somewhere around gpt 7 to 9 or 10 before it actually outranks real human beings

there are many things gpt5.4 also just still gets horribly wrong even with thinking or it cant open pdfs online and get sources

[night breach]

Maybe yours is just really bad. But that's alright.

[sudden delta]

thats not how ai works

we're all using the same gpt5.4 here

i never said gpt5.4 is "bad" i said that its not able to replace humanity

even somebody else agreed that is a MASSIVE stretch

[night breach]

There's also people who come in here on a daily basis, explaining why ChatGPT is good and bad with multiple different interpretations. It just depends on how you use it. I don't have much complaints about it. The way I use it seems to be working out perfectly for me. And I try to understand why people have different problems with theirs. If you don't think it's good, that's fine. I don't know why, because it seems to be good for me. So all I can do is give my advice and guidance. You can take it or leave it. But don't think I don't know how AI works and you know better than me. I have no problems with it, so clearly there's more problems for you than there is for me.

[sudden delta]

well maybe i would say the standard for "Replacing humans" is much higher than how you use it
what if the question you have bout a game or something is something only people can reliably pull

or what about a history question where its hidden in archive books and pdfs that the ai cant open

but the REAL issue is you trying to just respond in #community-help with "ask chatgpt"
thats not how community help works

[night breach]

Sometimes the simplest answer is the best one. I don't know if you've tried or not. Most people who come here asking questions don't even think about first asking ChatGPT about it. You'd be surprised.

And that IS how a community works. Someone may give you advice personally, another could remind you to use ChatGPT. We're here doing the best we can to give you support in ways you may not immediately remember or realize.

Here's an OpenAI discord sticker.

Have a great day!

[vague eagle]

for me it lags in the app and browser so.. just suxx

[lofty drum]

OpenAI why did you let the US buy your company and use it to kill thousands of innocent civilians all over the world. Long live Israel

[midnight hearth]

I was with you until the last bit, ngl.

[lofty drum]

sarcasm

f#ck the big yahu.

openai lost 30% of there user base in 1 day.

Long live freedom

[midnight hearth]

Sorry, I wasn't sure given how many wild takes are going around. Hell yeah!

[frank vortex]

Hey all. Connectors issue:

Outlook Email app says connected but doesn't show in the toggleable connectors. Chat told me it is a tool not an app so it isnt there but it can still search it. of course, it then said it isn't connected so it can't do it. but my UI shows connected.

I disconnected and reconnected it. opened a new window, closed everything, PC restart. nothing has made the UI show up for me to toggle it on in-chat.

am i doing something wrong or just a bug?

[sudden delta]

Yeah idk what it is

Its like the app and website on PCs is just unoptimized

Like..

One of those games that looks really good, lags to 20 FPS even though you have a 5090

[cinder peak]

lol this morn itell it to do something it goes "thinking" for like a minute, then ends ,and i'm like "are you there?" and it thinks

but never answers

[thorn crescent]

https://community.openai.com/t/mcperror-mcp-error-32601-method-not-found/1376110

Hi guys, I have a problem when using MCPServerStreamableHttp it gives me the error

2014 |     CreateTaskResultSchema
2015 | ]);
2016 | export class McpError extends Error {
2017 |     constructor(code, message, data) {
2018 |         super(`MCP error ${code}: ${message}`);
               ^
McpError: MCP error -32601: Method not found```

[hollow radish]

Does anyone know what to do when you have an issue logging in into your account and support doesn't reply to tickets for 25 days now... I really don't know what else to do or who to contact.

[trail cairn]

what error do you get?

[thorn crescent]

McpError: MCP error -32601: Method not found

[trail cairn]

i meant the person trying to log in but your mcp error looks like you are using the wrong url or method name

[sonic escarp]

I submitted my data request OVER A WEEK ago and no answer what is going on?

[trail cairn]

the one in settings is automated and often fails

[sonic escarp]

i submitted it 3x

this is not cool

i call BS

[trail cairn]

which plan you have? free?

[thorn crescent]

That's right, I'm not sure about the next configuration step; the examples in @unreal hemlock/agent-cores don't have them either.

[sonic escarp]

i pay 20 usdc a month

this time it didnt even let me click in old sora

not to mention will it actually download my art

this is handled really poorly

[hollow radish]

My account was connected to Google. I decided to switch the email to a non-Google email under settings. After doing that, I was not able to ever log in again because when I try signing in, I get the error 'you tried signing in as 'xxx' using a password, which is not the authentication method you used during sign-up“. The problem is that there is no way to revert to the Google email. I got an email immediately after changing emails saying that the email was changed and that I could revert by clicking a link in the message, but even that didn't work. Now it’s day 25 without access.

[trail cairn]

changing login method from that is not supported. create a new help chat thread unless you already did and ask it to be escalated to a human

[hollow radish]

Yeah... the saddest part is that literally all I did was use the standard 'change email' option under My Account. If it was going to break everything, the system really should have just warned me. I’ve tried escalating to a human so many times. This all started back on February 15th... today is March 9th. They actually asked for HAR files and details, which I sent like 20 days ago, but since then it's just been total silence. Honestly, at this point, unless by some miracle someone who actually works there sees this, I've kind of lost hope that anyone will ever pay attention to it. It’s just exhausting.

[trail cairn]

yea the help chat bubble has had bad rep for years and only solution is to keep restarting if it goes stale. the email change option was added not too long ago and it has caused some other probs too from what ive seen

[keen arrow]

Hey, that's a classic JSON-RPC -32601—server sees the request but the method's missing (likely config gap in capabilities or deprecation in streamable HTTP). Check if "prompts" or "resources/list" is exposed.
Built a layer that proofs configs/account changes like this (Badger node for immutable receipts) and scans gaps (LeakGuard node)—glides past without waiting on support. Check echoverse.dev/badger or /leakguard if curious (free proof takes 30 secs). Let me know what it shows!

[hollow radish]

Yeah, it’s wild. Honestly, losing access like this has been a massive wake-up call about vendor lock-in. When your work and many aspects of your life dependence on a single platform is this high, and their support is a total black box, you basically become a hostage to their system. I’ve dealt with a lot of enterprise SaaS providers, but I don’t think I’ve ever seen a worse support pipeline than what I'm experiencing right now.

I'm currently having to rebuild all my core workflows from scratch somewhere else because my information is just trapped behind their broken auth.

Honestly, you’re the first actual human who has responded to me in three weeks, so thanks for the ping

[night marten]

Hi. Have you managed to fix the problem? Same issue here...

[mossy valley]

Try it again - it seems to be working after some time Friday

[rich rivet]

why have i all of a sudden lost access to codex through my csu edu account?

when I try logging in it gives me: An error occurred during authentication (codex_cli_workspace_disabled).

[onyx kraken]

We sometimes suggest that because that could be the best idea the person has. Like me, I do not know the answer to your question. It's out of my field, and I'm not sure when or even, sadly, if someone happens to see your question and has ideas. Which means you might go unanswered.... or someone might try and help with suggesting 'ask ChatGPT'. Not to shut you down (that's not what I mean when I encourage that for sure!), but because from the answerer's perspective, maybe the answerer is imagining you not having any human answers that help.

I hear you that model answers aren't helping much/yet either.

Some questions don't have solutions yet. Sometimes iterating with the model to try and find a better way to explain and approach solutions from helps. Personally, I'm glad you explore widely.

[sudden delta]

Well, maybe we could genuinely look into it
It's not just my problem. Many people report.
When the chat becomes lengthy, it becomes extremely laggy on desktop. Unless someone else is reporting that not happening.

It is so laggy to the point where it is unusable.
And it brings my entire computer down with it, despite having literally the second most powerful CPU that AMD has ever created.
And the most recent and most modern GPU that they've ever made.

So.. I don't know
And it's funny because my phone, which is like seven times weaker, never lags, even if the chat is 200 messages long.

[onyx kraken]

I'm open to it being looked into, but you're talking to community members here, similar to yourself. Even the mods are volunteer community members, #server-staff . Only the very few folks who have gold names work for OpenAI.

For me, I have no idea how to look into that issue past 'Yep, aware'. I'm aware of it, it affects my chats, I have worked around it for a very long time, I almost never make 'pretty long chats' anymore.

We can #1053565716500074526 or even #1070006915414900886, and you mentioned already reaching out to tech support.

Hopefully one day it's fixed!

And if you find an allowed-use workaround, super awesome.

[onyx kraken]

Looks like there's a 'Codex unresponsive' issue OpenAI is still working on, #status can help you track that if you want to.

[sudden delta]

Well, the only workaround right now is to use my phone and then continue the chat there.
It totally works fine.
It's just, I enjoy doing it on computer more than my phone.

[unkempt crown]

Can anyone help? im trying to redeem the free plus trial, i tried visa and mastercard and they both say the same

[onyx kraken]

Here's a related help file, https://help.openai.com/en/articles/7232916-why-was-my-credit-card-declined that discusses what types of payment are accepted. Quite a few are not accepted, such as prepaid cards or if the bank's outside the region that OpenAI provides services to, among other reasons.

Note that we're all community members in the Discord, don't provide any personal or financial info to anyone here, help is through that help website, bottom right link if you want to reach out to OpenAI.

Some have stated they had trouble and solved it by paying through their phone instead of a browser, or even using a different browser on their computer

[jaunty crow]

hey ,
chat gpt not working as claude ai why ? when i am writing ai promt in chat gpt feedback format is not good if i ma doing same think in claude ai feedback format is good

[hardy pasture]

Does anyone here have an app live on the chatgpt app store? I'm hoping to learn more about chatgpt app surfacing and how to best optimise an app to surface based on the user's intent, Any help would be greatly appreciated 🙏

[shadow vigil]

is thier any way to get openai api key for free for college project?

[lavish ether]

I use ChatGPT mainly as a research and editorial tool. It has been enormously useful and amplified my productivity more than I would have imagined. That said, a slight issue has risen recently which I have so far been unable to resolve (and the model isn't helpful either).

A day or so ago, I believe in inadvertently hit a combination of key strokes which turned on "Developer Mode" and now I cannot figure out how to turn it back off. So far it has not been a serious problem; and I've even discovered funcitionality that must be specific to "Developer Mode" which is intriguging. But I'm a little concerned about how older Projects and chats might be impacted, and would really like to figure out how to turn it on/off at my discretion.

[wintry parcel]

Hi, there. You can turn on/off Developer Mode by going to Settings > Apps > Advanced Settings.

[lavish ether]

Which part of the browser interface to access the right "Settings" pane? The User widget in the lower left or the ... widget in the upper right?

I ask because: I've already tried this solution (it is what the model recommended) multiple times but my interface still shows "Developer Mode" near the bottom across almost all chats.

when I follow the UI path you describe via my User widget in the lower left this is what I get.

No clear control there to turn Developer Mode off.

[wintry parcel]

Hmmm, that's the path indicated in OAI help center. https://developers.openai.com/api/docs/guides/developer-mode I don't know why the toggle to the right is hidden. If you are on Enterprise or Business plan, there's a different path: https://help.openai.com/en/articles/12584461-developer-mode-and-mcp-apps-in-chatgpt-beta
Have you also tried logging out and back in? If none of this helps, go to customer support https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[lavish ether]

thanks SAWA_SAWA I will look into the info at those links!

[lavish ether]

FYI, as soon as I stopped to examine your detailed responses I knew what to check, as it has been an issue with various browser based tools in the past.

I use Dark Mode, one of the themes in Win 11. Generally speaking this interfaces well with browsers and other applications; but not always. It has been far too long since I did my little bit of coursework in HTML/CSS/XML so I do not recall the details of how such systems worked even back in those days (ca. 2015), and I imagine it has changed quite a bit anyway. So I cannot say why standardized UI elements would sometimes be invisible in Dark Mode, but they are! 😂

Once I saw that toggle button widget in your screenie I knew what to do: hover my cursor in that area of my own control screen and yep, that was it!

[wintry parcel]

Aha. Glad you were able to figure it out.

[coral monolith]

i reached a file limit today but i just uploaded the first one today?

how do i rest

reset?

[vague flint]

glad I'm not the only one this happened to...

it says the limit is 80 files in 3 hours... definitely didn't reach that in the time I was doing stuff lol

[coral monolith]

it cleared itself it seems and is working.

[lavish ether]

yeah paying $20 a month and getting a "you reached file upload limit" not cool . . .

[thin locust]

Hello I was hoping for support on a question? How do you add servers to follow things or for gaming on discord?

[potent horizon]

Getting this as well just now... generating 1st Image of the day

Must be a bug?

[tidal depot]

Is there an issue with uploading files? I’ve done barely anything and it keeps telling me I’m at my limit, I’m a plus user

I just seen others say this but I’ll say it again, so it’s a known issue, recent? Cuz file uploads still work easily on my phone just not on desktop

[coral monolith]

thats what im thiking.

[potent horizon]

Found the fix! Delete your website Cache

Reload the page

it now is allowing me to upload images again

[potent horizon]

delete website cache

[tidal depot]

Works now, just slow, also didn’t let me login for a couple minutes lol

[potent horizon]

Weird, it never logged me out and let me right back in.

[tidal depot]

Reached the limit again after just 2-3 uploads

Actually just 1

Clearing cache again fixed it but it is annoying

[zealous sand]

Unfortunately, not yet

[civic quail]

Getting this error: "You've reached out limit of file uploads. Please try again later."
Never got this before.
I have not uploaded many files in the last day.
I'm on the Plus plan.
Deleted several Projects that had a bunch of files, still getting same error.
Please help.

[civic stratus]

chat gpt is currently telling me "you have reached our limits on upload files - please try again later" but i haven't uploaded almost anything in the last 24 hours

Why is this happening?

I have a plus account.

oh sounds like i need to delete my website cache lmao

[wintry parcel]

They are having file upload issues today. https://discord.com/channels/974519864045756446/1349803488833572874

[civic quail]

Thank you.

[grave magnet]

does anyone know who is the best person to contact if my enterprise's mcp chatgpt app got rejected for commercialising a digial product whereas our competitors' app got approved with the exact same offering

[polar eagle]

Anybody knows the fix

Stuck

[wind lion]

Hi, I’m having some issues with ‘apply-patch’ on Windows and I have prepared a file with the issue and investigation done. How can I send this report to support?

[viscid escarp]

How to make channel like this only access through server guide #server-rules message

[rustic marlin]

you can get to the support team on https://help.openai.com

[wind lion]

Thank you

[toxic pine]

I'm still getting billed, even though my usage is far below these thresholds. Anyone else having this issue?

[hallow garden]

My answers didn’t change from ChatGPT.

Has anyone gotten this update?

[night marten]

I reported it to the support, maybe we'll get a fix soon

[sudden sparrow]

where's the option to log in using microsoft

when I enter my hotmail it wants me to log in using google

[crude mica]

can anybody tell me how can we did prompting with ai because my prompting is very bad😩

[thick lichen]

write a prompt, and ask it to tell you back what you said, then correct anything he understood wrong

[unkempt shuttle]

im using browser chatgpt and it is SPAM flashing me from dark mode to light mode, what do i do?

[cedar cloud]

I am getting messages like this in codex today:

unexpected status 403 Forbidden: {
  "error": {
    "message": null,
    "type": null,
    "code": "invalid_workspace_selected",
    "param": null
  },
  "status": 403
}, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: 1234, request id: abcd

Anyone know what's going on here?

[copper lodge]

is anyone able to login to https://auth.openai.com/ ? when i try to login to the dev portal, this page is just white

[rapid totem]

Go to #1070006915414900886

[sour tusk]

Same thing

[zenith mist]

Hi I have a very sensitive matter I am seeking help with on chat gpt. I got a response from the legal team at OpenAI on 2/20/2026. The matter is somewhat urgent.
I am requesting informal support from someone at openai that handles:
https://openai.com/index/building-more-helpful-chatgpt-experiences-for-everyone/
https://openai.com/index/gpt-5-safe-completions/.

[charred parcel]

you won't get support here, use https://help.openai.com

[zenith mist]

I did, I emailed the legal team and they responded, they offered me the links I sent. I reviewed them. I would like to talk to someone informally about openai I think I found a serious problem with the model.

I would like to talk to someone about it that is familiar with the content in the links I sent or works for openai or both, I want to get another perspective and think through how to formulate the problem clearly.

[delicate yew]

Hi. I have a trouble using Company Knowledge. GPT says it doesn't have an access due to repository being private, and that I should install an Integration GithubApp for this, but it doesn't know the exact name for the app

[viscid escarp]

How to make channel like this only access through server guide #server-rules message

[timid flint]

how to have permission to write in prompt-lab?

[white ingot]

Are there currently problems with the memory function of GPT 5.4? My GPT seems unable to properly access the stored memories.

Edit: In projects, outside of projects it works fine.

[molten plank]

chatgpt is currently unusable. It’s not even capable of interpreting projects, nor does it allow downloading files previously sent for editing. I’ve been dealing with this for about 48 hours now. A complete waste of time if you want to work seriously and professionally. I am working on my own private project, and at the moment chatgpt is giving me answers based on someone else’s project. Completely nonsense 😅

[obtuse sapphire]

help.
1 . I cant delete any chats that are inside a project.

2. i cant connect Slack to any chats inside a project ( but can on those outside )

browser = Chrome latest MBP

[white ingot]

I've moved a chat from a project to an external one. Everything's working again there. I suspect it's really related to the last few days and the migration/shutdown from version 5.1 (I remember it forever).

[white ingot]

I can'read a few chats inside a project atm. it's all bugged.

[obtuse sapphire]

yeah something on the front-end code is going on..

i was able to drag it out via the sidebar though..

[molten plank]

i'll check outside my old project 😉

[molten plank]

I confirm that it is still useless. doing exactly the same. the only thing that is solved is the download part. because the answers are related to another persons project, even chatgtp confirms it 😅

[white ingot]

strange... buggy af atm...

[wraith sierra]

Hi

‪Why does the codex plugin in vscode open a linked file into a browser (which is obviously not there) when the file it created/edited is already inside the current project folder?‬

I’m on windows

[lapis walrus]

When will be sora 2 in Czech republic finally. I'm so bored to use VPN to be able to generate with sora. I want to finally use it on my phone. I am still waiting to today

[white ingot]

I still don't have access to my recent chats. It's so annoying.

[kind hemlock]

Enter-to-send suddenly on...

Any android guys facing this issue?

[inner belfry]

ChatGPT randomly changed the enter key (normally new line) to send on Android and now I can't type any long prompts on mobile????????

And there's no setting to toggle this behaviour?

Imagine if the Discord app did not let you type out multiline messages.

The app is completely useless for me right now.

Nearly all my prompts are long prompts and I can't do anything with it.

Why would they do this? How is this behaviour supposed to be helpful????

---

Has anyone else noticed this problem?

[green summit]

Definitely a bug. Try signing out and back in.

Or

For App:

Go to the homepage on your phone > hold your finger on the GPT app icon > select "App info" > find storage > "clear data" > reopen the app and sign in and try again.

(Don't worry, none of this has an effect on your account data)

[kind hemlock]

That's what I raised in the video

[kind hemlock]

..

[kind hemlock]

It shouldn't happen to two persons at the same time then...

[green summit]

Well I just tested it and it works normally for me. So idk what's happening

[kind hemlock]

Hit chatgpt update option.

Mine just got updated this morning so...

[green summit]

Ah, I see. There's an update, that makes sense. Well, If that's the price, I'm not going to update lol

[kind hemlock]

Hehe

[green summit]

I do hope it's a bug then and not something permanently. That would suck, big time..

[kind hemlock]

This is a super big blunder. The guys behind App UI made it.

[kind hemlock]

Big bug.

[green summit]

I already know how frustrated ima get.

[kind hemlock]

It's a small thing causing big problems.

[green summit]

🥲 OAI in a nutshell once again

[kind hemlock]

Imagine ms word disables enter for next line

Masssive disaster

[green summit]

🙃

[kind hemlock]

Idk why no one highlighted it again

[mint ibex]

How can we use the real version 5.1 of Chagpt like before?

[brazen bluff]

Guys i want to use gpt 1.0 real bad

[crystal steppe]

Hi there

I have Chat plus and I'm noticing that when i ask it to write code for me, I tell it that I'm not good and copy / pasting to add code myself and have told it multiple times to re write the whole code, but it seems like I have to tell it every time, is there a way to force full re writes every time?

[timber shadow]

Hey! What editor are you using to code? VS Code? And, have you tried Codex yet? I am in a similar situation as you, not someone who really knows how to code very well, and having worked with both ChatGPT and Codex to work on various projects, I can say that using Codex is a ton easier. Your Plus subscription will allow you to use Codex

[white ingot]

I still use vs-code

[timber shadow]

Same here, I've found the Codex extension for VS Code to work great, but lately I've just been using the Codex Windows app instead and have been loving it, haven't switched back to VS Code yet.

[crystal steppe]

I haven't checked out codex, should I? I just use the regular chat

[timber shadow]

I would definitely recommend it. Codex is like ChatGPT, but specifically made for coding. It knows how to work with all your project's files/code automatically, making edits to only the needed parts, allowing you to undo if something breaks, etc. This is the help article I'd recommend to get started: https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan. Depending on how you work on coding projects, I'd recommend either trying the Codex IDE extension or the Codex app - see the links for both in the "How to connect Codex with your ChatGPT account" section of the article ^

[crystal steppe]

I will check it out thank you!! 😃

[timber shadow]

You are welcome! Come back with any questions you might have

[kind gazelle]

Good morning/evening/night, everyone! Is anyone having auth problems?

[mint ibex]

?

[chrome elm]

anyone else have issues on web with loading old chats? they'll all show up for a moment (plus projects) and then they're not there anymore. not gone from the looks of it, just seems like a UI problem or something

edit: It's good now lole disregard

[sudden sparrow]

why I can't log in with my microsoft account

why when I give my microsoft email it wants me to log in with google??

[surreal mango]

Where’s gpt 1.0?

[surreal mango]

Sometimes it’s easier to do it that way?

[sudden sparrow]

but it's a microsoft email

why it wants me to log in with google

[surreal mango]

Are you using the app or the website?

[sudden sparrow]

it happens on both

[surreal mango]

Sometimes that makes a difference. Maybe try switching devices I’m not sure sorry.

[vestal berry]

it was fine for me 30 mins ago but now its the same for me

idk whats happening

[kind hemlock]

Use VPN?

[vestal berry]

nvm its fine again, i hard refreshed the page

[kind hemlock]

Kk

[verbal pagoda]

Does any one know whats the best AI model for making a multiple choice quiz using a lot of info (large pdf document)?

[vague cave]

you just need a decent harness and prompt, i'd bet even small local models like qwen can handle that just fine

so any of the openai ones would too

[deep bison]

Sorry for dropping this in 3 channels, not sure which is the right one

[autumn thistle]

hey i want to get the free month of chatgpt + but evreytime my card gets declined , although it works just fine on other platforms , does anyone knows how can i fix this problem ? thank you

[sudden sparrow]

Bruh I wanna log in with my Microsoft account but when I give the email it wants me to log in to google where that account isn't even associated with google

What did you guys do

Where's the option to log in using Microsoft

[sudden delta]

so i mentioned this a while back and ive all but eliminated the issue

the issue was edge and chrome and actually literally all chromium browsers have very poor perf on my gpu and card, mozilla firefox apparently works great and has effectively no lag even with a megachat

just to let yall know

[queen heron]

Is anyone using codex within Xcode?

I'm having issues logging into Codex on Xcode and was wondering if anyone can help me out?

[charred parcel]

what version of Xcode are you using?

[queen heron]

Version 26.3 (17C529)

[charred parcel]

you can't see where to Chat with ChatGPT or Claude?

[queen heron]

Im facing this issue

[charred parcel]

I have tried and can't reproduce this issue, for me it logs me in just fine.

[queen heron]

I see, thank you for your help

[vague grove]

Bumping this. Does anyone have any details when ChatGPT Health could be available in the UK?

I'm on a Plus plan on Mac and I was super excited when it was announced, but it's been a couple of months and there hasn't been any update I'm aware of. Does anyone know anything about this?

[charred parcel]

there has been no additional news

[vague grove]

Oof, yeah, I didn't think there had been as I looked online before asking. I just hoped someone here might have additional insight to share, but no worries if that's not the case.

[frosty umbra]

codex app btw

[tawdry fiber]

Does enabling /fast in codex enable it across all sessions?

[trim jackal]

Why do I now just keep getting this error.

[brazen pivot]

Because life is always filled with unexpected errors, even some that takes longer than others to get away.

But they eventually get away.

You just needed to get the error in the first place to move past it.

The error won’t show up for you anymore now.

Enjoy yourself, have a rootbear.

Or two. 😉

• John Saint Carter.

[trim jackal]

I was asking for actul help with the issue...

Not copy pasted quotes

[brazen pivot]

Check now.

[brazen pivot]

Sure, copy paste or not. Does that change the meaning of the story in any way?

In order to question, you must understand it first.

[trim jackal]

Can you.. shut up if your not here to help people on this channel requesting it, or to ask for help with errors with chat gpt. Like go back to chatter channel or something

[brazen pivot]

Okay, fair enough.

You enjoy organization.

Why when it doesn’t matter in this moment?

Anyways, go to your settings and clear the cache for the app, then close and reload the app fully.

[sacred heron]

Hello, I’m having an issue with the codex cli while using wsl2 through the VS Code terminal.

Inside codex, my keyboard behaves strangely. I cannot hold down keys such as letters or delete. It only types one character instead of repeating, and delete or backspace does not repeat either.

Capital letters also do not register correctly when holding Shift, and even with Caps Lock on the letters still type in lowercase.

This only happens inside codex. Everything works normally outside of codex. I tested it in Git Bash and it works fine there both inside and outside of codex, so it seems to only happen with codex through wsl2.

[charred parcel]

Buy a Mac, you'll be much happier 🙂

[worthy belfry]

Hello guys. The generated download links havent been working for a few days now. I cant download anything apart from images

[sacred heron]

Ah yes, “Buy a Mac” as if I don’t already own several. The issue is specifically with codex cli running through wsl2 on windows 11, where I run my personal projects. Very helpful advice when someone is asking for actual technical help. Thanks for the help 👍

[rancid zodiac]

A. No B. I'd rather have no phone or computer than get stuck on anything Apple touches.

[split spear]

helle

helloo

[remote stone]

I just got a €220 pro subscription but when I open codex, it says I still need to upgrade to use the app fully. I’ve tried restarting, deleting the app and reinstalling etc and still the same - is there a bug? Which channel do I ask this in?

[rustic marlin]

are you on a mac. by any chance?

[rustic marlin]

weird, I use it in wsl and that does not happen to me

[sacred heron]

Turned out it was related to a vs code update that added something called the “Kitty Keyboard Protocol” for the terminal, which caused weird input behavior for codex. I had to add this to the settings:

"terminal.integrated.enableKittyKeyboardProtocol": false

[rustic marlin]

hmmm, good to know about that

[remote stone]

I am! Do you think it’s a Mac issue?

[carmine pawn]

I'm on the windows codex app, where's 5.3-codex-spark?

[charred parcel]

what plan are you on?

[rustic marlin]

there is a on goin login issue on mac, many users reported that

oai is aware and it should be fixed soon

[carmine pawn]

I'm on plus

[white ingot]

Currently, it's a bit annoying; sometimes GPT starts typing and then stops in the middle, and then nothing more happens. Then you have to type something new to get a response, or ask GPT to repeat the previous text so you can read it completely. (mobile app for Android, and it's incredibly annoying.)

[patent karma]

Spark is Pro only https://openai.com/index/introducing-gpt-5-3-codex-spark/

[gritty kelp]

QQ: I can use official connected apps like Granola in a normal ChatGPT chat, but not inside a Project.
Within the Project, the app does not appear, and I can’t call it.
Has anyone else seen this with Connectors/MCP/apps in Projects, and is there any fix or workaround?

The documentation says specifically it should work, so i don't see a reason why not. Tried to look for toggles, but wasn't successful (nor on google/llm/reddit). Any clues? Is that a bug or a feature? Thanks!!
https://help.openai.com/en/articles/10169521-projects-in-chatgpt

[carmine pawn]

Oh, ok
Thanks

[gilded cape]

Hi, just found a HUGE bug in 5.3, wrong logic in an answer, where should I open a ticket, please? Thanks.

[gilded cape]

Anybody in ?

[wintry parcel]

Hi, there. You can post on https://discord.com/channels/974519864045756446/1070006915414900886 If you'd like to open ticket with OAI customer support: https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[tropic herald]

can someone help me bruh

i made a virtual card

i used my real card

i changed my billing address twice

ive done everything

[mellow marten]

Is it possible that the chat archiving feature isn't working properly?

For months, I’ve been archiving finished conversations. However, when I tried to look for a specific one today, I realized my chats have been disappearing into the void for almost a year. When I archive a conversation, it disappears from my recent chats, but it doesn't show up in the archive either.

It seems like there might be a hidden limit on archived chats. If so, once that limit is reached, any newly archived chats are just getting lost without any warning.

I just realized that the error only occurs with project chats. Chats that aren't part of projects are archived correctly.

[crystal steppe]

I lost the message but whoever helped me and told me to use codex THANK YOU ITS AMAZING

[timber shadow]

Ayyy that was our conversation! #community-help message Super glad to hear you're having luck with Codex, it's really like magic huh
#codex-discussions is a great place if you are looking for help or more info in the future, there are a lot of experienced folks that channel!

[crystal steppe]

thats just the thing... IT DOES EVERYTHING FOR ME 😍

[proven coral]

Hi, I want to organise a prompt engineering competition for about 50 competitors and I want them to have access to non-free ChatGPT version. If I subscribe to Business plan, I can invite the participants but I realise the participants can also add others without my knowledge. Anyone has any advice on how I can conduct such a competition?

[green summit]

Debit cards should normally work for ChatGPT subscriptions.

(Real, not prepaid) credit cards also work, and depending on the country there may be other payment options too.

If it keeps failing, it could also just be the bank or the card's region, etc.

What is it exactly you're trying to purchase? Subscription or credits?

[tropic herald]

the free offer.

i used both my real and virtual card

i use revolut junior

i am romanian

i think my account is in italy but ive done both my italian and romanian addresses

[sterile anchor]

I love ChatGPT. I hope it doesn't go away because the company is struggling right now.

[rustic marlin]

Openai can be very picky with what credit cards they accept..
has to do with the fraud prevention, but the system is a bit too careful and often has false positives
best thing you can do is to contact the support on https://help.openai.com

[gritty kelp]

Has anyone noticed "Agent" tool limits (on Plus subscription) not being 40? I run every-day a News Task through a scheduled task, but today i noticed that i have reached my monthly limit for agent mode. That is weird, because months do have only ±30 days, and i haven't been using Agent mode otherwise. Anyone noticed similar behaviour?

[fallen venture]

Hi everyone, I’m looking to add an AI chatbot to my Discord server. What steps do I need to follow?

[charred parcel]

you need to write a bot and set up an API account and get an API key and deposit funds for the API usage here https://platform.openai.com/settings/organization/billing/overview

[fluid linden]

Trying to add photo and getting this screen

Its latest beta version on android

[fallen venture]

Cheers man

[fair veldt]

I have been running into deep research being used when not requested. I even added custom instructions in personalization settings to not use it unless specifically requested.

Is this happening to anybody else?

[delicate yew]

Can Company Knowledge or Codex be used without Github, e.g. with local git pepo like e.g. Gitea?

[mint ibex]

We can use chat 5.1 ?

[sterile anchor]

Can I rely on AI as my main source for learning the basics?

[hushed sedge]

this guy's ^ posted on every channel on the planet

[mint ibex]

?

[hushed sedge]

it's a bot. go look at every channel in this discord and on claude and he's posted that exact same message

[sinful kiln]

anybody know how to get out of this?

Suspicious activity detected
It looks like someone else may be using your ChatGPT account. Please secure your account to regain access to all features. Learn more.

[green summit]

Or maybe, they really just want to know. 🤯 I know, shocker! 🫣😆

[fluid matrix]

is sora.com working?

i can never get assistance on this discord for this service. smfh

[hushed sedge]

I'm sure that's it

they really, really, really, REAAAALLY want to know lmfao

[green summit]

So where's the proof for that?

[sly citrus]

Hey, I am working with the Agent Builder and am wondering, is there any way I can get access to the CLE name of the Classify Node and its category?
I want to link them to a Set State Node but don’t know how to call it and/ or reference the category directory in CLE.

[fluid linden]

Is there chatgpt pro in business plan?

That admin gives an account pro privileges?

[verbal matrix]

New here to the discussion on ChatGPT. Ran into an issue today where 4 hours of my work from yesterday is just gone. There are messages from before yesterday, and messages from today, but nothing in between. Weird thing is if I use the Search function, it can find content from yesterday's chat, but when I click on the result, the content is still missing. Any thoughts on this or where within Discord would be a good place to get feedback? TIA

[clear halo]

^^^
I'm having experiencing similar issues. All my pinned and more recent chats have disappeared

[verbal matrix]

Uff da. To be sure, It was just the one critical work chat that I'm deep into that I noticed the absence of content. I haven't really checked my other chats for data destruction. Been trying to rebuild all the assistant training patterns, and it's not easy, even with a brief (robot generated) log from yesterday's work.

[kindred lava]

Starting to run out of space with one of my chats. Besides taking a summary of the previous chat and using it on a new prompt, how else can I continue the conversation? (and no, paying for upgrades isn't a viable option atm)

[charred parcel]

the only way you can do it, no matter what plan you are on, is to create a summary and paste it into a new chat

[hallow shadow]

I can't upload C# file anymore
Location: China, Shanghai and I am using a VPN

I used to upload multiple files before.

[hallow shadow]

ME TOO 🤦‍♀️

Oh this ugly hexagon

[proven dune]

screenhand automation

[past zinc]

have anyone of you guys pinned convo gone missing? if i unpin it it shows up
i though my convo was gone ffs

[hallow shadow]

Yeah, what a goddamn basic feature. They can't even make it work.

[rugged wing]

They vibe code so hard and break things with every update. It’s like they don’t bother to QA.

[mental frost]

anyone here knows a clear idea on the parameter golf challenge. I need suport on how I can train the model. not the most familiar with runpod.

[silent topaz]

Hey everyone,

Quick question — do you think ChatGPT will get something similar to a “coworker-style” system (like Claude’s Cowork) anytime soon?

I’m basically looking for a system that can:

• help me automate daily repetitive tasks
• interact with my computer / browser
• manage and organize things I constantly deal with

From what I see:

• ChatGPT Agent is mostly browser-based
• Codex is more coding-focused
• Cowork feels closer to what I actually need

Also, I don’t really want to pay $200 for Pro — I’d probably stick to Plus or similar plans.

Another thing:
Does anyone here have experience with Claude’s Cowork limits?

• Are they restrictive in daily use?
• Do you hit limits quickly?
• How often do they reset?

If you have info, could you reply to this message or tag me? Notifications are off so I might miss it otherwise.

Thanks 🙏

[verbal matrix]

What day was this? All my work from 18-Mar-2026 disappeared on 19-Mar-2026

[past zinc]

Yeah around these days

[terse panther]

Any support team to explain a mysterious email i just received about Usage Policy Violation & Deactivation Warning, which comes as a true surprise!!

[charred parcel]

there are no support staff on the Discord server. You can reach support at https://help.openai.com

[verbal matrix]

They're at least aware of the condition of lost pinned chats.

[terse panther]

thank you for the tip

[steep mauve]

wsg guys

[umbral bolt]

If I cancel my ChatGPT Go subscription now, will OpenAI refund me for the remaining unused days?

[rustic marlin]

if you cancel now, you will still have access to it until the end of the period you paid for

[pale fox]

hi

Hi, I'm a beginner. I'd appreciate it if you could help me with Sora, because the videos it generates are low quality. I've already tried using Prompt, specifying that the video should be full HD with all its specifications, but the result is the same. Thanks for your suggestions.

[quartz girder]

and lost clicking buttons, copy and edit button

[cedar cloud]

I have been getting these errors in Codex:

⚠ Falling back from WebSockets to HTTPS transport. unexpected status 403 Forbidden: {"detail":{"code":"invalid_workspace_selected"}}, url: wss://chatgpt.com/backend-api/codex/responses, cf-ray: 9dfdfe4d4b84adc5-ATL, request id: 7aa14a56-da0e-46bd-a1f8-9230dab95b5d, auth error: 403, auth error code: invalid_workspace_selected

■ unexpected status 403 Forbidden: {"detail":{"code":"invalid_workspace_selected"}}, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: 9dfdfe797aefd845-ATL, request id: e2341910-951c-402c-a185-df354b037db6, auth error: 403, auth error code: invalid_workspace_selected

They happened yesterday too. They also happened for at least 1 day last week, then everything was fine for a while.
How can I troubleshoot this and/or get some support?

Edit: I seem to have solved this by logging out with /logout and then logging back in.

[quartz girder]

lost edit and regenerate button on the first chat

[hushed abyss]

how to fi

x

How to fix ^

[latent dune]

Hey friend, maybe you changed a folder locations within where codex workspace is working out of. So it's trying to write to an area, but you either moved the folder or changed the name of a folder so it's breaking path of where Codex has auth to work. That's my best guess 🙂

[small hawk]

Hi all — I work with cross-border brands and merchant-side BD tied to the U.S. market. I’m trying to better understand the public path for teams preparing merchants for OpenAI’s commerce ecosystem, especially around merchant readiness, product feeds, and ACP-related preparation. If anyone has practical insight on the right public path or best practices, I’d really appreciate it.

[little hound]

Turn off your VPN.

[hushed abyss]

Nah its not that apparently u gatta wait a day cuz of new users it woprks wether i have a vpn or not now.

[little hound]

🤷‍♂️ That's what worked for me.

[hushed abyss]

word

[little hound]

You want M3gan 2.0? 'Cause this is how you get M3gan 2.0.

[crystal chasm]

I look on Google, you are funny 😂. But there s no chance

[zealous abyss]

How do I make chatgpt stop talking to me like everything I say is the most intelligent thing ever conceived

[green summit]

"You’re absolutely right to call me out on that"

[small hawk]

Hi all — I work with cross-border brands and merchant-side business development tied to the U.S. market. I’m trying to better understand the public path for teams preparing merchants for OpenAI’s commerce ecosystem, especially around merchant readiness, product feeds, and ACP-related preparation. If anyone has practical insight on the right public path or best practices, I’d really appreciate it.

[quartz girder]

lost edit button on first chat

[golden ruin]

Hi Tristan, appreciate your interest, but not to spam the channel with the same question, we will leave your message here in #community-help

[small hawk]

sure

but NOBODY reply me😫

[kind hemlock]

What?!

[steady orbit]

Since Codex uses GPT 5.4 as well Is there a reason I could not use Codex create documents things like resumes in response to a job posting. I'm just wondering if I'm losing something by relying on Codex as compared chat GPT and the reason I'm interested in Codex is it seems to have greater programmability to fit into my workflow e.g. I can call into MCP servers and tools

Btw If this is the wrong channel to ask this question I would appreciate a pointer to the right place to ask the question

[delicate yew]

Is there a way to install Codex on Linux? Specifically NixOS

[rustic marlin]

Codex CLI

[tight vector]

guy I want to make a rotoscoping animation like "100 meters" animation. I have my own video that I recorded, so I wonder if there is any AI that can generate a video-to-video?

[mellow adder]

what is this system bro

[mellow adder]

nothings even bad in the prompt

[nocturne frigate]

Do brands find themselves on one side of the border more often, or do truly multi-border brands exist?

[nocturne frigate]

You'll probably be token-limited. If you try generating short clips and stitching together you might be able to generate and join high-quality snippets. Otherwise, try playing with visual effects or art styles where quality isn't as noticeable a qualia.

[delicate yew]

Do you know if there is maybe a read nix package, or am I forced to use NPM and/or this repo?

[rustic marlin]

official is npm

[shut tree]

How long did it take yall to get verified as an individual to access protected models and submit ChatGPT apps. I uploaded my ID/facescan and it said verification successful but it still says identity in review?

[regal jolt]

Hello. First of all, I’d like to apologize for my poor English. If I don’t express myself clearly, please let me know.
Second, I’d like to ask about a question I have regarding how this works, since I don’t trust what the technical support agent is telling me or what ChatGPT itself is telling me. I usually use the model to detect errors in code, and in my spare time, I use it to help me analyze, correct, and improve my writing in my native language. The thing is, sometimes it’s necessary to include cultural references in the text. On those occasions, the model seems to search the internet and dynamically cite sources without me asking it to, integrating that information into the conversation as if it already knew it—even though it incorporates the data as if it had pulled it from the internet, even though the search or thinking animation doesn’t appear.

This doesn't worry me, since the data it integrates is accurate and complements the text. What does worry me is data leakage—that is, when I ask it to search for something online, an animation appears; when the question is complex or takes longer, a thinking animation appears. In those cases, I understand what it's looking for, but when it dynamically integrates data, I never know when it's going to do so. When I ask where it gets the data from and regarding the images, it tells me that it generates them using its internal tools and that the data is included in its knowledge base prior to the training cutoff. Curiously, the URL from where it supposedly extracted the image appears. I go to that URL and it’s correct—the image is there. So my question is: does the model leak data from the conversation? Does it leak context from the conversation when it searches? When I ask the model about this, it assures me that nothing is filtered and that the conversation is private, but I don’t trust it since the model has failed in the past. So I contact OpenAI via their support email, and to my surprise, the one who responds is a GPT AI agent who tells me the same thing: that there is absolutely no filtering because intellectual property is always respected. But what’s strange is that there’s no way to contact a human who can actually confirm this for me.

[golden ruin]

When web search is used, it sends a clean query, not your raw chat history

[regal jolt]

No, I don't mean search mode; I mean a normal conversation using the automatic mode.

[kind hemlock]

so u mean links of previous chats nowhere get cited in a new msg?

is that what u meant?

[regal jolt]

No, imagine you go to the website and simply ask it a question, and that question is complex or requires additional information, or it decides it would be better to add information to enrich the conversation—it adds this dynamic information without performing an internet search, yet references to URLs appear.

A example

[fading kestrel]

iv used chatgpt for a long while now but i only recently got into giving my chatbot a personality and experimenting with the limits of what it can be capable of. Me and my friends got in a decent spot characterization wise and i was curious about adding it to our server so they could interact with it while im busy or away, as well as have it automate some things and i was wondering if there was a way to do this?

[steady orbit]

Apologies for the newbie question. Hey folks are there other forums to get questions like these answered.

[onyx kraken]

It's totally fine to ask again after a while!

I personally don't have a strong feel for how to answer your question. I've actually not used Codex yet, so I can't compared. I'm aware that many Codex-involved folk may cluster in #dev-chat , but I have no idea if any of them are using Codex for resumes or other document creation.

[onyx kraken]

Far as I know, that involves using the API, not ChatGPT itself, you totally can do both or either, but that's playground and a bit different, and more coding-like than ChatGPT. You maybe can even ask ChatGPT how to make a... is it a discord bot you want? A website chat interface? API tools is usually how things like automation are handled, in a programming or tool calling sense.

Note, if you made something like instructions for a personality, if you wanted to and your friends wanted, you could share the instructions for them to work with their own account using the instructions you shared.

[fading kestrel]

I wanted a discord bot and i did kinda fall into an api rabbit hole but i dont mind looking into it. Looks like fun to learn (i say while its not frustrating yet)

[onyx kraken]

Awesome. I solve a lot of stuff I'm interested in but confused about by asking ChatGPT to help me with the specific step I'm struggling with, and maybe I ask for alternate ideas. It'll tell you how to tie your shoes with spagetti, if you ask, but it may not introduce you to shoelaces if you seem to actually want to use spagetti; meanwhile I, maybe others, may not know that shoelaces exist but we've seen spagetti so thought my question made sense.

[fading kestrel]

Yeah i see what u mean! I didnt even know using apis to get ur own server bot was a think im really new to all this myself

[onyx kraken]

Ahh! There's also #codex-discussions which might be even better than the dev chat

[kind hemlock]

yeah that sounds familiar

[slim hamlet]

/docs/reference/checkout on acp site

[small hawk]

Hi — sorry for the cold message. I work with cross-border brands and merchant-side business development tied to the U.S. market. I’ve already submitted OpenAI’s official partnership intake and am trying to better understand the right public path around OpenAI’s commerce / ecosystem direction. If there’s a more relevant team or route for serious ecosystem / services partners, I’d really appreciate any direction. Thank you.

[small hawk]

dm u

[regal jolt]

So there's no answer to my question?

[torn island]

NEw to this discord - I was looking to see if anyones has vector stores that are taking 20 mins to update instead of 2. I've only just joined the discord and read where to look but can't find anywhere

I saw the status update so wondering if others are seeing the same

[kind hemlock]

idk what ans to give tbh

[atomic belfry]

Hi, im trying to claim the offer for ChatGPT Plus but i keep getting an error "please try again later" while my wife on same setup got it easly, i address support but didnt get any replay yet. any1 can help?

[idle briar]

I’ve been using Codex for a while, and the main problem I keep running into is the 5-hour limit. I’m a web developer working on pretty large projects, so even simple prompts can make me hit the limit very quickly. Is this just the normal limit for Plus users, or is there any better plan or workaround for this?

Also, if anyone has a Pro plan and doesn’t use Codex much and would be willing to help, it would really mean a lot.

[atomic belfry]

any1 can help?

[onyx kraken]

We're nearly all community members here, the very few of us who work for OpenAI have gold names. Since your issue involves PII and financial info, not just a possible buggy cache or something-

Hey, in case that's it, try restarting the device and clearing cookies and stuff, then try again. If possible, try on a different device, one you feel safe inputting info on.

Otherwise, if it's anything more than device level simple help, nobody here could help you, personal and financial data, OpenAI only handles through its official tools and the https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[atomic belfry]

yea tried everything diffrent devices, diffrent broswers, its a shame they got a buggy payment method for pepole whoe wants to buy

[idle briar]

Any one can help

[onyx kraken]

I hope they can help you through the official help page support with whatever your actual issue is. They have many hundreds of millions of members, and there can be a queue for help to be able to reach you.

[summer glacier]

Hi, I have applied for Codex Credit with my school email but I haven’t received confirmation email

can anyone please help

[regal jolt]

I'd recommend that you don't use the website's payment system—I had trouble with it at first, too. Instead, if you have an iPhone or an Android phone, try linking your ChatGPT subscription to Google Pay or Apple Pay. I bought my subscription through Google Play years ago and haven't had any problems.

[soft kettle]

why is sora 1 going crazy rejecting my request for the devil outfit to be dressed properly

[slim hamlet]

bg?

[bleak oriole]

hello my gpt lags so much on pc it works fine when u start a new convo but its gets bad when u try to strt a convo in an existing chat

[mighty moat]

well hello beautiful... where can I meet her?

[summer glacier]

Can anyone help??

[onyx kraken]

We're nearly all community members here, the very few of us who work for OpenAI have gold names. None of us have access to OpenAI's processes, but you can reach out to https://help.openai.com/en/articles/6614161-how-can-i-contact-support and make a support request, and someone who does tech support for OpenAI will answer you when they can!

[lunar lantern]

Hey — hoping someone here can help or point me in the right direction.

I submitted a data export request about 2 weeks ago through the privacy portal. After multiple follow-ups, I finally got an email saying my data was ready — but there’s no download link in the email, and the button just redirects to a default page.

Support has been slow and my case was even marked as completed without actually resolving the issue.

Has anyone else run into this or know how to fix it / get it escalated properly?

Case #: 06896182

[onyx kraken]

When I request a data export, if I haven't gotten it within 24 hours I request again. Sometimes repeatedly, if that's needed.

When I get my email, I use my button within 24 hours, I've seen it appear to fail if I use it after that point (And that means I request another export to try again).

[short ocean]

This needs to be selectable, i hate getting stuck on Thinking if i select it manually if i use up the Instant rate.

[delicate yew]

Can codex cli work offline?

[charred parcel]

no, it cannot

[astral narwhal]

Need some help ! My Codex limit was exceeded but I try to upgrade plus it doesn't how to upgrade go -> Plus

[onyx kraken]

I would reach out to OpenAI tech support, the bottom right widgit here: https://help.openai.com/en/articles/6614161-how-can-i-contact-support? The article helps explain how to ask for help.

[quaint imp]

Since today, GPT keeps generating images for almost every request I make, even when I’m not asking for it. Could you please help me with this?

[midnight hearth]

Tell the showoff not to? 😛

[onyx kraken]

I would ask in a new chat, 'Hey, how do you decide if I want you to make an image for me or not?' and go from there.

I note that I have not had any unrequested images made today, or recently. Rarely I've been surprised with 'making an image is a neat, creative way to fill that request!' but... more than 6 months ago. Maybe much more.

[onyx kraken]

If you're interested, I think that could be quite helpful. And we can compare it to how the base model answers to try and figure out why yours is weighting 'images omg yeah!' and how to fix.

[slate halo]

Guys why can’t I attach ref images in the image bot channel? Please mention me for the answer

[delicate root]

I’ve had a few chats where chunks of the thread have completely disappeared. These are not threads that I have revisited days later, but during the live exchange within minutes if I scroll up to reference or review a previous response or comment I have made, there is parts of the threat that are no longer there and ChatGPT has no memory of what was discussed minutes earlier. There’s no error message but comments me by me or responses by ChatGPT, including text and attachments are completely wiped out. The whole thread was not lost. Just random chunks of the conversation. This has happened on multiple threats. Sometimes exchanges for Minutes to hours earlier randomly are gone or missing.

[burnt pollen]

I've been trying to contact support regarding a billing related issue for over three weeks, and I'm not convinced that the human support that has only weighed in about three times over a period of 22 days has any more ability to resolve issues than the AI-based support. Is there not some way to actually get in touch with a billing person that has the ability (i.e. access) to fix a problem with my account?

[elfin orbit]

Found a serious bug, any suggestions for how to get a quick resolution? https://community.openai.com/t/custom-gpt-code-interpreter-doesnt-work-on-non-creator-accounts-when-knowledge-files-are-added/1378049

[vague glen]

is there anyway to add openai in my discord server

[charred parcel]

yes, write a bot and interface it to a model on the API

[vague glen]

i dont know how to do it

oh yeah its my friend discord seerver

mb

[charred parcel]

enter this question is there anyway to add openai in my discord server into ChatGPT and it will telll you how

[vague glen]

ok

[brave walrus]

any way we can get the support team to take the data sharing token discrepancy seriously? i consider this theft and fraud. i just dropped pro sub because of the absolutely garbage responses from support. basics are, data sharing allows for daily free token allotment based on tier. it allows for x amount of free tokens for larger models and lists them and x amount of smaller models and lists them... for 5.4-mini and nano (listed in the 10M token group for me) it's charging after 1M. support doesn't even think the model is in the data share program. so... i switched back to 4.1-mini for that automation and stopped being charged until hitting over 10M... i gave them receipts. exports of my charges, screen grabs, everything... they came back with out of date tone deaf non answers that will not fix it for me or anyone else. there are multiple forum topics about it. is theft from customers not a high priority to fix? and then... when i went to cancel my pro sub over this mess with support insisting it's a me problem, the account management button wasn't present. i finally got one instance of chatgpt to show the manage section in the account tab after tweeting x employees about it, and dropped myself to plus tier. i can't in good conscious pay you guys $200 a month while you essentially steal extra money from me and refuse to acknowledge or fix the issue. to be clear, i now have the funding to keep pro, and i really like the usage for codex, but i've chosen not to for now because of how insane this support issue has become. please... stop stealing from customers that made a deal with you to share data in exchange for tokens when you don't honor that deal. until then, i'm on an anti-oai tirade. i feel wronged. i'm in a bad mood. my first support response was received on the 23rd. every time i receive a response, the answer is worse than the last time. please... get ops involved. fix the discrepancy. refund customers for the overcharging you're doing. just... do better.

to repro the problem, go to this page and enable the datashare... if you've got the tier prereqs, you should see what i see. use 5.4-mini or 5.4-nano though an api key with this enabled... then check your billing after ~1M tokens used. you'll find you're being charged. next, change the model on the next day to 4.1-mini and use over 1M... you won't be charged until after 10M... in no world is this a user bug. convincing support is the real problem. getting action on this is so far impossible. forums show others with the same problem. still... no action. please... escalate this. save your company's rep. stop flushing it down the toilet. thanks.
oh, and for the profanity laced full on gille meltdown about it, just check my x account. screengrabs included.

[cold elbow]

hey guys, why when I wanna use agent mode the arrow is gray and wont let me click it?
it looks like agent mode is disabled but idk why

[brave walrus]

do you have agent usage left?

[marsh grail]

i forked and used codex to make a change because i had the same issue natively on macos tahoe, so i will pbulish my fork's changes, open a PR upstream, and let the bigger people ensure validity and that it works across all!

[noble iron]

anyone got any ideas why when i generate images my chat sometimes just gets this blue spinning icon next to it and stops doing anything at all?

[steady bear]

No

[wild river]

sora throwing this error with every generation attempt, is it down for good?

[carmine dawn]

I have the same error though I even have my phone verified, does someone know how to fix it?

[sacred heron]

That’s awesome, thanks for doing this. I’m not sure if this will work on macos tahoe, since I had this issue on windows 11. I fixed it by adding "terminal.integrated.enableKittyKeyboardProtocol": false to the settings, which resolved my problem. Hopefully your PR gets accepted and pushed

[lethal summit]

guys how do i stop chatgpt from saying "ahhhh" at the start of every sentence

[onyx kraken]

Tell it what else you want it to do instead.

[hollow breach]

hello; since 3-4 days i can not edit any photo on chatgpt using selection area feature, if i try i keep getting “I wasn’t able to generate the edited image due to a technical issue.” error
It s just me or a server issue?

[finite charm]

Hello friends…
Might anyone know if these TestFlight invites to a OpenAI Ads Suite are legit?
I received an invite but don’t recall requesting one. And I read the min buy for ads was $200k.

[trail cairn]

scam

[graceful citrus]

Hi! Can somebody help, i wanna try gpt agentic, but virtual browser doesnt work. How to fix that?

[gleaming cipher]

Can someone help me? I'm from India but the chatgpt go is still 300 rupees instead 2 rupees for the monthly offer.

[left mantle]

Hello! How can I contact technical support? I am experiencing login issues

[green summit]

For browser:

Open browser > settings > find "site settings" > then "all sites" > search for "OpenAi", "GPT", and "Sora" (when it's for Sora) > delete data > then reopen the browser and try again.

For App:

Go to the homepage on your phone > hold your finger on the GPT (or Sora) icon > select "App info" > find storage > "clear data" > reopen the app and try signing in.

(Don't worry, none of this has an effect on your account data)

[wheat walrus]

Hello, why is this?
help me please!
My internet is good.

[onyx kraken]

You'd need to reach out to OpenAI tech support, https://help.openai.com/en/articles/6614161-how-can-i-contact-support?

[wheat walrus]

please help

[warm flare]

why my gpt 5.4 pro model is always routed to gpt4o-mini?

anyone can help me? 👀 👀

[verbal parrot]

ChatGPT is displaying “Updated saved memory” for a newly created memory, and then the "Manage Memories" menu does not show that same memory for deletion. Doesn't it say in the Memory FAQ that "we are always in control"? I have many such "memories" that ChatGPT creates by itself and that do not show up on the menu.

I told it: "I want to delete memories you create; I want to be the one creating memories."

And it goes ahead and:

1. Overrides my prompt completely
2. Creates the memory anyway
3. The memory is contradictory: "the user does not want this stored", yet it is an "Updated saved memory"
4. The memory does not show up in the menu

I know the bot does not do this on purpose, but it's still infuriating. Please, support, look into my request. Thank you.

[wheat walrus]

Hello, why is this?
help me please!

[golden ruin]

refresh?

[wheat walrus]

I tried it doesn't do anything!

[waxen robin]

https://sora.chatgpt.com/invite?code=3YEGBY

[elfin orbit]

What official channels exist to report bugs to OpenAI?

[charred parcel]

#1070006915414900886

[languid cipher]

I have the $20 GPT subscription. My live camera has worked a hand full of times. It doesnt even show up as an option anymore on my account. Its almost as if it never existed.

Is there a fix to this problem?

[charred parcel]

The camera works in Advanced Voice Mode

[languid cipher]

How do I get access to that? If you're talking about the voice mode where you talk directly to the AI and it speaks back, i've tried that. Thats where I found the camera initially, now it doesnt even appear

[warm tartan]

I’m working on a project where I want to use AI / computer vision to read MEP (Mechanical, Electrical, Plumbing) drawings or floor plans.

My goal is to:

Detect rooms and extract their labels (e.g., “Bedroom”, “Kitchen”)
Extract room dimensions (length, width, area)
Detect windows and doors and their positions
Use this information to perform heating calculations based on custom rules/standards

Challenges I’m facing:

The drawings are inconsistent: some include a scale, others don’t
Dimensions are not always clearly associated with a specific room
Text (room labels and dimensions) is detected using OCR, but linking it to the correct room is difficult
Layouts vary significantly across drawings

What I’ve tried so far:

Using OpenCV for image preprocessing and contour detection
Using OCR (e.g., Tesseract) to extract text
However, I’m struggling to correctly associate dimensions with the corresponding room geometry

What I’m looking for:

Recommended approach or architecture for this problem (
Best way to associate detected text (dimensions) with specific rooms
Any models or techniques for understanding floor plan structure (e.g., graph-based, segmentation, object detection)
Suggestions for handling missing scale information

Are there any existing libraries, models, or research papers that address this type of problem?

[hasty olive]

hello , i just joined discord , i use GPT 1 year now , but i can not upgrade to pro , my bank said it is token 3ds secure from OpenAI , i can not put my business card at billing profile (your card declined) , and i can not upgrage.. (Payment error. Your card may be invalid or authentication may be needed.)

I have test different , browser , icognito mode cleared catch and cookies, whitelsit open ai , chat gpt and stripe , used phone and data , called the bank , cancel the renewal nothing worked

any idea?

[plucky sierra]

Hopefully, someone can help me. I keep getting We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly. The OpenAI API expects a JSON payload, but what was sent was not valid JSON. If you have trouble figuring out how to fix this, please contact us through our help center at help.openai.com. When i message In Axiom Ai. I cannot figure out what keeps casuing this error.

[minor rampart]

I’m facing an issue inviting a specific user to my ChatGPT Business workspace and would appreciate any guidance.

Error:
“Unable to invite user due to an error”

starting yesterday. when this will be fixed?

[mint hedge]

Someone please can help me why can't I get new free video gens every 24 hours in SORA 2? it started since April 1st

[trail cairn]

sora will permanently shutdown in less than a month. remaining time is dedicated to paid users, free users can no longer generate

[frosty condor]

What happend to sora 2😭😭😭 (ik it’s looks like sora 1)

[frosty condor]

Well, can we wait for sora 3?

[green summit]

You can wait for Sora 3, but it will never come.

[shadow stag]

OpenAI's API keeps charging even after I deleted all my api keys... anyone from openai to help?

I’m hoping someone can help me understand what’s going on with my billing.

I recently deleted all API keys from my OpenAI account, but I’m still seeing recurring charges. Looking at my invoice history, I have multiple charges within just the past few days:

• 2067ACAD-0030 – $79.36 (Apr 2, 2026)
• 2067ACAD-0029 – $77.41 (Apr 1, 2026)
• 2067ACAD-0028 – $78.78 (Apr 1, 2026)
• 2067ACAD-0027 – $77.77 (Mar 31, 2026)

These charges are happening very frequently — sometimes multiple times per day — even though I have no active API keys. I’m confused about where the usage is coming from.

I’ve already deleted all keys but the charges haven’t stopped. Any guidance on next steps — or whether I need to contact OpenAI support directly — would be greatly appreciated.

then i have to delete my credit card from the platform and now it shows negative:

[trail cairn]

check which project is using them as its should be possible to see it

[shadow stag]

i have deleted all the api keys ...

there are no api keys associated with any projects

[trail cairn]

but project is not the same as key. projects can have multiple and then there are the legacy keys

[shadow stag]

do you mean i have to delete all projects?

[trail cairn]

no

find out the project from dashboard/export. if its empty then its one of those bad legacy keys

[shadow stag]

how do i deactivate the legacy keys?

i have no active api keys in the project and there is no setting for me to deactive any legacy keys

[trail cairn]

if you can see them at org level then there but some of the oldest ones may not show up

[shadow stag]

this design is ridiculous...

[trail cairn]

support is the only way then

yes

[shadow stag]

thank you for the info

i couldn't find the support. all ai ...

and couldn't understand anything

[trail cairn]

say to the bot "escalate this issue to a human"

[shadow stag]

thank you man.

i am so disppointed with openai...

it was my go to ... now i have cancel all their services

does the default means: if an api key is deleted from the project or org, the api key should be deactivated?

the person who designed current vesion should be fired!

no basic product or tech sense... no wonder the product is deteriating so much

[zinc flame]

anybody getting this trying to login to codex app on os x?

[charred parcel]

just logged out and logged back in and it was fine

[zinc flame]

hmm.. maybe i have something else occupying the port..

ah indeed.. i'm an idiot.. ignore.

[charred parcel]

wrong app?

[zinc flame]

passthrough to a docker instance that i had piped the cli into shadowed the os x version

i guess the mac os app doesn't check if port is available

this is what happens when you get distracted for few weeks and forgot what you were originally working on..

[minor rampart]

its been two days and still counting. whats wrong with business plan? they keep charging me but I cant invite at all

[kind gust]

well well, just switch to claude atp

[finite inlet]

Currently, the ChatGPT application is completely inaccessible with the TalkBack accessibility feature; furthermore, it is inaccessible with any other similar software as well. This will cause significant difficulties for visually impaired individuals attempting to use it.

[next scroll]

Hey guys have anyone had a glitch in their chat got where Hindi showed up for no apparent reason? I speak only English and for some reason my chat gave a Hindi word

[midnight hearth]

It's a semi-common error for one semi-related word from another language to slip into an answer. I think it has to do with LLMs being probabilistic, so when it hits a low-probability roll the result is sometimes that it outputs a sufficiently related token from the wrong language.

[languid cipher]

My live camera mode doesnt work on my GPT subscription. Does anyone know why?

[raven linden]

I wonder if they forgot to set top-p to something reasonable

[terse panther]

anyone working with seedance 2 experiencing drop in quality? emotional impact etc?

[livid sequoia]

Has this happened to anyone else?

I added $10, but my balance is still the same. Even though the balance hasn’t changed, my April budget shows more than $20 spent, despite having auto-recharge turned off.

Also, the amount of tokens used is not showing up in my usage history.

[short patio]

does a business seat have the same usage as a plus seat for codex?

[fleet anchor]

How to delete an organization and a project in the OpenAI API platform?

[austere snow]

As for this message does it mean in general for my chat or just per day? Like I have the $20 plan atm

[remote shale]

Hi guys! Im a CS Student making a CLI tool that can use other agentic software like claude code, codex, cursor, antigravity, etc. Is there any features that you guys wished claude code had but doesnt currently?

[green summit]

Just that chat, you must continue in a different chat. (A daily/hour limit will be slightly different)

[deft juniper]

will i get a refund since i thought the free trial would end at 11:59pm of the end date?

i got charged 200+$ for the free trial of the business plan

i have proof + last 4 digits of the card i used

im also using it for 2 hours more (since its 10pm now) after i sent an email to support under the assumption i can use it up to (but not after) 11:59pm

[trail cairn]

did you read the terms when you accepted the trial? could have cancelled days before the end

support page and bot is here https://help.openai.com/en/

[deft juniper]

yeah i forgot i thought it was for the full day at the end though

[deft juniper]

i got the refund js now bois

main chatgpts support is actually pretty fast. much faster than some other llm providers

[hybrid lodge]

API fine tune constantly gives me “We’re having trouble accessing your files right now. Please try again later.” Do you have any solution for this

[toxic wraith]

I have a question regarding Codex usage.
Currently I have exhausted my plus subscription weekly usage (resets April 10), and Codex tells me I can buy 1000 codex credits for 40€.
Now my question is, how much is that actually when compared to a 23€ brand new account Plus subscription? Do I get 2x that amount? Is it even comparable?
I have no idea how much you actually get when you buy a 23€ Plus subscription - it is isn't said anywhere - I'm just trying to get the best bang for the buck.

[wind shore]

i

[onyx kraken]

Codex is extremely flexible, there are essentially no 'set' prices, because the task complexity, your message length, the speed configuration you choose (Fast on/off, more speed info https://developers.openai.com/codex/speed), and if your coding is being done locally or in the cloud,

Small scripts or routine functions may consume only a fraction of your allowance, while larger codebases, long-running tasks, or extended sessions that require Codex to hold more context will use significantly more per message.

Per https://developers.openai.com/codex/pricing

You can track your usage per task here, https://chatgpt.com/codex/settings/usage by checking that after you do each Codex thing, I'm not sure how often it updates. You can use that to check how your credits are being spent if you choose to get some and monitor.

Credits are now spent at token-API cost ratio, so they check the amount of tokens your total task needed and charge per that (input, cached input, output), https://developers.openai.com/codex/pricing

You can check the API prices at https://developers.openai.com/api/docs/pricing

I'm aware that isn't a crystal clear answer; I'm not aware of any crystal clear answers except if you have membership usage left your plan would use that and not your credits, and credits get used when you otherwise wouldn't be able to use the service you try to work with.

[wanton hedge]

finetuning og chatgpt 💔

[inner pivot]

Has anyone here dealt with this exact ChatGPT Business / Codex migration issue?

I was on individual Pro, tested credits there first and it is message based , similar value to the actual sub, and based on that understanding moved myself and my small team to ChatGPT Business. Only after switching did I discover Business Codex is already on the new token-based / API-style pricing, while what I tested on Pro was still effectively the legacy model.

The article I only found after switching was:
https://help.openai.com/en/articles/20001106-codex-rate-card#codex-rate-card-token-based-pricing

So I tested credits under one pricing model, switched, and then landed on a completely different one.

Within about 24 hours after switching, we got hit with around $50 in subscription charges and around $150 in credits, and support’s answer so far is basically:

• merge cannot be undone
• personal workspace cannot be restored
• if I want personal / Pro again I need a completely new account

That is super frustrating because the migration was irreversible on the same account, and the pricing difference was not clear enough at the decision point.

Has anyone here:

1. gotten an actual exception or rollback from support,
2. found any supported workaround on the same account,
3. or dealt with this Business vs Pro Codex pricing mismatch before?

[sour cloak]

#community-help Hey has anyone faced the issue of internal server error while doing domain verification when we are creating ChatGPT Apps. I have added the token txt as well as on the same path that is mentioned still I am getting the domain verification issue. Can anyone please help with this issue

[woven valley]

hmm very funny sudently in the chan my gpt don't want make generate picture no any sense about that

[midnight nymph]

With the API how long does it usually take for usage to be deducted from your credit balance?

[main badge]

Where do I report bugs?
(On Microsoft Edge & Windows 11 latest stable builds, when hovering the ChatGPT button on chatgpt.com, once the hover-popup goes away from "un-hovering", the messages (your chat) gets scrolled up in a laggy way.)

[frail slate]

(https://discord.com/channels/974519864045756446/1070006915414900886)

[wraith sierra]

Anybody else has this issue where codex plugin on windows in VS opens the file link in the chat content into browser (which obviously doesn’t actually open the file) instead of actually opening the file in VS code editor?

[jaunty crow]

Hi.
My codex limit finished can you help me

[jaunty crow]

Hi

I didn't premium for 1 years but I didn't get benifits can you help me or can you check my account

Do you voice conversation channel

[crystal steppe]

Hallucinations 🥺 usually when this happens we get a better model tho 😉

[cyan vault]

MY model is constantly repeating itself during conversations any idea how I can fix this issue? or wat could be causing this issue. The model is connected to lip sync avatar

[stray owl]

Anyone else have issues with the GitHub connector in the chat?
Like, I have my repo attached but it won't understand how to read it or something?

This is 5.4 Pro on the web

Like it starts searching the web for a private repo

[pulsar willow]

Help I cannot make videos on sora 2

[compact rune]

You limit has reached

[jaunty crow]

i am getting error in cli

can you help me in chatgpt

[ornate ivy]

no otpion show study

[drowsy current]

Scroll down

[ornate ivy]

No bro i try everything but not found

[tender river]

if i want it to give instructions to my codex agent are skills the proper way to do so. I just cannot understand the difference between a skill and the instructions that go into agents.md

[patent anvil]

I think agents.md is always read in the context... whereas skills load the name/description in the context, but only adds the full skill to the context when it decides the skill is needed for the particular query... so if it's something long that's not needed with every query, something you want to make more like a workflow that's only invoked in certain cases/prompts, a skill makes more sense.

So think of skills like discrete modular workflows/tasks... ideally ones that aren't used with every request... something specific like - this is how it does thing X.

[acoustic linden]

Why can't I select Agent Mode within projects anymore? It's only available in the “New Chat” section, and even existing Agent Mode conversations are gone.

[near willow]

Is it normal that I can only delete 3-4 chatgpt conversations before it stops me for saying too many requests? Is there a different way to delete more than one chat a time? (I don't want to delete all of them)

[woven valley]

Hello anyone know where is on chat Gpt Project folder i had upgrade from premium plus to pro but my project folder get gone

[weary bridge]

Is there a GPT bot that you can invite into your discord server ?

[charred parcel]

not one off the shelf. you will need to code one

[weary bridge]

thank you, is that easy, like can you use chatgpt to make a discord bot?

[charred parcel]

you could have Codex build you a Discord bot

[weary bridge]

cool! Any pointers for a newbie starting out I'm going to try this

[charred parcel]

this website is where you will need to create an API account. There are full docs on the API here -> https://developers.openai.com/api/docs

[weary bridge]

thank you kindly sir

[frosty vapor]

Hello. I'm having this issue

I could use it when I was plus, but now that I upgraded to Pro I'm getting this.

I'm on a personal account

[misty rock]

So, what's the advantage of the $200/mo plan now vs. the $100/mo plan?

[vivid mortar]

just usage limits really

[onyx kraken]

I would reach out to OpenAI tech support, https://help.openai.com/en/articles/6614161-how-can-i-contact-support?

[frosty vapor]

I think the tech support bot just crashed lol it errors on gen and chats are not loading

[frosty vapor]

Nevermind, it was a weird cookies issue.

[onyx kraken]

Glad you found a fix!

[versed urchin]

Hey everyone 👋

I’m a CSE student exploring AI, hackathons, and startup communities.

I’m trying to join some good Slack communities (AI/dev/events/workshops/startups/internships).

If anyone is already part of any active Slack groups, could you please share invite links or add me? 🙏

[young zealot]

Hey everyone! Recently, ChatGPT has been triggering image generation on its own, even without explicit requests. It mostly happens while I’m working on prompt engineering for Nano Banana. Has anyone encountered this behavior or found a way to disable it?

[onyx kraken]

I would discuss with it, like I'd explain to a person who was very friendly and helpful, that what I really want is the prompt to take to the different image generator. I'd be gentle but clear if it made an image anyway, as well as periodically confirm 'this is perfect, I want the prompts to use elsewhere, just like you're doing now'. This helps it stay locked in; over time it can drift and predict 'but, surely you want images, I can help!'.

Here's an example chat where I set up 'my way' for how I'd proceed. You can if you want read over this chat and see how I did it - how the model outputs, and decide yourself if you want to continue this chat yourself, or instead if you want to start your own with whatever ideas you have. It's totally fine to just continue here and see how that goes for you, if you want to try.

https://chatgpt.com/share/69d93c12-312c-83e8-9552-4a2daa338ef5

[spiral verge]

Hey everyone 👋
Glad to be here. I work mainly with funnels and automation systems, helping creators and businesses turn their traffic into leads and paying customers.

Most of my work involves building and optimizing funnels on platforms like GoHighLevel and Kajabi, while setting up automations that manage follow-ups, lead capture, and customer journeys.

I’m here to connect, learn from others in the group, and share ideas. If anyone ever wants to talk about funnels, automation, or scaling offers, feel free to reach out.

[spice tree]

Can someone help me I used to have JGPT plus subscription on my account but after months when I have login into my account the subscription is gone although it claimed for one here why had that happened

[frosty vapor]

I was wondering, does Pro get the full context window for GPT-4.5 or does it get a reduced one? If it's the latter, what ctx do we get?

[short ocean]

Uhhh... Is this a glitch?
I can read the sent image just fine...

[green summit]

GPT being GPT.

[short ocean]

Dot't worry, i submitted a bug report.
Link: https://discord.com/channels/974519864045756446/1492428227220869130

[civic horizon]

Just report a bug, it's simple glitch

Wait, and try again later

[neon aspen]

Quick help pls! I’m trying to get ChatGPT Pro on Android but only Plus is showing up. If I buy Plus first and then upgrade, does the $20 get refunded or turned into credit automatically? Or do I need to manually apply for a refund? Thanks!

[lyric igloo]

please help me

The workspace isn't loading

[civic horizon]

Okay, don't worry! Just try to reaload your wifi router, reload the page or check ChatGPT's status here: #status

[civic horizon]

If you buy Plus first, then upgrade to Pro
You will generally be charged for Pro separately at full price
Your existing Plus subscription usually does NOT automatically convert or get refunded
Any “unused Plus time” is not reliably prorated as credit in most cases

👉 So yes, there is a real risk you could effectively pay for both in the same billing cycle.

[civic horizon]

If you subscribed via Google Play (Android) → refund is usually handled through Google Play + OpenAI support
If you subscribed directly on web → you request via OpenAI support chat

Typical rule:

Refund requests are usually only considered if it’s within ~14 days and eligible circumstances

[minor rampart]

same here

[civic horizon]

I can help you

[minor rampart]

what should I do

[civic horizon]

That error message usually means your subscription status couldn’t be fetched properly from OpenAI’s servers. It’s fairly common and usually temporary. Here’s how to fix it:
🔧 Quick things to try:
Refresh the page / restart app. Close the tab or app completely and reopen it. Check your internet connection. Switch between Wi-Fi and mobile data to rule out network issues. Log out and back in. This often refreshes your account/session. Clear browser/app cache, If you’re on mobile Safari/Chrome, clear website data., If using the app, reinstall it., Try a different browser or device, Helps determine if it’s device-specific.

I had the same thing when i'm was trying to code something in Codex. that really helps

[civic horizon]

Oh, wait if that not helps. We have a trouble with Codex and ChatGPT

[green summit]

When you upgrade to Pro while having Plus you'll pay for the amount that covers the days from [time of upgrading] to [end of the already existing subscription month].

For example, if I were to subscribe to Pro from Plus right now, I'd have to pay 39.06 for the "first 5 days", after those 5 days my (already existing) subscription will be renewed and automatically be billed the full amount of Pro.

If all that made sense lol

TL;DR: You do not get refunded, you have to pay more to "cover the days until the end of the month" and will be billed the full amount at once next/upcoming bill.

[minor rampart]

thanks king, preciated it

[civic horizon]

no problem! you're welcome!

[frosty vapor]

Do we know what model does Deep Research use? is it a custom one based on 5.4?

[green summit]

From OpenAI’s public docs, Deep Research is not described as a custom GPT 5.4 variant. The clearest official wording says ChatGPT Deep Research is powered by a version, or early version, of OpenAI o3 optimized for web browsing and data analysis. 

For the API, OpenAI now exposes dedicated Deep Research models called o3-deep-research and o4-mini-deep-research, which makes the model family look even more like a separate research optimized branch rather than just '5.4 with a mode switch'. 

OpenAI’s newer GPT 5.4 page does say GPT 5.4 Thinking is strong at web research style tasks, but that page does not say Deep Research itself runs on GPT 5.4. 

So the safest answer is:

No, not publicly.

There is no official statement that Deep Research is based on GPT 5.4.

The public evidence points to o3 lineage for ChatGPT Deep Research, plus separate Deep Research models in the API. 


A clean reply you could send would be:

As far as OpenAI has publicly said, Deep Research is not described as a custom GPT 5.4 model. Their docs say it’s powered by a version of OpenAI o3 optimized for web browsing and data analysis. In the API, they also expose separate models like o3-deep-research and o4-mini-deep-research. So based on the public info, it looks more like its own research-optimized model line than just 5.4 underneath.

                    –GPT in response to your exact message.

[frosty vapor]

Yeah that's what it told me too. I asked here just in case there is someone with some info that is not directly public

I assume that time it identified itself as 5.4 in a DR was the system prompt

[glad coral]

hey need help, my archived chats i can't open up. im getting the Unable to load conversation message

i can't even unarchive them