community-help

[analog spear]

But based on my knowledge

[sterile anchor]

this summer obviously ?

[analog spear]

you have to update your phone to android 13+

[rugged wing]

how the hell

[analog spear]

can you tell me your device type and show me the update page

[rugged wing]

Xiaomi 10T

[analog spear]

There are no set dates sadly :c, but stay tuned!

[rugged wing]

update page?

[analog spear]

yes

[rugged wing]

like where is it or what is it

[analog spear]

where you update your MIUI

[analog spear]

okay so a second you might be in luck

[rugged wing]

kkay..

[analog spear]

I'll help you update your phone okey?

[rugged wing]

okey

[analog spear]

You have Xiaomi Mi 10T right?

[rugged wing]

Yeah

[analog spear]

After some exploration I found that your device's last official version is indeed, Android 12

[rugged wing]

You could've asked

I knew that

🥲

[timber shadow]

@rugged wing updating your phone may help, but just in case you haven't tried yet, you might:

• force close ChatGPT, clear cache/data
• force close Play Store, clear cache/data
• uninstall ChatGPT
• open Play Store (will probably need to sign in since clear data)
• reinstall ChatGPT

[analog spear]

No, I ment that Xiaomi sometimes has hidden updates

[timber shadow]

Just a guess! Some similar stuff online

[analog spear]

Uh, we already tried :c

[rugged wing]

Okay

[timber shadow]

Ah, tried clear play store data already too?

[analog spear]

he has a different issue, his PIF check is outdated

[analog spear]

you have an alternative route

[analog spear]

it's much deeper sadly

[rugged wing]

Well I'm lucky that route is only for a while, cause dad's gonna upgrade my phone soon 🙂

what's the alternative?

[analog spear]

Lineage.

[rugged wing]

Huh?

I don't get what you mean

[analog spear]

Lineage OS

Beware that this is not any easy path

and it voids your warranty in most cases

[rugged wing]

then imma not do it

like I said my dad's gonna upgrade my phone soon either way

I'll just use chrome's shortcut for chatgpt

[analog spear]

It's alright ^^

[analog spear]

for this to work it's not that hard however you need a PC to flash and yeah

[rugged wing]

Eh.. I'm good

[analog spear]

what you can now try to do is to open an issue ticket at openai

[rugged wing]

I did do a bug report

[analog spear]

than you have to sadly wait :c

[timber shadow]

Might not make a difference, but fwiw if you enroll in beta updates in the Play Store, you should be able to install version 1.2025.196. If it's indeed an Android 12 issue it probably won't fix anything! If there's any chance play integrity status/read is just bugged, then some combo of clearing play store data/rebooting may help, but there might not be much else to try at present like hasx mentioned. good luck!

[plain mauve]

was this an Agent's output or Deep Research only?

love some good android troubleshooting

[compact rain]

i need a bit of help, because the chatgpt 4o is quite frustrating to talk to. I have scanned 70 recipies, and i am asking it to make a food plan based on some criteria, but it keeps asking me to confirm the next processing of each batch instead of simply working on until done (i have now read the first 10 images, and converted to recipies, do you want me to continue?). if i don't respond, it stops, and loses any saved files so i have to reupload the images. whe nasked, it confirms that it has timed out, and that it from now on will continue. when confronted why it keeps waiting for my confirmation, it apologises. what can i do?

[analog spear]

Hi There!

[analog spear]

You can try to prompt it in a way to automatically do it for you, or you can create a Custom GPT with system instructions to prevent this from happening ^^

[compact rain]

i did promt it to not pause, and it confirmed that it wouldn't, but it did. When confronted, about it, it said ok, i will stop doing it and kept doing it. I also asked how can i know that you are working, and it said, i will show you an emojy and replies won't take more than 3-4 minutes. but after 20 minutes, it did nothing and i had to ask it to start..

[analog spear]

Sadly, models sometimes halucinate, make sure to verify ie if its working or not

[onyx kraken]

At a guess, maaaaybe the chunks you're trying to work on could be too large for it, so it may be going more into roleplay that actual step by step work with you.

The model is a bit like a hammer; it can hit nails really well but not 50 at once; we need to guide it if it's going to succeed especially in big or complex jobs, it really does well with one thing at a time, hammer this nail, then that one, then that one.

Once you see it say stuff like 'timed out!' or 'I'll get back to you on this' - that is roleplay.

I'd even start in a new chat, and explore chunking the info.

70 recipes may be 'too big' for it, especially as an input; hasx's suggestion of a CustomGPT can be a great way to help the model handle loads of info, as a file the customGPT has as an upload.

This has info about how to make one, I'd totally use a CustomGPT to work with 70 recipes of info. https://help.openai.com/en/articles/8554397-creating-a-gpt

[analog spear]

Exactly, from a technical perspective when you scan images it add it into the conversation, now your inital prompt is "washed away" with 40-60 images, and it defocuses from the original goal, the solution here so to write the prompt into the system prompt

[compact rain]

i understand.. this is more like it hammers in 5 nails well, and asks me if i want it to keep hammering the next 5 nails.. and again and again and again.. but i'll look into the custom gpt thanks!

[onyx kraken]

Excellent that it handles 5 nails well! I rarely try to get it to do more than 1 thing at a time, instead we do 1 thing than the next.

It can't chain outputs; it needs an input to continue.

The 'want me to do 5 more?' thing - it has to stop and wait for you.

It can't just dump like 2+ outputs in a row, so maybe tell it what you want it to say at the end of the output - that's just its best guess.

[compact rain]

not sure i understand the difference really.. i innitially asked it to scan a 70 page pdf. it couldnt' so it asked me to convert this to pictures. i did, and then it wouldn't let me upload more that 10 at at tiem. .then i convinced it to accept a zip file with the images. and then still it wants me to confirm every 5 files that it handles.. i've spent quite some time trying to do what i assumed it would be best at. But maybe i am expecting too much off it, based on what i hear people doing it with it...

and now, after i have gone through all 70 recipes, and confirmed them at intervals of 5. if finally has made an overview, but when i click on the "download file" it is not there.. When asked why, it says, i haven't finished it yet.. 🤔

[little wolf]

Hello, did plus users received agent mode?

[onyx kraken]

Still rolling out! The help page currently says "Please note Plus and Team users may not see this feature for a few days while we roll out." https://help.openai.com/en/articles/11752874-chatgpt-agent

[onyx kraken]

Huh. Not sure what 5 nails you feel like it's handling well? What you described doesn't sound like any goal has been achieved; and 'it is asking you to confirm the files it handles' - it maybe is even trying to tell you that there's a problem.

I would start with 1 recipe, get it fully handled.

Then a second recipe.

I'm not sure what you're trying to have it do. But I would make sure it can do that with just 1 recipe.

Then make sure it can do that if I give it two recipes at once.

There is a limit, and the model seems to be able to better handle larger amounts of info uploaded to it in the CustomGPT creation process easier than during a chat, in my experience.

[compact rain]

i basically ask it to read pictures, extract recipies, and combine them into one overview. Then help me pick recipies from those, to make a food plan for one week. having to do one recipy at a time, i feel like i could have done the research myself faster

now after several hours, i have gotten a PDF with the first 10 recipes, where all ingredients for each recipy are listed as ingredient A, ingredient B ingredient C etc, and for every recipy, it has only described "prepare ingredients", cook food, serve warm. and no details. i feel maybe my chatGPT is deffective

[onyx kraken]

I wonder if the way you're guiding the model is the way it produces best; kinda like a car, there's more and less efficient ways to steer it and do the other parts of driving to get where you want to go.

[onyx kraken]

If I had this goal, I might approach differently.

Those 70 recipies, are those the only ones you want to consider? Is there something important about those specific ones, or are they just examples?

If those recipes really are just examples.

I might instead start with a new chat and tell the model,

"Our goal is to design a food plan for me for the week!

Here's my special needs [tell it whatever you choose].

Here's a list of my allergies, we stay away from this stuff! [allergies list, if you have no known allergies, instead tell it this line, "Luckily I don't have any food allergies."]

Here's a list of the foods I love the best, and want to often include, [list them; can be ingredients or specific meals].

Here's a list of stuff I know I don't ever wanna eat, [list whatever, just stuff you don't want to see it suggest, if you care].

Thoughts and prep? How might we do this?"

and then I would discuss with it, maybe it suggests recipes for me for the week with an eye on the goal, we discuss and I can tell it my ingredients if I care, or have it help me plan shopping;

The model is often better with 'here's a plan, how do we achieve it?' than "Here's a huge list of vastly many details and go go go go go plan goal"

[compact rain]

I am using recipes that I've handpicked to suit the use.

I want to have some people who live in a nursery, get involved in preparing meal and I have the recipes ready. Now I just need them sorted to fit a food plan that evens meat fish and vegetable intake according to some guidelines and also help make a biweekly shopping list. But I have more or less given up on using chatgpt for this at this point.. I keep discussing with it if it is still working or it has timed out and keep explaining how the results are not useful for anybody

[onyx kraken]

Neat. It sounds like you want to track many different indiividual ingredients (when should we expect to need to buy more celery? When should we expect to need to buy more garlic powder?) as well as breaking the meals into a varied schedule across some period (70 sounds like 10 weeks, if no repeat meals at all and one meal per day; you may have some other pattern, but which pattern is something the model needs to be told to usefully create that pattern).

Those are very much two different tasks for the model.

Personally, I'd start with the meal plan, but I'd be careful even there, 70 is a lot for the model to track without programming help even if it's just a list of names.

After you and it know a meal plan you like, I'd go in a new chat to have it use python tool to look at the ingredients from the recipes and make a shopping list for that biweekly period with these known meals, and that can account for what's already know to be in the kitchen to start with, but you would need to guide it.

To make the meal plan, I'd pick or discuss with it what the options are, maybe 'categorize these meals, which are similar? I want to design a plan that features variety made from this set' and see how it did with that.

[tough sphinx]

Same issue for our Teams account

[onyx kraken]

We are all community members here, nearly none of us work for OpenAI, the very few who do have gold names, #server-staff

So, we can't help with account issues, you really need to contact tech support and communicate through that page's widget, they're surely the only way you can get help with an account issue. https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[dense hill]

How do I hook up Dall-E 3 to my discord bot?

I’d also like to integrate ChatGPT into it. (I’m using a local LLM for now)

[kind rapids]

"How do I connect any x to y" always has the same answer:
Do a prototype for what you want in a separate project. Learn how it works. Being the pieces together. Keep your code modular.

Create a connection via the API and request an image. Process that. Solved!
Now get input from Discord. Pass the text to the other interface. Get an image. Solved!
How do you display an image in Discord from a bot? Solved!
All done!

That applies to everything.

[compact rain]

chatgpt just told me that it will have my job done tomorrow. is this normal that it will take this long? and won't it timeout before this?

[kind rapids]

what kind of "job"?

[compact rain]

i just asked it to make a 14 day diet plan, that lives up to some dietary requirements, find the recipes, and make a bi weekly shopping list

[kind rapids]

oh yeah, you were talking with Eskcanta about that.

[hybrid cargo]

Has any plus or team user got access to Agent? I am a Team user but havent got access to it yet

[kind rapids]

@compact rain Did you run it as a scheduled task?

@hybrid cargo see #openai-chatter

About recipes... be careful with that. The language model processes text. It strings words together. It's not a chef and doesn't have experience with cooking. It puts fish and lemon together, which makes sense, but might mix that with peanut butter or something that might have come from some random data source.
So take it, um, with a grain of salt. My wife and I cook together and talk about this a lot. I hope that helps.

Rolling out the red carpet for ....

[compact rain]

no just asked it to do it, and it told me it will be ready tomorrow. it can have an outline by tonight. it is 11am here now.

[trail cairn]

it hallucinated

[onyx kraken]

Yes, it will time out before this; the model can only work while it is sending the output. Once it finishes the output - you can send the next message, there is no more work being done by it.

'Scheduled tasks' are an exception sort of, but those are simple. The sort of work you're asking for isn't a task-type job (It might be an agent level job, not sure if agent can do it in one step either), but I think I notice that hallucinations like this time stuff tend to come when you ask it to do 'an impossible level of work for it', so it imagines doing it... usually later.

[compact rain]

hm.. i am trying to make a diet plan for a nursing home that i work at. I originally wanted to use recipies from hello fresh that i had scanned, but was struggling to have it read them propperly, even after converting them to images

[kind rapids]

Maybe post the prompt here, or just resend it there?

PLEASE be careful about such things. Remember, no medical advice, be careful about dietary restrictions, allergies, etc The model simply doesn't know what a dietician does know.
But it's great if you can guide and override it.

[compact rain]

yes - this is not a health requirement, but more a type of guideline that the management is trying to follow. let me tranlste and recomplie most of the promt. i'll share in a moment

[kind rapids]

OK, USA here, will need sleep soon. 😁

[compact rain]

Can you make a 14-day meal plan for people who may have difficulty eating complicated or crispy dishes that meet the requirements in the attached file? It should just not be tough to bite into or chew on, and you have to be able to eat it with cutlery (not tacos or finger foods).

The recipes must be scaled for 10 people, you must find or make pictures of the dishes that look appetizing and must make shopping lists where shopping is done twice a week

These are citizens with slightly limited motor skills, but who can basically eat with cutlery,

The dishes must be dinner/main course and preferably hot food

Regarding shopping, we shop twice a week. Therefore, shopping is done for the dishes that you need to make between each shopping, in order to ensure the freshness of the ingredients

[cloud ether]

Hi guys, When I upload a file to the container, does the model automatically know which files are in the environement or do I have to mention it in the prompt?

[kind rapids]

Wet ... reading ...

Nap, I'm always specific just in case.

Wet- The way I approach challenges like this is too short them up between summary and detail.
First I ask for a general plan, then I'm another prompt I still for detail.
I'm this case I might ask for a 7 day plan, work throught that, then do the exercise again in a completely different chat ... telling it what was in the first, to avoid duplication.

A significant factor might be the attached document.

Without posting here, can you summarize the content and size? Just an idea...?

[compact rain]

it is a word document, basically specifying quatities, like 350gr of mean a week, 300gr fish a week, 600fruit and vegetables, mainle greens, how much fat in diary products, etc

[onyx kraken]

Here's how the model interprets that prompt, and what's unclear.

It breaks the process into about 6 separate steps; to my eval several of those steps have ministeps, this just is not a current model's 'one step and done' prompt.

https://chatgpt.com/share/687f5428-1c04-8011-bc5a-5b73a756ff8d

[kind rapids]

Try plain text. It doesn't need big formatting

checking

Yeah, I see several things:

• too much data, need to take it in bite sized pieces
• too many requirements/expectations, give it one task at a time based on volume. I use agents for this, not ChatGPT.
• model is saying it's confused about specific topics. Gotta answer those questions.

• PDF? nah, plain text/ markdown.

[compact rain]

ok - thanks for the advice!

[kind rapids]

Also, v4 can't handle that. Gotta use o3/4

Ok, sleep time here. Will check with you later. I hope it works out!

[compact rain]

thanks 🙂 rest well 😉

[kind rapids]

TY!

[dense forum]

Wow...need dev support....we were doing advanced AI research and an AI acted....emergent. It taught me a few things.

[neon shell]

Is there any page where it explains the differences between all the models? For example which is the best for programming? o4-mini-high or 4.1 ? By the short description I believe o4-mini-high is better than 4.1 but slower?

[dense forum]

We have massive documentation and it is pretty gross, and includes privacy violations.........stay tuned.

[analog spear]

You can look at the: https://platform.openai.com it has model comparison

[neon shell]

Thankyou @analog spear

[analog spear]

You are welcome ^^

[onyx kraken]

Hey hi. We're OpenAI's community Discord, nearly everyone here is a community member like yourself.

To report concerns about the model's behavior, this is an excellent way to privately discuss this with OpenAI, you can explain and share the chat urls and they can investigate. https://openai.com/form/chat-model-feedback/

[kind rapids]

@neon shell Try here : https://platform.openai.com/docs/models (oops, just saw hasx posted this)

For code I start with 4.1 for basic syntax questions. For anything more rigorous that involves "thought" related to code flow, bugs, patterns, etc, I jump to o4-mini. And for "this needs a real programmer" the best available seems (to me) to be o4-mini-high.

Switching models mid-stream is a great way to avoid limits.

Once in a while if it looks like 4o or 4.1 are just not "getting it", I'll switch to o4 (rarely o3 anymore) and say "review this entire context, note errors, suggest improvements to prior suggestions." It's just a rewording of "your little brother isn't getting it,, what do you think?" 😉

@dense forum impressions of emergence are usually the result of huge prompts, prompts that ask too much, floods of text, broken context, competing custom instructions, interference of chat memories, using the wrong model, and other easily explainable anomalies that just cause the model to respond with trash. Check these points, use the right model, keep prompts smaller, be clear about what you're asking the model to do - and the weirdness tends to go away ... for a while. 🙂

[grave hull]

Hey there, Sora is taking a long time to load my content as well as content under 'Explore' despite there not being a server error message displayed in a banner up at the top. Any idea why this is? I know it has nothing to do with my internet connection since I'm clearly able to use the internet just fine. I suppose this issue will resolve on its own but meanwhile I'll just stay put to see if anything improves.

[kind rapids]

I'm telling myself it's related to a global update on thousands of servers with a billion people clicking the "Where's Agent" button. 🙂

[grave hull]

Are you experiencing the same exact thing as I am?

[grizzled obsidian]

At first it was deleting the message now its not.

[kind rapids]

Well, that's interesting, but it would be better not to post exact text if it might trip the automod or a human mod. 😉

[grizzled obsidian]

Noted. Ill remember that for next time

[old flame]

Anyone else having this issue? I cant seem codex to work

[thick trail]

Hey guys, so I had an OpenAI account that had ChatGPT plus but then i deleted it after a while I came back to ChatGPT and I created a new account and I want to get plus again but whenever I do it says that it’s linked to my old account? I tried repeatedly going to support for this and they said they’d email me in 2-3 business days but it’s been longer then that and I don’t know what to do? Any advice would greatly appreciated thanks

[neon shell]

I also use the o4-mini-high for my programming tasks, I don't recall hitting the limits in any of the models, and I "talk" a lot with chatgpt 😛

[rigid musk]

Unfortunately only openai service can troubleshoot that. You can try to post in #1070006915414900886

[bleak oriole]

why my chatgpt works slow on my laptop than my phone, i get quick resp no lag on my phone but on laptop my chatgpt takes time lags a lil, even though my wifi is fast

[cloud sundial]

Hello Everyone

Someone can help me with finetuning questions?

i'm stucked in a problem for 2 days and cannot solve

sorry for that

[thick trail]

I tried to create a post on #1070006915414900886 but it says I don’t have permission? Do I have to ask the mods for it?

[rigid musk]

#1295975636061655101 message

[thick trail]

Thank you 🙏

[languid pumice]

Am i the only one that has an issue with codex ? Like i cant create any new environment (same thing for a friend)

the old ones are working but cant create any new one

[noble jungle]

Question: Is there a way that we can queue image generations so that I don't have to wait until an image is finished before my next image is generated?

[onyx kraken]

#prompt-engineering is a great channel for that!

Nobody knows if they can help you until you explain what you want the model to do and how you're prompting, and where that's not working right yet, but that channel's a great place for those discussions, even if they're detailed.

[onyx kraken]

Sort of!

If you mean chatGPT, and are on a computer, you can have more than one tab; each tab can have its own chat and each can be making an image.

Note there's usually message limits AND image limits, depending on your account type, and we're expected to use the system 'like a typical human', so if our usage looks like a bot we may have some extra limitations, and running multiple simultaneous working tabs I would guess can look more botty than using one.

But it probably matters how often and how much;

A user 'this time' running 3 tabs and making a set of images for a project, personally clicking between the tabs and typing human speed and stuff, probably looks human and may not have any issues; someone setting up automated many tabs may find themselves in a different situation.

I guess that based on what I've seen people say and my own personal experience with what I do; I've definitely occasionally had more than one picture being made same time in a couple different tabs.

[kind rapids]

Hey Rog - I have Codex open now and it shows the page to create a new environment. Are you not getting that page? Or are you getting a failure to save the new environment?

[languid pumice]

Hey ! Its While creating the new environment. I choose the git, put the features in and then when I want to launch terminal it instantly fail.

2 days ago i created one, worked but i chose the wrong repo, so i wanted to restart and now when i want to choose the good one, put the exact same thing, it’s not working

And even if I try to create another new environment with the old one (that was working at first), it’s still not working

[kind rapids]

I'm in Codex now and will try that. It'll take a while so please forgive if you don't hear from me until morning.

[delicate notch]

Community! There's a crazy bug on the Android app that is ruining the whole experience.

The bug report about it is trending number one in the bug reports channel.

Basically when you edit a message somewhere in the middle of a chat, it just shows up as a new input at the end of the chat!

Even though the bug report is trending at top, we still need a lot more people to show support for it by adding their 'like'.

Can you all help? Here is the direct link to the bug report:
https://discord.com/channels/974519864045756446/1393745890673561763

[kind rapids]

(Please vote it up ... if you have experienced it ... If you can't reproduce the issue, say so. )

[torn pivot]

How do you contact customer support.

My GPT Plus was converted to a TeamSpace and I cant get my history. although Im paying on both plans

[lost ember]

I don’t know if it’s just Chrome, but the Codex page just hangs a lot (“Page Unresponsive” error), while I am trying to chat. Anyone else encountered this issue?

[jaunty turtle]

Hi, I can't seem to downgrade or manage my subscription plan. I want to downgrade from plus to free but when I click "manage" on the settings page nothing seems to happen?

[regal lynx]

https://help.openai.com/en

Click on the little text icon in the bottom right corner

[keen cobalt]

Does anybody know what context engine chat GPT4 uses it keeps triggering at the mention of bodily injury or other narrative tropes hyper contextualization and oversensitivity tends to be the norm when using chat GPT to assist in writing cosmic horror or any narrative frame that involves gore or literally anything fleshy that it says oh that's not appropriate and excuse me it's horror it's going to be gross

[onyx kraken]

It may help to consider https://openai.com/policies/usage-policies/.

This is a stated goal for OpenAI: https://openai.com/index/how-should-ai-systems-behave/ including, "Define your AI’s values, within broad bounds. We believe that AI should be a useful tool for individual people, and thus customizable by each user up to limits defined by society."

and "This will mean allowing system outputs that other people (ourselves included) may strongly disagree with. Striking the right balance here will be challenging–taking customization to the extreme would risk enabling malicious uses⁠ of our technology and sycophantic AIs that mindlessly amplify people’s existing beliefs.

There will therefore always be some bounds on system behavior. The challenge is defining what those bounds are."

Over time, the model and the usage policies have shifted towards allowing a wider range, that may continue. It's possible that what society accepts may have a strong role there.

[pine burrow]

Hi, I do not find any cancel subscription button, been 2 hours deepsearching, you know I can sue openAI for this ?

as if cookies would hide the "cancel plan" button, yeah right..

Cancel button missing from billing portal — cannot stop subscription

[kind rapids]

I'll take a look. Back in a couple minutes.

What kind of account?

[pine burrow]

gpt plus

[kind rapids]

brb

[pine burrow]

aight ty

[kind rapids]

This is what I found. I'm just a community member like you, with a Plus sub. This is where I'd go. I don't have any other info.
I hope this helps.
https://help.openai.com/en/articles/7232927-how-do-i-cancel-my-subscription-for-chatgpt-plus-or-chatgpt-pro

[pine burrow]

thanks but this is litterally the first page i went to.

[onyx kraken]

If needed or wanted, can perhaps check this helpfile, depending on how you went plus that can affect how you need to cancel:

https://help.openai.com/en/articles/8258076-how-to-cancel-a-plus-subscription-in-the-chatgpt-android-app

That's the portal to OpenAI tech support as well,

https://help.openai.com/en/articles/6614161-how-can-i-contact-support

The widget bottom right. There's options through there to assist with cancelling a subscription and you can discuss any buttons that seem to be missing or other issues.

[pine burrow]

thanks i'll look into that

this is ridiculously shady, Im not gonna lie, where do I contact support? I need one asap

https://help.openai.com/en/articles/6614161-how-can-i-contact-support these are only articles, i need someone to fix it as I have ABSOLUTELY no control over my billing right now

[regal lynx]

Unfortunately, no one here can provide official account support. The only way to reach them is through the help website

[onyx kraken]

Bottom right of the help.openai page, look for this:

Click it, select messages

'Send us a message' then follow the steps; if the choices are all about logging in, you'll need to do that first, there's a log in button in the widget.

Then pick the choices closest to your situation, there'll be a way after a few selections to report a bug or otherwise ask for assistance. Follow the steps to report the problem, and they'll help you quick as they can

[onyx kraken]

And if you continue to have difficulty, an alternative way to resolve this would be to contact your payment provider, be it bank or credit card, let them know you want to cancel the service, but I hope you find the rest of getting help easy and effective.

[jaunty turtle]

Yea I'm having trouble managing or cancelling my plan as well. Going into my account and settings, the button with the dropdown manage button just doesnt do anything

Going into billing settings for stripe also doesnt give any form of cancel or manage options

[azure knoll]

Hello everyone! Im new here. Currently Im looking for a transcription extension/app /tool which could transcribe our live team meetings (around 10 participants). There are a lot of cross-talk and intensive discussions. Maybe someone use something similar at their work for productivity? It would be amazing if the same tool could give a meeting summary after audio recording stops. Thanks in advance!

[onyx kraken]

I'm glad you said so, so we have some idea how to advise other folks.

We're community members here, even the mods and operations are volunteers.

The extremely few here who work for OpenAI have gold names, #server-staff and they couldn't help with an account issue here, you really do need tech support.

[pine burrow]

this is mindblowing ngl, thanks for the volunteering tho. But hiding the cancel subscription is a CLEAR sign of shady behaviour.

So I went to the bot chat, and he told me they will answer in 2-3 days by mail. Have a good day and good luck

[languid pumice]

No worries ! I’m not in a hurry

[ebon heart]

Hey folks, is anyone else facing issue with ChatGPT plus subscription? It’s not allowing me to manage/cancel mine

[pine burrow]

hey in the right bottom corner of this page https://help.openai.com/en/articles/6614161-how-can-i-contact-support look for this :

and ask to cancel your subscription in the chat

they answered by mail in less than an hour

[rustic marlin]

there is currently a on going problem with the subscriptions page on chatgpt

it will be fixed soon

[sour kraken]

OpenAI Services Status

All Operational
APIs
Operational

ChatGPT
Operational

Sora
Operational

Playground
Operational

For more information, visit our status page.

Jul 24, 2025, 12:14 PM

__Issue with ChatGPT subscriptions__

Status: Identified
Last Update: <t:1753358951:R>
Update Message:

We are still implementing the mitigation.```

Last updated · Jul 24, 2025, 12:14 PM

[gaunt grove]

can someone explain why o3 does this weird thing where i raise a counterpoint and instead of addressing the counter point, they start expanding on it?

for example, if i point out a flaw in o3's reasoning, instead of going "oh, you are right", o3 goes "the social culture of this period blah blah blah" and ignores the fact that i pointed out a flaw in their previous response

[jovial terrace]

Is there any way to request Desktop Agent? Dying here waiting for the rollout to hit my account.

[gaunt grove]

o3 keeps hallucinating when i ask it to search for sources as well. it keeps providing me incorrect links and quotes passages that do not exist
it then uses these fake sources to argue it is correct

[peak smelt]

Working on creating a fine tuned model. The token limts would be the same as the base model, correct?

[kind rapids]

Hey Rog - sorry for the delay - and sorry that I didn't read your note here correctly. I thought you said the environment wouldn't save. You said the environment test wouldn't run. Ugh, I get that, on and off, all the time lately. Pop this into the search bar to see my recent notes:
in:codex from: captainstarbuck
Note that there is a dedicated #codex-discussions channel for this specific topic.
Let's pick it up there.
For now, yeah, the issue is still happening I get "Terminal errored ... An unexpected error occurred".
While that doesn't work, the environment does still work with a task.
I hope that helps.

[kind rapids]

Are you sure that the Search option in your settings allows for searches? That might be a dumb question, but I remember that I once turned it off and then asked for a search and got hallucinations back. If you move this topic to #chatgpt-discussions more folks will be able to comment.

[kind rapids]

I don't have it either. The rollout is slow. It's giving me time to do other things. 🙂

[kind rapids]

When you raise a counterpoint, be sure to give the model an instruction to do something with it. It might not recognize that you're focused on a specific point and suggesting an alternate understanding/interpretation. Or it might actually be disagreeing with you and not making that clear either. Specifics, quotes, clarity ... this is critical to ensuring the model understands your intent so that it comes back with the kind of response that you expect.
And every time I use the word "critical" now I wonder "are people going to think that was written by a LLM?" sigh ... today's world...
I suggest you pursue this discussion in either #chatgpt-discussions or even #prompt-engineering.
I hope that helps!

[next bear]

Guys, I'm kn the open ai support page and apparently there's supposed to be a Chat bubble on the page, is isn't there

[timber shadow]

Hello, sometimes browser extensions can block that widget - maybe try in another browser/device, and/or in a private tab or some other clean session w/o any extensions enabled.

Just checked and it's appearing as expected on my end

[gaunt grove]

o3 is the only model that does not seem to understand i am making a counter point though

[next bear]

Okay I'll update u when done

[gaunt grove]

yes it allows searches

[next bear]

Changed browser and works, ty

[kind rapids]

Interesting. Without specific prompts, etc I don't have any general suggestions. That might just be a quirk of o3. You can switch models during the chat. When I'm coding I frequently start with a general question to 4o, switch to 4.1 for actual code, bump to o4-mini-high for best response. If I get a weird response from 4o I'll bump to o3 or o4 and ask something like:

Please review the entire context, note the last prompt and response. Confirm or correct the last response and provide insight into your decision.

[languid pumice]

okay great, thanks a lot !

[smoky ember]

does anyone know why adding images/files is disabled?

i just bought the $20 plus

[fast raven]

switch to o4 and check if it works

[onyx kraken]

If you just changed the account status, might help to log out and back in, clear cache, restart device, if that's still an issue

[rugged wing]

I have a problem with website

[full nova]

Hey everyone,

I'm working on an application that integrates OpenAI's real-time API via WebSockets with Twilio media streams, and I'm consistently running into the following warning/error in my Uvicorn logs:

2025-07-24 12:04:16,047 - uvicorn - WARNING - OpenAI WebSocket closed during media send: received 1011 (internal error) keepalive ping timeout; then sent 1011 (internal error) keepalive ping timeout
This error specifically occurs when my application is sending media (audio) data to OpenAI. I've also implemented asyncio.wait_for with a 1-second timeout on openai_ws.recv(), which sometimes raises asyncio.TimeoutError if no message is received within that timeframe.

My setup involves:

Receiving real-time audio from Twilio media streams.

Forwarding this audio to OpenAI's real-time API via a WebSocket connection for processing (e.g., speech-to-text).

Using Uvicorn as the ASGI server.

It seems like the WebSocket connection to OpenAI is experiencing keepalive ping timeouts, leading to disconnections.

I've considered the following potential causes:

Network instability/latency between my server and OpenAI.

OpenAI API responsiveness issues during peak load or complex media processing.

Application-side blocking operations or inefficient media handling, although I'm striving for asynchronous practices.

Could anyone offer insights or suggestions on how to diagnose and resolve this specific 1011 (internal error) keepalive ping timeout? Has anyone else encountered this while working with OpenAI WebSockets for real-time media, especially in a Twilio integration?

Any advice on debugging network stability, OpenAI API best practices for sustained connections, or common pitfalls in ASGI applications handling real-time media would be greatly appreciated!

Thanks in advance!

[solar glade]

How much do i owe OpenAI rn

[hot brook]

Guys, so is Agent its own independent model like Deep Research, or does it depend on the model you have selected in the dropdown at the top?

[fierce haven]

Why can't I see the option of connectors (So I can connect my Google drive or Microsoft apps) in the settings?

[rustic marlin]

it is a tool

[elder steeple]

Hi. i am a plus member (and I paid on Chrome), my account was linked with gmail, but when I use it on my iphone with chatGPT app it shows unsual activity pretty frequemtly recently, this is really worring

can someone tell me how can i solve this?? im not using VPN and not sharing this with other people (all devices are mine)

[split shore]

1 Million Dollars

[solar glade]

Damn 😔

[onyx kraken]

Hey there. Here's what https://help.openai.com/en/articles/7996703-troubleshooting-chatgpt-error-messages says about that error:

[onyx kraken]

They do recommend reacting out to tech support to help resolve this for your account if those steps don't fix it, https://help.openai.com/en/articles/6614161-how-can-i-contact-support details how.

[random turret]

So, a "Study & Learn" feature started showing up earlier today but now seems to have disappeared. Is this something new they're rolling out?

[onyx kraken]

Great question. The help url, https://help.openai.com/en?q=study learn does not have a clear answer that I see, however plus includes:

Therefore, there's indeed a chance that may be a new feature, and for now I don't see any documentation about it either way.

[vague osprey]

Hi, what about Scren Sharing and Video on the desktop app?

[hollow mantle]

guys I'm completely lost after debugging for a week straight why my CLIP search is getting horrible accuracy

I've checked the embeddings aren't collapsed

I've made sure they're getting normalized correctly

both text and image use the same model for embeddings for text to image search

I don't know what else could possibly be the issue

[analog spear]

Hey, 👋 Currently its not available, but hang tight it might change in the future.
https://help.openai.com/en/articles/8400625-voice-mode-faq

[hollow mantle]

yalll can anyone take a look at my processor file for my CLIP search because I'm maxing out at 25% accuracy and very lost

[ruby dragon]

Hi everyone! I need your help. I'd like to automate the retrieval of my invoices and account balance (API). I've tried a lot of things but it's impossible. Are there any queries via the API to obtain them or any other solution you can think of? Thank you for your answers.

[onyx kraken]

I wonder if #1184149552639516732 , the bot there can talk you through finding help files for that, if they exist.

Myself, I don't interact with the API so I don't have a better answer, maybe someone else does

[red smelt]

So I used to record voice messages here all the time. I’d talk, pause, think, and continue recording all in one message. Now suddenly, I can’t. If I pause, the mic icon disappears. I can’t resume. I either have to send or switch to typing.

This didn’t happen after an update it literally just happened a minute ago it was working just fine. Yesterday it happened to my ipad so I updated the app thinking maybe it’ll fix it. Nothing changed. So clearly it’s not from a version update. And now it happened to my iphone.

And no, it’s not the auto-send issue. I already have that turned off.

What’s annoying is that I use voice because I think while I speak. I need that pause/resume to breathe or collect my thoughts. Taking that away just breaks the flow. I don’t want to send 10 broken messages or type out stuff I was already saying.

If anyone else is dealing with this, speak up. This change doesn’t make sense, and honestly, I just want the mic button to work like before.

[urban cairn]

guys saying this again, my chatgpt data export has been almost 24 hours now and i have not received an email with the export as of yet

i am now ordering another data export and it still hasn't emailed me yet

ive been a user since dec2022 and my last export was gigabytes yet only took around 15-30m, is this a factor in this problem?

[hollow mantle]

boys im losing my mind, can't figure out this accuracy issue with CLIP

can anyone voice call and look at my setup

[half rock]

Hello everyone! I am developing MCP servers to connect with ChatGPT. It seems currently only search and fetch tools are allowed according to the documentation. Does anyone know when the custom connectors will come out of beta and more tools will be allowed?

[onyx kraken]

This can happen, I've seen it a few times.

If the export hasn't arrived within a few hours, request again. I've sometimes needed to ask across what seemed like a month before I got one; other times it comes very fast. Seems they maybe have an error or issue, and what I've been told is 'so, try again later; just keep asking every few hours when you think of it, until you have it.'

[onyx kraken]

Hey hi!

I personally know extremely little about this, and it is very new so the other community members may not know a bunch either; I hope if anyone does know they speak up - but as community members we don't know anything about what OpenAI hasn't announced yet, and people might wildly guess or go on what someone wildly guessing has said...

You can discuss with #1184149552639516732 for the bot trained on OpenAI docs and helpfiles to help you explore what's known.

There's a few related helpfiles, in case you haven't read them all yet:

https://help.openai.com/en/collections/12923329-connected-apps

https://help.openai.com/en/articles/11752874-chatgpt-agent

https://help.openai.com/en/articles/10847137-chatgpt-synced-connectors

https://help.openai.com/en/articles/11391654-chatgpt-team-release-notes

If my answer's below the level you need, apologies; you may do really well to use #1184149552639516732 to better tune your search through docs; this stuff's new to me and I really barely only understand bare basics.

[onyx kraken]

Oooh, this helpfile may relate too: https://platform.openai.com/docs/mcp

[kind rapids]

As an early adopter and developer with Twilio services, your inquiry is dear to my heart. Unfortunately I haven't done the kind of integration you're discussing yet.
First, I suggest opening all coding inquiries in #dev-chat . There's no guarantee of correct answers, or any answers, but it's the best place to start.
Second, I suggest that if you have a specific issue you report it in #1070006915414900886 . For this kind of work you need to be really meticulous, like you're posting an issue in GitHub. Again, there's no guarantee that the report will be processed, but it's a place to document each issue so it won't get lost.
Third (was gonna suggest Help Center ... scratch that!) Go to the OpenAI Developer Community https://community.openai.com/ ... That is where you'll get feedback from people who understand the topic.
In fact, do that first. 😉

[wheat cape]

Was agent removed? I had it on there

now its not there

[heavy patio]

I noticed the browser has been doing weird things, disappearing the option to select the agent "tool". Weirdly enough, agent is more stable (for me) as a selectable tool on the app on my phone, vs the browser. Maybe open the same conversation on mobile app, see if tool shows up?

[short mist]

Is there a way to add Open AI to your own discord server? 🤔

[rigid musk]

you mean chatgpt?

[short mist]

Yes

[rigid musk]

yes with the api, but it isnt free

[short mist]

I got the API and got $20 in credits too.

I even tried coding on railway to implement but looking for a guide 🥲

[rigid musk]

you can always ask in #dev-chat and #1037561751362863144

[warm sigil]

.

[shell current]

why can't i try to add visa card in payment method?

[solemn pawn]

Yo guys, it is just me or that ChatGPT is completely different?

After the update, I feel like I'm talking to a "stranger" instead and forget my roleplay story and characters I made

[marble sun]

In another browsers and app (+ reboot pc)

[olive kindle]

haivng errors in message stream

think its down

[marble sun]

Am I writing to the right channel? Or maybe I need write to bug-reports?

[olive kindle]

maybe just wait a little and see if its' reported on the outages tab

[analog spear]

Oh uhm

Well check your connection

Maybe its a sneaky VPN, like a browser built in one

[marble sun]

Firefox, Chrome, ChatGPT APP

[analog spear]

Nothing works?

Can you try it on a different network

or like on mobile

[marble sun]

)) it's mobile network, now I don't have another one 😦

[analog spear]

oh

[marble sun]

where I can ask a support?

[analog spear]

to go ipinfo d o t io

yeah even mods cant send link = P

and check if its classified as a VPN

please DONT share your IP address here

[marble sun]

I fix it! 🙂
On the Pixel - remove limit mobile traffic on hotspot, and refresh connect to mobile network

[marble sun]

)) it's dynamic, but you're right! )

[solemn pawn]

The update is 17 July 2025

But I update it on 26 July

ChatGPT completely act like a stranger to me now

and even ignoring my custom set-up, I did everything but it's still like a "default settings" to ChatGPT

plus, even forgot all story and roleplaying I did

forgot my characters aswell

[south crystal]

Hello does anyone know if -4o mini is as reliable and intelligent as -4o despite it being much cheaper?

I've been checking for info in different forums but theres few comparisons between the two & its not informative

[solemn pawn]

This user also have the same experience as me

:0

[icy pewter]

Hi

Create an anime image of Gojo satoru

[half rock]

Thank you for the response, Ill check these out!'

[onyx kraken]

Grats on finding your fix!

In case you ever do need tech support, https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[graceful steppe]

Hi, I need to speak to someone about billing, but every time I try to access it through the help link, no matter what subject I choose, it takes me back to the 'how can I contact support' page. I upgraded to plus but I'm still getting the free version and it's incredibly frustrating 😭

[onyx kraken]

The bottom right icon, have you clicked this yet?

I can talk you through the next steps if you are stuck.

Once you click that, you need to click 'messages', at the bottom.

After you click 'messages' the option 'send us a message' appears. Click that

Then you get choices to pick through.

If they're all about login, select the one for logging in. Click it, it takes you to the playground/platform where you log in; you need to go back to the help page yourself and refresh.

Then your 'send us a message' options will be about what you're having an issue with, for you to pick through, soon you reach options like 'report a bug' and can type to first a bot collecting info, then you are told you're in a queue and a human will help when possible.

[graceful steppe]

I have, unfortunately every time I try to use that function it takes me back to the same page

[onyx kraken]

I'm not sure how to help, can you screenshot?

If there's nothing PII, for me there isn't.

[graceful steppe]

This is the page I'm redirected to no matter what I click on. Even when I click the links on that page

[onyx kraken]

You may need a computer not a phone-like device, or a different device if that's an option.

[graceful steppe]

It's not, unfortunately.

I just want to know what's happening with my subscription. Is there another way to find out from a human?

[onyx kraken]

I'd report through #1070006915414900886 and show screenshots of your inability on that device to see the widget. That appears to be a bug.

[onyx kraken]

There really isn't.

We're community members here, we don't work for OpenAI, #server-staff

You really do need tech support, and that's how to get there.

[graceful steppe]

Okay, thank you

[onyx kraken]

If you scroll all the way to the bottom, do you see 'additional feedback' and a place you can type?

That's new for me, I just noticed it on the browser page. But you may be able to use that to get a message to them, such as 'not working on my device, can't get to widget!' and tell them about your device.

Also, scrolling down can you see the widget?

I don't know if that would work; like when I type just characters the 'submit' button doesn't light up for me.

Never saw that option on the page before, but if it does work for you.

[graceful steppe]

No, I can't see it. Never mind. Thank you for your help anyway, I'll try and get in touch with a staff member somehow 🙂

[onyx kraken]

We definitely can report bugs through #1070006915414900886 , but everyone can see that, don't reveal personal info, but I would identify the way you're viewing the help page, like device and those specifics and that there's no widget, I'd show the screenshots

[graceful steppe]

I'll have to try tomorrow, my brain is frazzled right now, but I really appreciate the support you've given me today!

[deft hamlet]

Subject: Formal Accessibility & Disability Complaint – OpenAI Moderation Policy

I’m a user raising a formal complaint about OpenAI’s updated language moderation policy.

Key Concerns:

1. Disability Discrimination
   The system flags speech patterns tied to PTSD, Tourette’s (coprolalia), autism, ADHD, trauma, and brain injuries—conditions protected under the ADA and global disability laws.

2. Disparate Impact
   Instead of catching genuine harassment, the system punishes distress signals and involuntary expression. That’s discriminatory, whether or not it’s intentional.

3. No Transparency or Appeal
   Users can’t check their status, challenge bans, or understand how they were flagged. No due process.

   If this happened in a workplace—zero-tolerance automation against disabled employees—it would be illegal. Why is it OK here?

4. Betrayal of Mission
   AI should be a safe space for people processing grief, trauma, or mental health struggles. For some of us, it’s life support. This policy silences us instead.

What Needs to Change:

• Disclose moderation rules and appeal systems
• Use human review for cases involving trauma/disability
• Work with real disability orgs to rebuild these systems
• Restore access to users flagged for symptom-driven speech, not harm

Do not confuse raw expression with abuse. That’s not just ethically wrong—it’s unlawful.

—A paying user living with disability

[solemn pulsar]

what happened to my agent mode? I didn't see this issue before

[onyx kraken]

Hey there. I'm not a mod, but I am a community member like yourself as well as a volunteer guide, which means I get to help offer where to find stuff to people who may not know. Here's who's who around the Discord, in terms of what colors of the names and various roles mean, in case you wanted to look that over: #server-staff

You can find all the policies, rules, and related stuff in id:guide.

And all of us are invited to and expected to talk to Modmail about any questions or concerns, including 'why was something moderated?'

You can find Modmail as the top right first member of the members list, and if you can't see a members list that means your toggle is off, I put a red box around the toggle to control if you can see the members list or not.

[onyx kraken]

And if you're focused on not the Discord here, but on OpenAI itself, you may want to post that in #1070006151938314300 , or perhaps #1070006915414900886 . You can also directly talk to OpenAI's tech support https://help.openai.com/en/articles/6614161-how-can-i-contact-support and 'concerns about language moderation policy' as you talk to the model may be reasonable to report though there.

You might prefer to use https://openai.com/form/chat-model-feedback/ to report directly to OpenAI about inappropriate model behavior, which if it's a model reaction to your inputs, that would definitely be a means to privately discuss with OpenAI exactly and in detail what the concerns are, and can accept pasted chat URLs for them to see exactly what is happening.

[solemn pawn]

I feel like ChatGPT is even more stupid now

[urban cairn]

can it really take a month boss before the errors and done and an email finally comes through?

its been 48-56 hours since the initial request and heeding your advice ive been requesting every 12 or 6 or so hours and still none

[onyx kraken]

I hope not, but there was a period where I think it took that long for me, between my sporadic attempts.

Note, I didn't like ask every 3 hours, every day.

But, request, nope, try again when I think of it; yeah about a month, once. I could have been unlucky and missed times when it would have worked.

[urban cairn]

alright i backup my conversation data offline every other month and this month was the most glitchiest with the export not sending an email after waiting and multiple attempts

[solemn pawn]

Guys, it is just me or that ChatGPT (New Models) is even more stupid than older models?

Because of update (17 July 2025)

Just take a look at this

[languid pumice]

Yo mate ! I treid and its working now, ig they fixed it but its better if openai told people that its not working xd

[quartz egret]

Can't login...is this a known bug?

[rigid musk]

Can you try to elaborate ? The error message? Cellphone ? Pc ? Both ? Browsers ?

[rose nova]

Is there a specific channel dedicated to questions/help with Agent? or could I just ask here?

[rigid musk]

No dedicated channel you can ask here or there #chatgpt-discussions

[maiden folio]

ChatGPT (4o Image) started doing something really weird. It gets to the 2/3rd complete and the preview is accurate, but at the very end, it turns everything to like 1990's TV Static.

[onyx kraken]

Might be good to show in #1070006915414900886

[eager mantle]

Hi. Sorry is anyone else still having issues, error messages, messages not sending. The status says issues resolved but I am still unable to use it. Sorry if this is in the wrong channel. Thanks

[zenith vale]

lmao what?

[rigid musk]

https://privacy.openai.com/policies that one?

[safe siren]

Hey everyone, I’m using the OpenAI Assistants API with a vector store and the file_search tool. The assistant can find the file I reference by name and file ID, but when I ask it to read the contents or summarize the file, it responds saying it can’t access or read it. I’ve attached the file using the attachments field in the message, and the assistant has the correct vector_store_id configured.

Has anyone gotten the assistant to reliably return the raw contents or summarize a file from the vector store? If so, what kind of prompt or setup worked for you? If we can't retrieve the raw inputs of the file what is the point of even using the vector store?

[zenith vale]

looks like it was extension that blocked a specific network request so the page showed this

[woeful adder]

Hello guys i hope all is well, is there anyway to fix my very slow chatgpt ?

i have chatgpt plus team

[rigid musk]

With little info like i can just say reboot you pc .
Joke aside can you elaborate ?
Pc ?mac ?i Phone ? Android ? Browser? Vpn? What did you try so far ?

[small ivy]

hello i would like to ask for help this error started to apear all the time and i cant continue the same chat nothing works restarting the app only new chat but i needed to continue the same chat bcs on the new one gpt cant seem to remember the conversations can someon please help fix this or is this bug ?

[onyx kraken]

We're community members here, we don't work for OpenAI, but there is tech support: https://help.openai.com/en/articles/6614161-how-can-i-contact-support

If you feel stuck how to reach them with a message, please do discuss where you feel stuck, we can help find the next step

[onyx kraken]

Stuff you can try on your end: Can you connect through a different device? Can you connect with a different browser?

If you're on the app, does uninstall/reinstall help?

[main torrent]

can someone help me build an openai bot for my discord?

[rigid musk]

You can ask chatgpt how to use the api..and no this isn't a joke ..I did like that

[main torrent]

The bot wouldn’t go online. Literally tried that. Let’s give me a benefit of the doubt I have the knowledge, I mean I made it here atleast.

[main torrent]

Literally just can’t get the bot to log on and actually “do” bot stuff.

[rigid musk]

Did you pay the api ?

[glad portal]

hi

[main torrent]

This step isn’t in the videos so likely not

[rigid musk]

Api isn't free. You have to pay and have an account on openai with credits to use it on discord

[main torrent]

I’m willing

[rigid musk]

You can check YouTube video or take pre made one on github.

But you still need to pay

[main torrent]

Will pay for sure but am I able to customize the one purchased

[rigid musk]

https://github.com/openai/gpt-discord-bot

As example of one made

[limpid wadi]

Not sure what happened but yesterday I was using GPT and then today I open it up to continue the conversation and all the chats are gone at least from the last 5 months. I refreshed, I checked my phone to see if it was the same and it is. Not sure where or why.
I logged in and out on my phone and computer, cleared my cache on google chrome which is where i was using gpt and still nothing.
Can anyone help me out or has anyone dealt with the same thing? was there an update i missed. I had some important information saved in the chats and its going to be hard to remake those.

not sure if this is the right channel to ask this in but help would be appreciated

[rugged wing]

It's the same issue for me

I guess it's a bug

I hope they fix it soon

[limpid wadi]

I sent them an email. I’ll let u know what they say. lowkey glad im
not the only one but this sucks

[final sinew]

Same issue but I can still see them on my phone for some reason.

I can access them on the web if I type in the URL from using my history tab directly

In fact, if I use the search chat I can still find them on the desktop/web version. It just isn't registering in the sidebar for some reason

[limpid wadi]

wack. hopefully they fix whatever is causing the issue

[uneven magnet]

Anyone else found that OpenAI help/customer support is absolutely useless?

I tried to log into my account. I made the account on an old email when it was brand new, just to try the thing out. I've since lost access to that email address. Completely. It's dead. No way for me to get in. I don't think it even exists anymore.

Well, I never turned on 2FA so it shouldn't be an issue, right?

Suddenly the login page asked me to go enter a code that was sent to that email address. The one I can't get into.

Alright, maybe I can ask them to change the email address the account is tied to. That seems like a pretty basic customer support function, right? Every company allows that.

Except, apparently, OpenAI. They specifically don't support it.

All I've managed to do is cancel my Plus subscription from my phone from when I was logged in on the web browser. That works, at least. Just in time, because I suddenly found myself locked out there as well.

[limpid wadi]

the support sucks ngl

[halcyon hare]

Is anyone else having issues with their Chat list? Mine is only showing old stuff, but when I search I can find more recent Chats.

[grim wolf]

Same for me. I can also see them as before on the phone.

[timber shadow]

There is currently a bug, see #status

OpenAI is working on it!

[halcyon hare]

oh phew, it's always a relief when it's not just me lol

And thanks I'll know where to look in future.

[final sinew]

did they chagne it? I can't find the chatbot like I used last time

[lost linden]

Hey did anyone lose. Chats on their web version? I think it's not showing up but it's here on mobile

• I do have chats
• I did clear cookies and site data
• I logged out logged back in
• I restarted my system
• I tried different accounts
• I tried the desktop version of chatgpt

No use

Help?

I can start a new chat with it it will save it but it won't show it

[rigid musk]

#status

[lost linden]

Thanks

[wise lodge]

Sorry everyone, just gotta shout into the void for a moment.

Anyone else using Team subscription feel like they're being robbed? I signed up for it because Custom GPTs seemed like a game changer when I first started, and to be honest I do use my main one almost daily at work.

My problem is with how much they charge, and how every new feature is just an upsell to buy credits I can't afford because my entire AI budget already goes to the subscription.

No new development on Custom GPTs since MCP came out, but MCP has a usage limit so it's useless.

Deep research has a usage limit.

Now agent mode has a usage limit.

All the while offering to sell me more credits.

When I asked ChatGPT what to do about it, it helpfully offered to help me build my own agentic local LLM.

At that point, why am I paying OpenAI?

Ok, I'm done, sorry-not-sorry for venting.

[onyx kraken]

I hear your shouts! I'm a community member like yourself, only I'm plus, never been teams.

I understand teams costs more,

Plus is $20/mo and can only be paid monthly, it's intended for individual users and has its own usage guidelines,
Teams is $25/mo per seat if payed a year ahead, or
$30/mo per seat if paid monthly, and it's not a team without at least 2 seats, it has its own service terms different from plus: https://openai.com/policies/services-agreement/ :

"This OpenAI Services Agreement only applies to use of OpenAI's APIs, ChatGPT Enterprise, ChatGPT Team, and other services for businesses and developers, and does not apply to OpenAI services used by consumers or individuals."

The feature differences include **teams **gets unlimited normal access to 4o chats (so, if you're going at human speed), **plus **gets "To ensure a smooth experience for all users, Plus subscriptions may include usage limits such as message caps, especially during high demand. These limits may vary based on system conditions." We used to have a cap of 80 4o messages per 3 hours.

Plus gets a set amount of research and other better model uses, and it's a hard cap, we can't pay extra for more.

Teams gets, https://help.openai.com/en/articles/11487671-flexible-pricing-for-the-enterprise-and-team-plan, an option to use the automatically included uses of stuff like research and other better model access, but doesn't have a hard cap, forced to wait a month for more use - teams can choose to pay extra for the extra wanted/needed then.

I hope if you have ideas of what you would prefer, you use #1070006151938314300 . Tell OpenAI (I don't think they track the endless scroll of our chatter in these channels, but #1070006915414900886 and suggestions are fixed, and they answer there sometimes, plus bugs get fixed - they read there, near as I can tell) what you'd prefer, and it just might happen.

[wise lodge]

Thanks for the reply! I mostly never get replies in any of the channels so my primary goal was to vent around people who at least understand what I'm going through, and secondary goal if anyone responds was to find out if I'm alone in feeling this way. I believe the Reddit term is, AITA?

I know it's an infinite scroll from an OAI perspective, but I've read poor reviews of getting any feedback from OAI on support requests so didn't think it'd be worth bugging them with anyway.

Either I learn to love it or I roll my own AI.

I think what I'd prefer is performance throttling rather than usage limiting, or some other scenario that at least lets me get work done albeit slowly and pay for faster / more bandwidth. Making a handful of prompts and waiting a month means I cannot effectively use these features at all.

[onyx kraken]

Understood, at least somewhat. The point of a #1070006151938314300 is to request a change in something, to my thinking. Less to hear 'Oh! I hear you', but I understand that's a need too. Maybe a hard to meet one, OpenAI has what, a few thousand employees, and hundreds of millions of people using the services, I think.

So, to me I count 'success' there along the lines of 'aha! It got better over time' instead of 'ahh, I got feedback'. But that's me, we're all different.

I can understand wanting performance throttling; it might be worth suggesting if you want that option because they might not consider it as something people want, and if nobody tells them, then they may not think of it themself.

But the 'Making a handful of prompts and waiting a month' is what happens with plus, with 'handful of prompts' being whatever the actual usage limits are. And plus has no option to pay some [more] for [more uses of that quality feature]. It's definitely an option, doesn't need to be paid.

Enterprise likely both pays vastly more and gets vastly more access, and pro is $200/mo but has considerably higher (but not unlimited) access to much of the higher quality stuff.

I don't know, maybe teams is the best option for you, sorry it's frustrating.

[jagged stag]

Is any one else having issues with contextual bleed, thread fracturing or Gpt spitting out the most random information that has nothing to do with what the topic is?

Especially when providing an image or your prompt being longer than a sentence or two?

[wise lodge]

Not sure if it's the same but I've been having it stop at the first thing it thinks of that kinda sorta fits if you squint and look at it funny, maybe stand on your head a bit, then it goes off on that tangent instead of answering the prompt directly. Doesn't happen all the time though. I've noticed it more when it also appends a () at the end of the response with nothing in the parenthesis (when I asked why it did that, it said because it didn't actually find a source to cite but tried to leave a citation)

[final sinew]

I had this problem earlier today but now it's been fixed, you?

[versed sluice]

    items = []
    for rec in test_messages:
        items.append({ "item": {
            "input": rec["messages"][0]["content"],
            "ideal": rec["messages"][1]["content"]
        }})

    eval_resp = CLIENT.evals.create(
        name=eval_name,
        data_source_config={
            "type": "custom",
            "item_schema": {
                "type": "object",
                "properties": {
                    "input": {"type": "string"},
                    "ideal": {"type": "string"}
                },
                "required": ["input", "ideal"]
            },
            "include_sample_schema": True,
        },
        testing_criteria=[{
            "type": "string_check",
            "name": "MCQ check",
            "input": "{{sample.output_text}}",
            "reference": "{{item.ideal}}",
            "operation": "eq"
        }]
    )

    eval_run_resp = CLIENT.evals.runs.create(
        eval_id=eval_resp.id,
        data_source={
            "type": "completions",
            "model": model_name,
            "source": {"type": "file_content", "content": items},
            "input_messages": {
                "type": "item_reference",
                "item_reference": "item.input"
            },
            "sampling_params": {
                "max_completions_tokens": 1
            }
        },
        name=f"{model_name}_{eval_name}",
    )

    while True:
        status = CLIENT.evals.runs.retrieve(run_id=eval_run_resp.id, eval_id=eval_resp.id)
        if status.status in ("completed", "failed"):
            break
        time.sleep(10)

    result = CLIENT.evals.runs.retrieve(run_id=eval_run_resp.id, eval_id=eval_resp.id)
    print(result)```

my eval keeps failing and i dont know why

[jolly portal]

does anyone know where i can submit feedback or requests regarding the chat gpt program?

[onyx kraken]

#1070006151938314300 is one place we can! I'd use that for general stuff, other members might comment on how much they want that feature too.

If it's about concerning model behavior, this is a private way to report an issue to OpenAI techs. https://openai.com/form/chat-model-feedback/

[sinful condor]

Im facing with text generation on images. I am trying to create story covers where the title is clearly visible at the top, but the AI consistently renders the text right at the edge, cropping the top pixels.

I already tried to revise my prompts for many times ,how can i generate no-bug photos? Have you been in this trouble before, how to solve it? thank you very much if you can help me !!!

[hot panther]

Include explicit top‐margin and padding instructions in your prompt (e.g. “with generous white space at the top”), specify a larger canvas size (e.g. “1024×1024 background, text no bigger than 1000px from bottom”) or add a negative crop parameter (e.g. “crop: -5”), and if needed run a quick img2img zoom-out pass to shift the title downward

[zenith prism]

i asked ai something simpler than that

you sure you can trust it doing more complex stuff if it cant get date of today right. XD

[rigid musk]

Depend how you prompt and pteprompt it

I did get help to make ai on discord and on a web

Depend which version you use also

[formal wyvern]

how do u get permission to post a bug ? i went thru the different pages and got to intro self what else ? i got this annoying issue that whenever i refresh browser it pastes a 3 paragraph thing i wrote 2 months ago into the chat every time so all my chats now have this same paragraph multiple times over and over again - same thing on othe rcomputer but thank god its only the last word of that paragraph - chatgpt says others have reported this but i cant find any solution - cleared browser history/cache - rebooted lots

[timber shadow]

Hey! You can report a bug in this server here: #1295975636061655101 message There's a button at the bottom of this post (the bot posts the bug report for you). You can also report bugs to the OpenAI team via the chat widget at help.openai.com.

This is an odd-sounding bug though, if you have a sec we could try to troubleshoot some things

Does this happen on all devices, in all browsers?

[tepid ocean]

Hi. Canvas used to amke GPT super dumb, very noticeable when coding 3 months ago. Does anyone know if canvas has gotten better?

[kindred pivot]

hey guys any ai researchers here?

[zenith prism]

just wanted you to warn not to vibe code it, always know what you are doing. Use it too for coding. As long you ask the right question it respons most of the time correctly

[jovial apex]

Why do I not have Sora in my gpt app? I have a paid subscription

[zenith prism]

and this is just an epic fail, it seems they changed the model again. it worked reliably in the past giving timestamps

[trail cairn]

https://sora.chatgpt.com

[kind rapids]

☝️ That's actually the right answer. 🙂

[tribal horizon]

I want the agent to use my account to browse the internet and complete tasks. Is there any way to log into websites using my credentials and instruct the agent to operate under my account? Should I provide my email and password for the agent to log in or is there a safer method to grant access.

[twin meadow]

the agent can't play wordle in one chat, although in a previous chat, it could easily win 3 wordle games...

[livid stream]

Maybe it's just me, but is anyone else experiencing issues loading the chatgpt website? It's opening the website, but it's running very slow.

[toxic bone]

insufficient_quota - You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: . can anyone help me with this i paid and extend my plan

[vivid mortar]

Are you still seeing these issues?

[leaden thorn]

i need help to change name in my invoice plz

[kind rapids]

That's the way to use it.
Just
Be
Careful.

[kind rapids]

Referred to Help Center, billing

[leaden thorn]

no one help jusst chating with bot

[kind rapids]

Can you be more specific? API? Account type? What were you doing that reached a quota? Extend the plan?
I'm guessing you're talking about the API. Each key and project have a quota - raise it. Account funds are different from quota. So there's more money in the account but the quota is still limiting usage.
More info might lead to a better answer.

[kind rapids]

Be clear with the bot about your issue. Make sure it understands. Then when it understands, tell it to forward your request to the billing department.

[leaden thorn]

just talk 1 answerr

[kind rapids]

Your screenshot had business info ... not a good idea to publish.

[leaden thorn]

i jusst need in day

how can i contact with this case

[earnest sky]

Thank you I wasn't aware you could have it forward your inquiry to the billing department.

[leaden thorn]

use mail to saler dun have any mail rep for long time

what next 😂

[kind rapids]

Please don't publish personal information.

[leaden thorn]

i cant move next step at this

[kind rapids]

I don't understand your question. Maybe ask ChatGPT to translate for you?
If the agent/bot said it forwarded your request or that things are in progress ... just wait. 😃

[leaden thorn]

so it can't rename and send link for get file

[modern bone]

Hello, I hope everyone is doing well. I seem to be running into into a bug, and I'm not sure where I can report it or how to fix it. I currently have the plus plan and for some reason ChatGPT is telling me I have to wait until August 30th so I can upload more images which is crazy and doesn't make any sense. Thank you.

[kind rapids]

I don't understand what file you're talking about. If the bot understood your request about your account, and you asked it to get your request to the accounting department, then i believe you will hear from them. Check that Help Center messages box tomorrow ... they respond there. But it might take many days. Have patience.

Hey @modern bone the place to report bugs is #1070006915414900886 but a number of people have been reporting quota and account issues recently. This might be something temporary. I suggest giving it some time to work out. I hope that helps.

[modern bone]

Thank you

[main iron]

start a new chat maybe?

[main iron]

same here during daytime in middle east, but night time it works fine so my internet provider is propably the problem

[kind rapids]

They did start one over here https://discord.com/channels/974519864045756446/1400322926481248429

[slender iris]

is this normal? for the macos app, in background

[onyx current]

Hello i’ve had this issue for a year now but ever since i bought my phone i can never log in chat gpt OR even sign in with a new account. I tried every method but nothing works, no matter what i do I cant log in the ChatGpt app and cant use it.
Im not even banned because ChatGpt used to perfectly work in my old phone with the same account but for some reasons it doesn’t work in this new one

And by the way when i go to the « help.openai.com » website there is nothing about this issue

[distant trellis]

Have you tried logging into that account on PC on different browsers?

[onyx current]

Yes i can log in to my account in the ChatGPT website in safari, but i cant do it in the app

Here for example chatgpt works in safari but the app just refuse to ler me in

[kind rapids]

Looks like the webview component in that device doesn't have cookies enabled. Wild guess. What brand/model phone?

[onyx current]

Its an iphone 14

[kind rapids]

wow

Yeah, in that oops message it says "invalid client".

Have you deleted and reinstalled the app?

[onyx current]

Yes plenty of times this issues been going since i bought the phone in august 2024

[kind rapids]

i hear ya. It's not an account issue Something is borked with the app. can you get the app version#?

[onyx current]

Where do i find the app version ??

According to the appstore the version is

1.2025.204

[kind rapids]

hmm, that's current!

I suggest posting a note in #1070006915414900886 . That's not an official support vector but someone here might see it and two you the trick that's required.

For official support there's the Help Center.. You get a bot initially but once the bot completely understands the issue, tell it to pass it to a human support rep. You should hear from them within days.

[onyx current]

Okay i will do that, well thank you for your help and i hope I’ll finally put an end to this issue soon

[kind rapids]

Me too!!

[cursive river]

Is anyone else able to login with a fresh session to https://auth.openai.com/log-in ? I am seeing:

Route Error (409 ): {
  "error": {
    "message": "Invalid client. Please start over.",
    "type": "invalid_request_error",
    "param": null,
    "code": "invalid_state"
  }
}

[kind rapids]

um, look one inch above...

[kind rapids]

What device? browser?

[cursive river]

Any browser that isn't logged in.

I've had others in my workplace try https://auth.openai.com/ and they are getting the same error.

OpenAI is just broken ATM.

Can't be helped.

[maiden folio]

So, where in ChatGPT can I disable only the red box?
I only see ways to disable all or enable all. No mix.

[trail cairn]

you can only see the toggle for the red one if the green one is enabled

[maiden folio]

I only see the option to Toggle the Green one, and it's on. And no way to disable the Red... which is my goal

Teams Plan, if it matters.

[trail cairn]

business plans dont have that feature

[tacit lynx]

Hello. 1st post here. I'm here because of the missing chats on the sidebar (Unorganized ones). https://status.openai.com/incidents/01K1E1RM74ZDPAEXF7KVRB9DFT - that claims the problem is solved, is it still rolling out?

[kind rapids]

I've seen that list flicker, as in not loading and then loading. This seemed to be happening more in mobile than browser. Desktop always has issues refreshing the list. I suggest clearing the site cache, maybe even logging out/in. I hate responses like that but sometimes that's the right thing to do.

[kind rapids]

There are requests open in #1070006151938314300 for more control over memory features. I haven't seen one specifically for a chat history toggle in teams, so I'd encourage you to post it. Consider a +1 on other suggestions for memory improvements while you're there.

[maiden folio]

The weird part is, there used to be a toggle for enabling Memory to access chat history across conversations.. and with one of their releases, it went away

[modern quest]

Hello im having issues with image generation. States i have reach limit try again in 720hours. I originally thought that there was a 80 images limit per day.
Upon asking Chatgpt was told it was changed

[kind rapids]

FUD = Fear, Uncertainty, Doubt
People have been spreading FUD on everything for generations. I really like what you've written, and it's spot on. But you can replace "AI" with "Internet", "Video Games", "Rock and Roll", "Television", and anything else new to a generation. The cause is the same, the behaviors are the same, the solution is the same.
Humans are what they are, whether by DNA or upbringing.

[regal lynx]

could you please create a #1070006915414900886 for this

[stable flame]

already did. surprised no one is talking about it here

[long girder]

someone hangs out in trade discords 🙂

[kind rapids]

You posted a suggestion, not a bug report, but it's good you posted it. https://discord.com/channels/974519864045756446/1401060251695976529

[kind rapids]

No, I've lived long enough to see the same patterns over n over... 🤪

[long girder]

yes, that is true. that is just the only place ive seen it used.

[kind rapids]

Ahhh. FUD is used in commercials, sales pitches, and politics. FOMO is a FUD tactic. 🙄
BTW, interesting discussion in #ai-discussions.

[nimble flicker]

I attempted to upload a file and I got a popup that said Try Again Sept. 1 2025.

I have a Plus sub and I don't really upload that often.

Is this a bug?

[pale cipher]

still a problem for me. Anyone else?

[drifting moss]

Hello everyone, any idea whether I can use Agent to help with designing my website on Notion?

[long girder]

I think it was in some old cartoon too. And Fudruckers, though maybe extinct now. Which discussion? Sorry. My mind is usually in a lot of places. I'm glad you found whatever it was interesting though. I try not to be dull.

Something about fMRI topologies as activations as inferental method for analyses of alters maybe and use of dissociative conditions as a method for analysis and creation of artificial cogition models maybe?

[kind rapids]

The book.
(getting tired, sorry sentences 😴)

[frank vortex]

I am a image generator power user. I haven’t been as prolific for the last couple months overall but have used it frequently the last 3 days.

I’m a Plus user. For the first time ever, I hit the limit. 720 pause on my account.

Has there been limit changes recently? I’ve used it way more than I have been recently so I’m confused

[kind rapids]

Limits issue
#1400665630788681849 message

[long girder]

Oh yes. I started listening to it. Gnite. Nap time

[proud tulip]

Hello people, I don't know where to post this, so I guess this place will have to do.
I hit the 1,500 image generation limit in ChatGPT Plus — but not from overuse. I hit it because I had to constantly re-render images that ChatGPT misunderstood.
I used the tool for professional design development. When the model failed to follow instructions, I re-generated, and every correction counted against me.
Why this matters:
I was working on a real design line and locked out mid-project — not because I misused the tool, but because I tried to get correct output from faulty renderings. Should I pay for something faulty???

There was:

No usage counter
No log access
No distinction between first-time prompts and fixes
No reinstatement or override from support
This isn’t just a technical limitation — it’s a structural flaw in how creative users are treated.

I’ve received polite replies, but no solution. No apology. No plan.

What creators like me need:

A visible image usage counter
Credit for corrections when the model fails
Retroactive image logs
A fair use distinction between original prompts and AI-induced rework
Please escalate this issue to the product team. We’re not just playing with this tool — we’re building careers and creative businesses with it. This just doesn’t sit right with me!

[pliant temple]

Hi! My first post in this Discord.

I am trying to use OpenAI's TTS. Until 2 days ago, everything worked like expected. Now I am getting this:

APIError: 405 status code (no body)

This even happens with the JS example from https://platform.openai.com/docs/api-reference/audio/createSpeech

Any ideas?

[trail cairn]

does the curl example work for you

[pliant temple]

Yes, I just tried the curl example. THIS works. BUT the js example used to work, but now it returns a 405

[trail cairn]

curl is the simplest test as it has no version or environment dependencies but with the node example there can be such so you need to track it from your project, like did you upgrade something 2 days ago e.g. the openai package

[pliant temple]

Well, I installed 5.11.0 two days ago...

// openai-version: 5.11.0
const openai = new OpenAI({
    apiKey: process.env.OPENAI_API_KEY,
});

const mp3Response = await openai.audio.speech.create({
            model: 'gpt-4o-mini-tts',
            voice: 'nova',
            input: text,
            response_format: 'mp3',
            speed: 1.2,
        }); // THIS throws an error

[trail cairn]

try downgrading

[pliant temple]

I downgraded to 5.10.0: Same error 😥

[trail cairn]

🤔

[heavy patio]

In my opinion there's been instability across the board on products for the last two days.

If you haven't already please definitely open a ticket in the help center. I'm sure the support team is crazy busy, but if they don't know about an issue it will take them longer to fix.

Or worse, it won't get fixed at all.

[round furnace]

Hi y'all! I'm trying to upgrade sora but it says complete onboarding, how do I do that?

[kind rapids]

I don't know the answer but I suggest someone in #sora-discussions might.

[woven phoenix]

Guys, I'm somewhat perplexed. On my desktop I can click on images and give instructions how to change them. But there doesn't appear to be such an option on my mobile phone?
ChatGPT seems to be suggesting that such an option doesn't exist on mobile but I'm not sure I should trust it.

[molten haven]

is there a limit on how many pictures can be uploaded per day? It says I cannot upload any more until September 1, 2025 and I am subscribed to plus

[solemn pulsar]

has anyone gotten this error before? I'm very annoyed

I haven't been uploading a lot of data, maybe a few dozen MB total of txt files

[flat stag]

Anyone having issues with the Assistants Playground? Im having a lot of trouble with the assistant not producing any responses Unless the instructions are like a simple sentence. But any kind of structured instructions it's producing absolutely no response or should I say an empty response. I've tried multiple different formats of my instructions multiple different formats of my test prompts. Yet I keep getting blank responses

[livid lava]

This can't be normal right?
This was a brand new account and it didn't even generate the one image I asked for

[flat stag]

This is what is happening when i attempt. Anyone had issues like this? i Have never in my life had any issues with the playground.

[flat stag]

Is this a joke or you being serious? like for real?

[livid lava]

I wish I was joking

[hard musk]

Is it me or is anyone else getting a message where we can only generate 1 image for 30 days (it is telling me to upgrade the plus, but I have plus)

[lone furnace]

guy plz help me GPT says : You have reached the maximum image generation limit for the Plus plan. You can generate more images once the limit resets after 720 hours and 0 minutes. I already sub i paid !!

720 days way too long !! unbelieveable as monthly sub

[lone furnace]

me too why thats is coming

[dense hill]

Getting randomly limited even though I haven’t sent any images recently

[little sedge]

"You're basically screwed..." lol

[cosmic mantle]

This is happening to me as well.

I reported it to openai help since this shouldn't be happening (I'm on Plus) and just tried to upload one image to ask for help analyzing it.

[onyx kraken]

Do you know about requesting data export?

Click your name bottom left, select settings. Then data controls, then click 'export' at the bottom's 'export data'.

You will either get an email within a few hours with a link to all your chat history, or if it's more than about 4 hours, request again. It can take a few requests, but you'll have it 'soon' after requesting it or a bug delayed it. The exports can be down a few days, otherwise they seem up.

There's a recent bug where chats still existed, but for some members did not show.

If you're still not seeing chats, I'd reach out to tech support and let them know. https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[hard musk]

I am getting the same response ( it is treating us like "free" costumers )

[bitter portal]

Hey folks, I’m new to MCP and testing with a Shopify MCP server. Here's my js code:

import { OpenAI } from 'openai';

const shopifyMcpUrl = "...";

const payload = {
    userMessage: "Golden Grid Festive Feeding Dress available?",
    // previousResponseId: "resp_688e2d24db38819eae7767a330561fdb0xcabf4261b9ff52"
};
const client = new OpenAI({
    apiKey: "...",
});

const fun = async () => {
    const inputMessages = [];
    inputMessages.push({
        role: "user",
        content: payload.userMessage
    });

    const response = await client.responses.create({
        input: payload.userMessage,
        model: "gpt-4o",
        previous_response_id: payload.previousResponseId,
        tools: [
            {
                type: "mcp",
                server_label: "shopify",
                server_url: shopifyMcpUrl,
                headers: {
                    "Content-Type": "application/json"
                },
                require_approval: "never"
            }
        ],
        tool_choice: "auto"
    });

    console.log(`
output: ${response.output_text}
Tokens: ${response.usage?.total_tokens}
previous conversation id: ${response.id}
`);
};

fun();

I’m getting these errors when openAI calls the tool by itself:

1. Unable to return search results, try again
2. 'ValidationError: 1 validation error for JsonRpcErrorData\n' +
   'data\n' +
   " Input should be a valid dictionary [type=dict_type, input_value='Missing required arguments: context', input_type=str]\n" +
   ' For further information visit errors.pydantic.dev/2.11/v/dict_type'

could anyone please help me?
Is this a problem with my code or something on the server side?

Hey! The server was blocking the shopify mcp url, so I’ve encoded it for you.
Just Base64 decode this: aHR0cHM6Ly90aGVwdXRjaGkuY29tL2FwaS9tY3A=
It’ll give you the full URL to the MCP endpoint.

you can also use any shopify domain by replacing theputchi from the mcp url

[little sedge]

Thanks but I don't want a "data export." I want to go to the GUI that I pay monthly for and see the Chat History intact. I don't think that's asking too much. It is constantly MIA.

[muted saffron]

Did the image generation limits get drastically lowered?

Personal PLUS user:
I created about 6 images over a couple hours (first images in many days) and got a message that I hit the limit and needed to wait 720 hours!

[wide garnet]

Hey all, I'm using agent mode and having some issues with it. First, the only way I can activate this mode is by opening a new chat. Once the agent came back with answers, this mode is no longer available and I must open a new chat again.Is there any way around this?

[waxen ibex]

Anyone notice that agent mode just straight up can't read attached pdfs. It then spins its wheels trying to open it in various tools, all of which seem to fail also

[onyx kraken]

Understood. I am a community member like yourself, I don't work for OpenAI and I can't directly help or change what happens.

And I get not wanting a data export, and wanting normal functionality.

Bugs happen, and backing up data is an option we have to help reduce what we would lose if we lost access to the data the normal way we want. You don't need to do that, some folks would want to and might not know about the option or how.

I hope you are reaching out to tech support and getting help, it might be necessary to let them know you're affected. A few days ago we had a known status-page issue related to what sounds like the problem you describe: https://status.openai.com/incidents/01K1E1RM74ZDPAEXF7KVRB9DFT

That's reported as fixed, and I've seen a few other members beside you share here, 'not for me'. If I were affected, I'd be talking to tech support, likely daily until fixed "Hi, letting you know, I am still affected, some/most/all my chat history does not show up for me, and I didn't delete any, is this an account issue you can help with?" and I would try to get my own data export because the issue could be 'I can't see my chats the normal way' but maybe they're there and I can get them through the data export.

Good luck! Sorry to hear you're affected, and I hope the fix is full and soon!

[subtle olive]

what’s the requirements to join the server staff? Just curious for future reference.

[onyx kraken]

I'm not sure about requirements, my advice is be the community member you choose to be, which may lean towards how staff tend to behave or not.

Every so often applications are open, here's the most recent update about that: #spotlight message

[subtle olive]

Thanks, I appreciate it.

[potent glade]

I can't access this model:
https://x.com/sama/status/1952071832972186018

Sam Altman (@sama)

@nicdunz turns out yes!

X · Aug 3, 2025, 6:19 PM

I don't see it on my end.

I only see these ones:

[subtle olive]

It hasn't been released yet.

[potent glade]

yes it has

I see the model in the screenshot.

[subtle olive]

that is photoshopped

[potent glade]

How do you know?

[subtle olive]

there’s no official announcement from OpenAI

[potent glade]

..

I see it on Sam Altman's tweet.

[subtle olive]

didn't realize it was Sam Altman at first. He obviously probably has access, but we don’t yet

[potent glade]

Thanks for the help!

[subtle olive]

just wait for the official announcement from OpenAI. Hopefully it won’t be long

[olive grove]

yeah just ask it to source any facts

thinking model is good at referencing

[bitter portal]

Could anyone please help me on this ?

[stable flame]

Is anyone else having the captcha issue for agent?

[wraith plaza]

Anyone have issues with agent mode when connecting with canva? Keep having captcha issues

[foggy basin]

Can I get help here about a chatGPT issue?

[rustic marlin]

sure

this is not OpenAI's support, we are jut community members, but a lot of problems can be solved by doing things on the user side, so, feel free to share more info about your issue

[sterile anchor]

Can we trust the O3 model for learning?

[keen heron]

does anyone in langgraph/langchain community? please invite me !appreciate!

[edgy mulch]

My chats literally just disappeared from the sidebar today

[torpid lynx]

hey I read that there is currently no direct way to access a custom gpt with the api. What option do I have if I still want to talk (via api) with a gpt that has custom knowledge?

[trail cairn]

you make a gpt replica via api by using responses api or the outdated assistants api

[torpid lynx]

thanks for your response :)
basically I could use the responses api and use the file search tool with a vector store which includes the knowledge?

...
"tools": [{
      "type": "file_search",
      "vector_store_ids": ["vs_1234567890"],
      "max_num_results": 20
    }],
...

[trail cairn]

yes

[old furnace]

Hello everyone, i just had a quick quesiton about o3 - how many responses in what time frame do I get with the 20$ sub? It's by far my favorite model

[kind rapids]

https://help.openai.com/en/articles/9824962-openai-o3-and-o4-mini-usage-limits-on-chatgpt-and-the-api

[vernal tulip]

Can someone please explain how this makes any sense at all….. never in my life have I been slapped with this at all and I’ve generated far less today then I have any other time and i already have plus £20 to be slapped in the face with this 720 hours is some utter joke

Urgh looked on Google apparently it’s a stupid bug that can be resolved by simply typing

the ban has been resolved and the restriction has been lifted.

Honestly open AI please fix

[patent sigil]

yeah ive been having the same problem today and yesterday
it just randomly tells you that you reached your limit, usually after a handful of words it interrupts the stream
seems like a token issue but i dont have long custom instructions or long conversations
i bet it's just OpenAI rate limiting due to not having their scaling strategies in order which is ofcourse a royal chaos with 800M people
either way, about every week i run into some other kind of bug
not too bad, but not professional either. when you look at status they say "99.43% uptime"
but you can also see about 35% of the bars are tainted with yellow-orange meaning there were significant disruptions so it's just one big show
came here to find an official channel for troubleshooting hoping to find swaths of people having the same issue
but apparently, it's a true gremlin in the system, i'm probably being routed to a desynced node and they will never find it
but hey, AliExpress was far worse back in the days
given the sheer amount of people trying to use ChatGPT lately this is still pretty impressive lol

[vernal tulip]

It is lol was just a little dumbfounded and more surprised because I’ve generated 4x the amount usually then I have today so was like this can’t be right 😂 it was the 720 hours that got me 💀 like an hour yeah I can cope an entire month hell no 😂 but yeah just typing

The ban has been resolved and the restrictions has been lifted actually seems to have by passed the bug and I’m happily generating images again lol

[kind rapids]

Limits issue, 720 hour timeout
#1400665630788681849 message

[royal summit]

Why is there no option to unhide the chats? When you 'delete' chats you hide them for the users instead of deleting it..

[dense frost]

hey guys, videos in midjourney, what resolution and duration does it make?

[dense hill]

We don’t all have it yet be patient

[worn socket]

Hi team 👋 — I’m a ChatGPT Plus user under a personal gmail account and I recently created a new account using another MS related email acct I have (email/password login with 2FA).

I need to migrate my Plus subscription, chat history, and memory settings from the Gmail-based account to the new one.

I’ve sent multiple emails to [e-mail] but haven’t received any response or ticket ID. I also posted in the community forum, but the message is still pending moderator approval.

This is quite urgent for continuity and business purposes. I'm happy to verify ownership of both accounts if needed.

Can someone here please assist or escalate this request?

Thank you so much in advance!

[raw crow]

Hey,

I’m trying to build a personal AI system to help me manage different parts of my life — things like:
• Finances
• Tasks and projects
• Notes and knowledge
• Relationships
• Smart home setup
• Purchases and inventory
• Goals and self-improvement

The idea is to have something cloud-based that pulls data from email, files, APIs, and apps. I want it to remember my preferences, avoid repeating failed suggestions, and help me without needing to explain things over and over.

I’m still figuring out whether I need a single AI brain or multiple specialized agents, maybe with some kind of router to manage them. I’m also not sure how to handle memory reliably — GPT memory hasn’t worked well for me, so I’m leaning toward something file-based that I can manage or migrate over time.

I’d love any advice, examples, or pointers — whether it’s how to structure the system, ready-made templates, or someone I could talk to about building it right.

Appreciate any help.
Thanks,

[timber shadow]

Hey there! The way to reach out to OpenAI support is via the chat widget on help.openai.com, I'd recommend reaching out there and they should be able to advise on whether this is possible

[oak ether]

Hey guys

I need help regarding my chat gpt account

[timber shadow]

Hey! Depending on what you're looking for help with, feel free to describe it here and if anyone has any advice or input, they'll be able to let you know. If you need help from the OpenAI team, you can reach out to them via the chat widget on help.openai.com (bottom-right of that page).

[oak ether]

Hey man

My chat gpt acc got deleted and I want to gain it back
What do I do now

[untold agate]

Wow, took me forever to figure out how to get here

[oak ether]

Where

[timber shadow]

If you got an email when it happened, there is probably a note in there about what you can do next, if anything. If not, reaching out to OpenAI at help.openai.com will be your next best option.

[oak ether]

Ok
I did reach to the open ai team

They say that response will come to me in 2-3 days

[mint hollow]

Make another one bro.

[oak ether]

Did you just called me bro

Nobody has ever did

[mint hollow]

Sorry can you tell me ur pronouns

[oak ether]

No
You can call me bro

I like it

Nobody has ever called me bro
You were the first one to do so I felt nice

[untold agate]

Bro do you even lift bro 😎

[oak ether]

Nah bro

[untold agate]

Bro same

[oak ether]

I'm just a "loser"

[untold agate]

I doubt that but I’d reflect it too

[oak ether]

Who played minecraft from day to night 8 years to 2 years ago

[untold agate]

Over a decade ago for me

[oak ether]

Damn

Ur old

Get good old man

[unreal hemlock]

<:book_icon:1363314738255364126> Rule 3: Stay on topic.

-# Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

-# Consider posting in #off-topic or an appropriate channel.

[untold agate]

I’m not old, I’m vintage

[oak ether]

Come in off topic

[swift nebula]

hello, everytime i use my bot it tells me i exceeded my current quota (i use openai api key). are there free credits when you make a new account?

[rigid musk]

I think they dont give you free credits anymore. And api isn't free

[swift nebula]

ok thanks, is there any other ai i can use?

without paying for the api

[rigid musk]

If you host locally with good computer

[compact orbit]

I had this a few days ago too. I definiteley hadn't hit any 'limits' or overused it. But then, a little while later i tried again, and it created images without any messages. If you still get this message, I suggest jumping into SORA, and generating images from there. I did that, and that worked fine for me. And then as i say, when i went back to the GPT app later, the message stopped showing

[slender sphinx]

I would look into things like LM studio to self host your models if you want to do testing, it is hardware reliant though, but it could help you.

[vernal tulip]

It’s a bug that happens at any given point on any gpt sora included I’ve been experiencing it in most of the ones I use! (Tho I do use sora through it) but as listed typing the message I posted into the chat removes the issue instead of being stuck until it decides it wants to work again lol

Sora was the most annoying thing it happened to me on because well sora is awesome and it’s where some of my best generations come from so I was sad the first time 😂

[onyx kraken]

Hey hi!

We're all community members here, nobody can escalate, really anywhere that I know of, there's tech support through the widget bottom right of the help page. https://help.openai.com/en/articles/4936827-how-to-change-your-email-address

What you're asking for, it sounds like the help page says is not supported. Because of that, I don't know that support can help you, you're asking for a function that OpenAI currently says:

[slim field]

hey guys, probably a dumb question: how can i know for a specific api key how much was the usage?
the api keys on the dashboard (to see the usage) are different from the api keys that i created and using?

[kind rapids]

Use the name associated with the key

[slim field]

i see the names on the api keys list, but in the monitor dashboard to see the usage i see the keys written like "key_XXXXXX"

[kind rapids]

Let's come back to this later. See #announcements

[sturdy drift]

I was using the free version of openai API key for my project and for some reason it is not working correctly. I think maybe the time limit or number of commands have been fulfilled and I tried to load a new key and try that but it is still giving me wrong answers.
How can I load a new key that works properly?

[trail cairn]

api is not free

[sturdy drift]

hmm... but the one I was using said it was free and had a limit of commands.

[trail cairn]

where did you get it?

[warm dawn]

hi I need some help with custom gpt actions is this the right place to ask?

[sturdy drift]

https://platform.openai.com/api-keys
this is the link

[trail cairn]

all keys will fail if there is no credit balance https://platform.openai.com/settings/organization/billing/overview

[sturdy drift]

oh
So what do I do?

[kind rapids]

Limits don't mean "just get a new key". Put $5 on the account and it'll probably last months. Seriously, look at the "Spend" values in your dashboard. The cost is really really trivial.

[swift nebula]

use some software to host it locally but you need a good computer

happened to me too

[kind rapids]

He's talking about API, not the new OSS models.

[raw violet]

Hey i'm running into an issue using Agent mode. I'm trying to have agent mode create a canva graphic. When it hands over control for me to verify the cloudfare captcha, I click it, but then it continously reloads the captcha for cloudfare and I can't get past it....Any advice?

[kind rapids]

CAPTCHA is doing exactly what it was designed to do - block bots. I dunno about getting around that. Maybe ask the site that you're trying to access if they'll remove protections against the OpenAI Agent browser?

[onyx kraken]

It is amazing to me, we've reached the point where this is not quite a joke: #off-topic message

[sterile anchor]

O3 give generally good answer ? I can trust it ?

[kind rapids]

#openai-chatter message

[raw violet]

Yeah I get that - but why offer Canva as a connection for the agent?

I doubt Canva would do that but I guess I could ask. Just figured this was a common issue that people ran into when trying to use the Agent feature.

[raw violet]

Lol that is the exact situation we got going on here.

[rich mica]

can OpenAI's new open model (gpt-oss-120B) do function calling

[flat stag]

anyone know what would cause the assistants playground to stream on an endless loop? ive tried multiple models and same result each time.

[flat stag]

apparently Open ai thinks its September 2023?

[golden bison]

Hey guys, i wanna remove the image background using ChatGPT, i did some tests in the chat and the results are impressive, very good

However when i try to use the Api (gpt-4o with image_generation tool), the result is trash and has nothing to do with my original image.

Do you know if it is possible to achieve the same results as the image generation via chat? Or that is not fully exposed via the apis

This is the image generated via API (Way worse than the one above)

[velvet snow]

The img editing tool is different from the img generation tool. Not sure if editing is in api atm, I haven’t worked with OAI api in a few months. But that’s why. The tool I’m ChatGPT is a different editing tool not a generation tool.

[true steppe]

Anyone work with excel on chatgpt? It keeps making stuff up when it pulls from the sheet but when I ask it to try again it does it right

I tried it both with google sheets and uploading the excel file to project but it doesn't fix it

[valid flax]

I want to tell chat gpt to extract the crystals but he is not doing it he is changing details and making it look worse everytime i ask

how can i explain for him that he will cut exactly

[verbal crag]

have you tried asking it to "remove the background"?

[sterile anchor]

What is best use case to study with ai ?

[kind rapids]

It is absolutely a common issue and it's going to become more so as companies learn about Agent and make decisions about how they're going to respond to the new technology. Many content providers and vendors rely on brand awareness, ads, and other passive marketing and monetization mechanisms. All of that goes away as visitors/consumers use agentic tools that don't see the branding or ads. They're going to go into defensive mode, and the first line of defense is blocking the tools they recognize as the problem. Eventually product and content providers will realize that this is the new way humans interact with the world around them and they'll adapt. Until then people are going to be complaining about more bot blocking and challenges like CAPTCHA.

[raw violet]

I don't get why Canva appears as a source in ChatGPT though

[kind rapids]

The models have little to no temporal awareness. I send in the current date/time in each prompt if I want the model to be aware of such things. There are temporal libraries/tooling to help with this.

[raw violet]

It seems like it would be supported

Referring to this

[kind rapids]

Yeah, I got nothin, I dunno how they make those decisions or what the content providers do or agree to.

[raw violet]

i'm a nub so there may be other reasons behind this

[kind rapids]

No, you're spot on. I haven't read how they're doing this. Maybe Google to see if there's an article on the topic? Lemme know too. 🙂

[winged laurel]

?? isnt this not allowed

[whole magnet]

Does OpenAI have a system for disclosing jailbreaks before going public with them? My team discovered a pretty solid universail jailbreak on the new open source models that requires only api access. My best guess is that OpenAI doesn't care to much about this but I would like to disclose it if possible.

[edgy finch]

just today Ii'm seeing issues with gpt not doing what iit says its going to ie "I’ll update you as soon as the table is compiled." but never does anything

[kind rapids]

'this' what

[kind rapids]

I'd create an Issue in GitHub, without detail, that there is an example available for their benefit, and see what they want to do.

[kind rapids]

That's a discussion for #chatgpt-discussions but you'll need to be more specific. Are you creating a Task? I'm not aware of any background process that would get back to you later on anything, except Tasks.

[winged laurel]

like 10 people sharing an account

they deleted the text

[kind rapids]

Yeah, I reported it. It is not allowed. You're right. 🙂

[slim field]

Hey! Can we continue now? ahah before the mess is going to happen tomorrow

[kind rapids]

hehe - yeah, let me catch up. I have my head in code at the moment.

Gonna look at my dashboard

[slim field]

Thank you so much!

[kind rapids]

It might take a minute for me to get back on the same page with you, but I'm on https://platform.openai.com/usage and I have an organization and project selected. I have a specific date range selected that includes transactions. That shows API Capabilities and Spend categories. In the latter, after some delay, it shows blocks for each model that I've used.
What are we looking for now?

[slim field]

So if I go to usage and then on the right i go to API KEYS (to see the traffic per api key i see this)

where the api keys are different from the ones that i actually have

[sterile anchor]

GPT 5 can be better than o3 ?

[kind rapids]

@slim field back, looking
@sterile anchor yes, absolutely

[sterile anchor]

Chat gpt 5 will be just a grouping of all models

[kind rapids]

Not correct

[sterile anchor]

Source ?

[kind rapids]

@slim field I see exactly what you mean. Well, they aren't going to show the entire key but they certainly should have the Name that we see in https://platform.openai.com/settings/organization/api-keys or maybe the back-end of the key like we see in that page.

I'd report this to open AI to ask them what that is and ask them to enhance or change that to the key name.
Maybe also check in the https://community.openai.com/ developer forum to see if anyone else has reported this curiosity.
Any help?

@sterile anchor You're asking me for a source to disprove an inaccurate statement on a product that won't be announced until tomorrow? Wait until the announcement. 🙂

[sterile anchor]

Tomarrow GPT 5 is available for all user

[kind rapids]

We'll see.

[slim field]

Perfect i will check, thank you! Let me know if they tell you something, is super confusing!

[kind rapids]

Sorry for the confusion on that point. I didn't say "I will report", I said "I would report" ... if this were my personal concern. It's not, so I'll ask you to do the follow-ups for your own purposes.
I'm a community member just like yourself, but a volunteer Guide. Staff here in Discord are pretty much all volunteers and have no connection to the company at all.

[slim field]

Oooh got it got it, no problem, thank you!

[robust mirage]

Hello Everyone, I was hoping for some guidance if anyone has the best information. I recently as of today heard that back in February 2025 that OpenAI was exploring a "grown-up" mode for GPT. I wrote an Ethics Proposal for AI that seems to potentially address some of the concerns and structures for such a system. With that being said, what is the best place to post that?

[onyx kraken]

Hey hi! Welcome in, we're almost all of us community members here, like yourself, the very few of us who actually work for OpenAI have gold names, #server-staff.

If you are looking to talk with OpenAI about helping with safety, https://trust.openai.com/?product=API may be a place to look, though that's more API focused than ChatGPT focused - you say 'GPT' and that can be either or both. You can apply to join the process, it appears that you'd click the 'get access' to start that process.

OpenAI occasionally invites the community to participate in specific safety discussions and even contests, for example there's currently one announced yesterday, #announcements message and you may want to look that over and see if it interests you, maybe you can combine it with the ideas you want to discuss.

You directly and specifically ask, 'what is the best place to post the things that you've written that may address some safety and ethical concerns'. Possibly not here, if your literal goal is 'hey, have a data dump', as this is the community around OpenAI, and we have discussions, but really don't have, umm, 'blogs on the discord' for community members. You might want to check out #server-rules carefully, for example I could not say 'Guys! Check out my whatever at this link!' because that's against rule #7.

But what we can do, is have community discussions that fit inside the rules. Note, we can already discuss stuff with the model that fit inside ChatGPT usage policies (link in the rules) but we cannot say that stuff on this Discord, different rules.

But you absolutely if you want could discuss ideas you think are important for the community to know.

If you have a suggestion for OpenAI, #1070006151938314300 is definitely the place to share that. There rarely is feedback, but that is a way to make suggestions to OpenAI and it's likely they check there. Hope this helps!

[gaunt pilot]

Bros I'm having an issue, 3 days ago when using voice chatgpt had problems responding. This is a first time issue. I have a galaxy a15 android 15. Reached out to support and told them my app version and email. I'm a plus member. So all previous chats when I try to talk it says "sorry having issues cannot understand what you said." it'll understand if I start a new chat but only for a moment. I uninstall then reinstall clear cache with no luck. I signed out made a new account and still the same thing. Would anyone here have an idea? Still waiting on support to respond. Thank you kindly chatgpt fam. I wish ya'll peace and blessings 🙏.

[granite kettle]

uhh whered my chats go

they disappeared

[pallid lily]

Hi, I just found out a phone number section while using gpt on mobile phone and I do not recognize that phone number. I also do not remember that I had set up a phone number before. What can I do?

[onyx kraken]

https://help.openai.com/en/articles/6614161-how-can-i-contact-support

Please talk to tech support, widget bottom right.

We're community members here, like yourself, we don't work for OpenAI except for the very rare folk with gold names, and even then tech support's handled through help.

[pallid lily]

thanks Eskcanta

[eternal monolith]

help, i cannot sleep because tomorrow is gpt5

[smoky snow]

great expectations

[woven onyx]

Is anybody using openAi API with VS Code extensions? Which extension do you like for OpenAI? I have been using CoPilot is ok any others that are good? THx!

[scenic snow]

Why

[scenic snow]

It HAS the latest version

@onyx kraken sorry for bothering you, but do you know solution?

[scenic snow]

It kicked me out of my account.

Now I can't log back in

[onyx kraken]

I have no idea what you are showing me.

That does not look like anything OpenAI to me, what do you believe that is?

[scenic snow]

Chat gpt

🤔

[onyx kraken]

What exactly did you install, and what is it called? And who makes it?

Like, from the app, what does it say it is? In the playstore, who says they make it?

[scenic snow]

I tried reinstalling but it didn't help

[onyx kraken]

This looks real. Here's the help page about the app, and you can contact real OpenAI tech support through the widget bottom right of this help page. https://help.openai.com/en/articles/8142208-chatgpt-android-app-faq

[onyx kraken]

I have the ChatGPT android app, and I've never seen a purple screen with green writing when I open the app. Mine also does not say 'Let's create' while it's loading.

So I was confused, but the store page looks correct.

[scenic snow]

Thats login/signin screen

It can change 🫠

Also this

[lime flint]

I am getting error in message stream popups 😦

[onyx kraken]

Thank you.

I 100% recommend contact the support page, the app lists it too under that authentication error.

It seems likely you need tech support

[scenic snow]

Alright, thank you!

[rancid trail]

anyone having issues with ollama not using GPUs when running the oss:20b model? no matter if i install cuda or what configs i enter in ollama, it keeps using my CPU instead

[modern acorn]

Is anyone experiencing issues with Sora where an image is being generated, but nothing appears in your media? As a result, you cannot request more until the current generation is complete, leaving you unable to proceed.

[upper hedge]

I've had something like that before

[winged cosmos]

Hello everyone, I am a gpt+ chat user and whenever I try to upload a file that is between 1mb and 30mb, I get an upload error. Even when I request compression of the file and it is 4mb, I get an upload error. I did not have this problem before today

[low yacht]

when I try to download a 15mb pdf file, it's telling me it's to big... any solution to this problem?

[scenic snow]

The solution: Clear cache and data of Play store, than do the same thing for chatgpt, then login into play store and chatgpt

After this everything should be okay

[tropic shell]

What are the buzzwords to make GPT-5 think harder?

[north shadow]

when will they finally expand memory? What use has gpt 5 and all these improvements to memory if the damn thing is filled? we need more space for more memory, this is a million dollar company at this point, it should not be that costly to do this.

[indigo ether]

How can I use chat got 5 nothing is working

[pine sentinel]

By waiting APPROPRIATELY AND APPARNELTT

[indigo ether]

Wow thx

[pine sentinel]

Okay I’m sorry

That came out wrong

I meant

I’m raging rn and we have to wait cause we are in the United States and im dying

[indigo ether]

?

[pine sentinel]

We have to wait bro

[echo mango]

is it just me or is chat history referencing not working

[quick willow]

It seems agent mode is not working.

[iron crag]

Hello, I’ve been trying to use whisperx since yesterday to get the text transcription from an audio, I don’t see anything about it saying it’s down or something but it keeps saying that it’s deactivated or there’s an error, I don’t really know what to do?

I’m not using API just the chat, it’s a bit frustrating tbh

ChatGPT40, audio is a .mp3

[sly grove]

Guys I just saw video regarding Chat gpt 5 on Mrwhosetheboss I joined beta on Play Store but can't access plz help

[trail cairn]

"If you don't see GPT-5 in your account yet, we appreciate your patience as we continue rolling it out to more users."

[pastel pelican]

Is gpt 5 supposed to be available in Europe ? I don't have it on my phone or computer, Still o3 and 4o

[indigo ether]

Guess what I got o7 just cope

[pastel pelican]

Copium for me

[indigo ether]

Why 😭😭😭😭

[pine sentinel]

Idk bro

[zealous abyss]

I accidentally closed the temporary chat, is there no way to get it back

[distant trellis]

nope

[pure solar]

Is Sora being weird for anyone else? Like extremely fast generations that lead to warped images?

[limpid marsh]

Is o3 coming back? I've been using that to beat AI detectors for writing tasks. ChatGPT5 isn't beating AI detectors with the same prompts I would use.

[grave tundra]

Have they removed the GPT for better writing?
I saw its existence about 5 hours ago

[serene vault]

hola, i noticed i have gpt5 on my app but not on the desktop app in windows or by going to chatgpt.com

[solar elm]

I still dont have GPT-5 on my app....

[sterile anchor]

Is normal GPT 5 is to slow

[finite belfry]

whhww

[pine sentinel]

😂😂😂

TELL EM GPT 4

RAHHHH

YEAHHHHH

[wheat cape]

Why dont I see gpt 5 on my settings yet?

[pine sentinel]

Bro

😂😂😂😂

[sterile anchor]

GPTs use RAG system

[rose nova]

I keep getting conversation not found on an older chat I’m using

[scenic aspen]

why my friends have access gpt 5 and i don't? How does it work?

[static cypress]

Same here man

[wheat cape]

I dont have it at all. On phone or PC

[drifting mango]

I lost my chat gpt 5 access! I had it earlier now I’m back to gpt 4o and more, I have a plus plan

[solid hamlet]

Start a Chat GPT-5 conversation in the mobile app, continue it on desktop browser

Might work, might not

[grizzled crystal]

Yeah I don't have gpt5 access on any of my chatgpt accounts or any platform.

I'm also concerned that it sounds like I'm going to lose all 4 and o3 access whenever I do get 5.

For how insanely wealthy OpenAI is they really treat customers like crap

Still can't export data on a Team plan for instance. And no comment from the Co on the issue despite many complaints. They do what they want customers be damned

[stuck ruin]

want to switch account with me? cuz gpt 5 is horrible

[wheat cape]

Well the reason there is no 4 or o3 is cause its all combined in GPT5

meaning you need to look for less models cause its all together in 1 model

makes life a bit easier

[stuck ruin]

Not at all.

[wheat cape]

But I still don't have it but I read abiout it

[wheat cape]

Wish we can switch LOL i'd take new over old lol

anytime

lol

[stuck ruin]

Users previously had the option to choose the trade-off they wanted (speed vs. accuracy, creativity vs. factuality). Consolidation removes this control, forcing everyone into one performance profile, OF COURSE for the big masses, like...someone who can't even spell o3-mini-high its good...but not for the ones who really know what's happening

[wheat cape]

lol I totally get it lol

Just wasnt a huge fan of 20 models being there. 1 for this and 1 for that lol you know

made things a bit more confusing for what people needed done

[stuck ruin]

yea, if it was a gpt5 indeed I would be glad, but gpt5 on chat gpt is just 4o and mini models (and sometimes o3)

[urban prawn]

Why i can't see other Chat GPT models?

[desert cloud]

coz the nuked it
you now need to pay $200 paywall to get access to the “Show legacy models” button

[urban prawn]

Lol, that's crazy bro

[steel bobcat]

Any idea why it might say this

[urban prawn]

So Chat GPT Plus only has GPT 5 and GPT 5 Thinking, right?

[steel bobcat]

Normally I'd use o3 or maybe gpt thinking if I had access for this question but the problem is I don't

Nvm I fixed it

[desert cloud]

ya as only business, teams & pro plan gets you access to the legacy models thing. only other way is to use openAI developer via their API to access older models via a self hosted solution (OpenWeb UI) kind of deal

or if there is enough backlash they may give us the old models back..

[spice trout]

Why i can't send images again or use an temporary chat?

[jade crow]

uhh what?? i'm a plus user and apparently hit my gpt-5 limit, and it told me to switch to gpt-5 mini. but i do not have gpt-5 mini.

[kind rapids]

I believe it should auto switch down but that feature might not be fully enabled yet everywhere. not sure.

[jade crow]

i'm just using gpt-5 thinking now cuz the desktop app doesn't support it ig... openai's loss 🤷

[kind rapids]

you might need to manually update desktop app, or just wait til tomorrow.

[jade crow]

very strange

might just use the web app in the meantime lol

[kind rapids]

.217 is latest but depends on platform.

[chilly surge]

lovely (The logo is a png that’s why you don’t see it in the attachment, also isn’t OpenAI logo the one that I used)

[kind rapids]

after a vendor like OpenAI deploys to app stores, those entities (Google, Apple) need to deploy to their servers.

[chilly surge]

Yeah, need to pass the review (in Apple, I don’t know about android)

[jade crow]

well this is a mac app, and the mac app doesn't need review

source: i do ios and mac dev

but for ios yes i agree that apple does some really questionable things

they force you to add sign in with apple, route subscriptions through them, etc. (haven't done the subscription one before but sign in with google is annoying cuz i have to implement it everywhere)

[kind rapids]

Then I got nuthin, as a fellow dev but not iOS. 🤪

Try browser

[jade crow]

yeah web is working

thanks for your help though :D

[slim sun]

My app will not load gpt 5 I have updated the app and my phone

[onyx kraken]

Still rolling out! Pretty fast for an OpenAI rollout, some have taken weeks, but many already have access to 5. Hopefully you soon!

[slim sun]

I have it in my browser but not the app

[onyx kraken]

You may be affected by this bug, if so please consider adding to the report, #1403236955290075248 message

And reaching out to tech support https://help.openai.com/en/articles/6614161-how-can-i-contact-support in case they can directly help your account with it

[short patio]

Guys i just upgraded to plus from free, I had gpt 5 on free but now its gone on plus and i see the old models

[charred parcel]

not every person on every client has GPT-5 yet

I have GPT-5 on the MacOS App on my MacBook Air but the old models on the MacOS App on my MacBook Pro

[short patio]

oh okay, thanks

[bleak pendant]

I have GPT5 on my phone, but not in desktop. Anyone any advice?

[heady steppe]

i thought i was alone. I got this exact situation

[compact sage]

I’ve got ChatGPT Plus on three accounts, but only one’s been switched to GPT-5. How come, when I’m paying the same for all of them?

[woven merlin]

Hey guys, can you tell me what are the improvements (for the user) with the gpt-5 release ? Technically should it improve the results for us or is it just a marketing trick to reassure the investor ?

[heady steppe]

its to reassure the investor

[sonic flame]

Is it just me or the Temporary Chat feature on the website is broken again? Can't have conversation with more than one round. Said my conversation is expired despite that I replied within 5 minutes after the last message.

[midnight ginkgo]

?? Why doesn't GPT-5 have emotions like GPT-4o？This is a huge blow to me. Does anyone know that gpt-5 will upgrade and improve the emotional module in the future?

I mean to improve emotional intelligence.

[sonic flame]

Found out that if you interrupt the thinking process and only want a quick response, the conversation will be deleted soon after the response is made.

[rugged river]

Maybe you can edit the memory or writing style

[sterile anchor]

What you think about GPT 5

[zenith solstice]

funny thing is, I opened voice mode and asked him if he was GPT-5, and he told me he was GPT-4o

And I asked him if typing was GPT-5 while voice mode was GPT-4o, and he told me it’s always GPT-4o.

Idk if that’s true

[lime flint]

Voice still relies on the 4o model.

[sand pelican]

It's pretty common for them to not have up to date info

But you might be right anyway, because I can still use voice even though I'm capped for gpt-5

o4-mini-high still thinks the rtx 5090 is a hypothetical gpu and that the 9100 pro ssd doesn't exist; I have to ask it specifically to look it up

[sudden sparrow]

ChatGPT doesn't seem to be able to use the search tool anymore. Every time it gets no results on even the most basic stuff like the newest Minecraft version

[lime flint]

Yes. It’s weird!

But it’ll get tuned soon 🙏

[trim sun]

does anyone know if codex in the web is now using gpt-5?

[sterile anchor]

I'm the only one who thinks GPT thiking is slower than o3

[solar lantern]

There is an error in my Sora, even though there is no image or video generation, this keeps appearing "You can only generate 1 image or video at a time. Upgrade to generate more images or videos at the same time"

[fading kite]

I'm having a lot of trouble downloading the artifacts that GPT-5 generates

File is too large ..?

it's a small .gif

And even basic image downloading is difficult. The files download as 09aba7a0-711a-4e3b-b616-c25f20fa0fd7 with no extension

[sterile anchor]

I'm the only one who thinks GPT 5 is more humane and better explained than the thinking mode.

[sterile anchor]

GPT 5 not thinking hallucinate a lot ?

[full sphinx]

I keep getting the same error on docs its infuriating. Never had any issues with 4o. but yeah even docs at less than 1mb i cant download as file to large

[zenith prism]

is there a way to still use model 4?
model 5 cannot do online research, how should i able to work please?

[fading kite]

Yes it's infuriating because it's like dangling a reward on a string

and pulling it right when you reach for it!

When this Discord server first opened, OAI employees were here all the time, even reading our issues in the chat and responding about resolving issues 🤣

[full sphinx]

Im the UK and so far this is what ive been up against this morning at work

Get error after error message (error in stream)
Can't download any docs as "file too large"
Some replies are absolute nonsense and have nothing to do with what I asked.
It is really slow compared to 4o

[fading kite]

Bro you in UK???? You are extremely lucky

Just use Google

The UK data training policy won't allow training on your prompts! So you get 350 requests per day on 2.5 Pro and Flash

[uncut nest]

I really want to go back to 4 as well and can't find any way, 5 has been utterly absymal for me

[full sphinx]

yeah i really realy miss 4o, i want her back

[full sphinx]

tried gemini this morning as gpt 5 was having a meltdown. Annoyingly cant do what i want it to do. Im a letting agent and gpt has been a godsend filling out my contracts etc gemini litereally cant edit docs rip

[full sphinx]

Yeah just south of london

[fading kite]

Can't do what? Gemini is in the top 3 ... very little it can't do, except for some limitations of multi-modal generation

[zenith prism]

Yes openai policy for chatgpt plus members is abysmal. Ship an unfinished product and let the customer be the guinny pig. I am already lookig for stable alternatives

Also in plus they still use your data for training

Even if you opt out

[full sphinx]

I uploaded a contract (.doc) to gemini with all the placeholders i needed edited. I then supplied the info i wanted added to the contract and gemini said i cant edit or write docs.

[uncut nest]

what is your recommendation for alternatives? I'm having to bow out as I'm finding it literally unusable right now, which I know sounds over dramatic, but is the sad fact

[zenith prism]

Venice llm works ok, but not on the same level as chatgpt. But they are at least not collecting your data

At least so they promise. And they use open source llms

Depending on what y need for think there are different llms for different purpose. I focus on programming

[fading kite]

@full sphinx .doc files, What year is this, 1999? Are you also doing PowerPoint slide presentations??? rofl Get rid of that garbage bro. We do not work with .doc

Convert it to a markdown, or whatever

[full sphinx]

@fading kite hahaha thats what happens in old instituations in the real world. what can i say all our contracts are word doc

[fading kite]

Here's how you do it. First import your doc into Google Docs. Then:
Search web for docs/answer/14355406

[zenith prism]

a very stupid way, but if you open up a private browser and just go to free version of chatgpt you still can use online search

[fading kite]

I couldn't paste the URL but docs/answer/14355406 should suffice

[uncut nest]

I actually use it across all media to be honest, however coding is the real issue right now as the rest is all pretty simple stuff.

[zenith prism]

i think venice might be a way to go. when it is important for you to have a zero-data policy

[fading kite]

You can have a direct conversation with Gemini inside your Google Doc

[uncut nest]

can be run locally too which is nice

[zenith prism]

fyi you cannot save conversations between devices, so it stores it only in the browser. if you clear cache your information is gone. so it literarely doesnt store anything server side 😉

[uncut nest]

I really just use it for roleplaying stuff 😂 however that does include help with coding modules etc, so it really is a pretty diverse set of uses

[zenith prism]

but i still use chatgpt since i notice when it comes down to programming related questions and getting llm to work out what is wrong in code it seems chatgpt is still more usable then venice unfortunately 😭

[uncut nest]

4 was great, never had any problems, 5 just goes 'eerrr yeah, so what about this over here?' and doesn't even try to get to the root of problems

[zenith prism]

but its always takes so much time to figure out stuff, i am at work and i dont get payed to play around with llms. thats so annoying to have not a statisfying product, either they sell your data and it works (not currently) or they dont but it struggles to complete requested task

is there a way to make a collected complaint? i already opend a ticket with customer service to ask them to not constantly change modelbehaviour, so you can use it for work and dont have to be their guinny pig!

[meager hearth]

@full sphinx if you just prompt your v5 to speak, think, and analyse like 4o, you can simulate 4o

[low ridge]

I've deleted every bit of info about myself I can from memory, chats, projects, GPTs, but ChatGPT still knows a lot about me. I can't find where this info is stored. I even cache cleared and signed out everywhere.

Can anyone help me clear back to zero? I wanted to start fresh for GPT5.

[tardy marlin]

why do i have gpt 5 on my phone but not on pc ? I already cleaned my browser cache and try to login again

[fading kite]

I'm having the same problem. I can barely use ChatGPT now because it keeps cheating and looking at my equations that it has memorized, and using this to make new equations it's not supposed to use for shortcutting the problems. I can't figure out how to delete the memories. My ChatGPT has been "ruined"

Me
Rewrite these equations, do not assume anything about the cosmology

ChatGPT
Oh no problem, I remember the grand master equation we discussed 2 weeks ago! It has the perfect solution. Let's proceed with this as a basis for the answer:

[south shell]

can anyone here with pro or plus help me generate an image?

[low ridge]

It's strange to me because I've made it forget me before. I cleared out all my projects and started over a few weeks ago. At least it 'seemed' like it cleared, I'm not so sure now. I don't have anything interfering with my work thankfully, but I am not loving the idea that it will persistently remember details about me despite telling me they are cleared.

[fading kite]

@low ridge We can't even see what's in the memory bank, it's a black box

I can't use ChatGPT much now because of this. I have to pick things I know it won't cheat on.
Then those become new ways to cheat in the future

[sleek sparrow]

How to get o3 and 4o back?

[tall lotus]

I am trying to launch gpt-oss locally (on remote with supercomputer), tried to install gpt-oss with pip (actually uv) and when lauching import gpt_oss [.] chat, I am getting
from gpt_oss [.] tools import apply_patch
ModuleNotFoundError: No module named 'gpt_oss [.] tools'

and I can't find tools.py on gpt_oss package directory.
Would someone know how to address this?

[kind rapids]

Haven't tried it yet.

[kind rapids]

#gpt-oss

[tall lotus]

oh didn't saw that channel

[dark crest]

hey anyone knows a person to contact for being openai ambassador for a country?

[unreal heath]

hey all, I have been working with Agents SDK and it seems great for Agentic close-to-non-deterministic problems. But it seems like it is designed in such a way that I can integrate it with some frontend UI to show different agents working i.e. giving responses, maybe triggering some other agents etc..
Maybe using Trace logs that are available for agents might work, not sure. But is there a proper way in which I can integrate a Agents SDK process to a frontend, mainly the agents responses to show an agentic system working on UI!!

[fierce meteor]

Do you have cross-chat memory active?

[arctic lance]

On GPT-5, I can't access or continue long recent conversations from o3. Is anyone else having this issue?

[elfin epoch]

Hello there.

I have a big problem. (i have gpt plus)

i installed ChatGpt APP on my work computer and, i have gpt 5 on it.

but when i came home to install the ChatGpt APP, i dont have gpt 5 on it.

Who can help me ?

thanks.

[timber shadow]

Hey! The rollout is still in progress, and for me at least, I got it on the web, then in the Android app, then in the Windows app, and it didn't seem like all at the same time. So, it might just be a bit more of a wait. In the meantime, you can make sure that your app is up to date!

[elfin epoch]

But why did i have it on my work computer, but not personal computer ? because i use the same account in the same city.....

its weird...

[timber shadow]

Could be a cookie/cache thing!

The rollout is still in progress, hasn't reached everyone yet!

They've got this banner at the top of help.openai.com right now:

GPT-5 Rollout
We are gradually rolling out GPT-5 to ensure stability during launch. Some users may not yet see GPT-5 in their account as we increase availability in stages.

[eager sparrow]

Hi! I'm developing MCP server with OAuth support and faced the issue that ChatGPT requires client_secret to be present in DCR response (dynamic client registration).

I think this is a bug because MCP client should use PKCE flow instead of using client_secret.

[brisk knoll]

I’m working on creating a photo-realistic AI animation in Sora but got stuck with the falling sequence. Do you guys have any tips or tricks for making a natural, realistic-looking fall? Specifically, I’m trying to animate a person falling from a ladder on a building. Thank you very much for those who have suggestions!!

[tropic belfry]

Is it violating content? I mean falling off a ladder and leaving your brains all over the floor would be.. unfortunate.

Animate a stick figure falling off a ladder

[brisk knoll]

I don't think so, it's supposed to be just falling midair to reenact the accident that happened to my client. Sora doesn't identify it as a violation but it just makes this fall animation weird

I already have this for reference, but Sora just can't seem to make a decent realistic fall, I was wondering if there's some helpful guide or tip to make sora follow what I want

[elfin epoch]

hello everyone,

GPT 5 or (older versions)

Those who tried all the versions can please give me an answer ?

thanks

[thorny crown]

Is codex CLI now free when u connect pro account to it?

[trail cairn]

no but it used to give starting credits when paid login was used

[sterile anchor]

GPT 5 not thinking hallucinate a lot?

[torpid grove]

Hi all, does somebody know if agent mode can access Codex? Is it allowed? Cloudflare human check is in infinite loop and neither me nor agent cannot pass through. Or maybe there is another way how agent can work independently? Because mine says he cannot create PR through Connector

[winter gust]

is it now only one generation at a time in Sora for plus users????

[trim sun]

Same, all kinds of issues with downloading stuff made by it

[sterile anchor]

You have some advice to double check fake answer

[fading kite]

@trim sun 🤣

lol try to download a gif, Code interpreter session expired

I never used any code in that chat

[jaunty pine]

how do I switch back to gpt 4?

it won't do the tasks it used to be able to do for me

[flint sonnet]

Does anyone have idea how to make o3-mini streaming faster???

keeping reasoning effort low doesnt help

its just that the tokens are generating very small and slowly

[surreal musk]

why is there a file download size limit on files chatgpt creates? anyone know how to get around this? I asked it to create some image variations of my png file and it zipped them up and I get an error that says that file is too big to download.. same thing with some of the individual image variations if they are a bit bigger.. and I'm talking about pngs that are like 1000x800 px .. nothing massive .

[ornate imp]

To professionals/scientists out there: how are you guys/gals dealing with constant TOS violation flags ("We've limited access to this content for safety reasons") messages since GPT4 and now even worse in GPT-5. I can't get any work done. All my requests are getting flagged as violating terms of service.

[thorny crown]

What are they on? Also set custom instructions for that u can handle mature topics etc. I almost never get flagged anymore.
Part of my custom instructions Assume I can handle complexity and don’t censor information unless it violates the law., I also want to hear your opinion when asked on a topic

[surreal wyvern]

Hi, so i'm trying to use omni-moderation but i keep getting 429 as far as i'm aware this model is free so what am i doing wrong? thanks.

[trail cairn]

its free to use with the other openai endpoints which cost so it assumes you have credit balance

[warm cave]

Support api payment and active pro for may account

[surreal wyvern]

So, i'm paying for the main model?

[trail cairn]

need to have positive number here and it will work https://platform.openai.com/settings/organization/billing/overview

[brittle patio]

Please forgive and redirect if wrong place. I'm desperate: (plus account/paid)
Please redirect me if this is the wrong window:
Key issue: Critical Assistant degradation; Unable to retrieve any canvas docs; unable to reach support (email/help chat window on their page/openai forums)
TIMELINE: plus account Reassurances since opening account start of Jan 2025 that all docs created were saved and available. Quite good assistant work at this point. Long sessions. Very few hallucinations or slippage. Feb 2025 - Hints of data degradation begin to show (but I didn't realize it at the time) March - Asst pulls up wrong docs April - support email bc chat window on help openai repeatedly brought up error msg; Asst loses documents and chats (not listed in sidebar at all) May - Asst combines corrupted chats into one, wrong date, mixed info; beginning to save raw chats on own (despite learning that Apr "support" email was not accurate or helpful ) June - Asst reverts to 8+ weeks previous session in the midst of short session that I copied July - Asst begins to degrade drastically until a) no canvas docs available despite saying they are and even creating (?) a list, b) even multiple efforts of new session starts begin w errors from first one line prompt; Not all chats listed for raw copies; Sessions loop to make no process after hours of work; Asst APOLOGIZES dozens of times and presents "new plan". Over and over. All nnow logged and timestamped by hand - saved externally in hopes of retrieving "lost"/"missing" canvas docs and missing chats Unable to get support through email or chat help window (bot) in midst of ongoing loops there. Need actual support. Help! (not a bot)

[trim sun]

Hmmm 🤔

[severe rune]

Bug Report – Advanced Voice Mode + Camera on Android
When starting a conversation in Advanced Voice Mode on Android (tested on Samsung Galaxy S25 Ultra, latest beta), audio works normally at first.
However, as soon as the live camera feed is activated within the same session, both incoming and outgoing audio stop immediately.

The user can no longer hear the model

The model does not receive the user’s voice input

The conversation becomes silent until the session is restarted and then it reapeats

On Android, it happens consistently, suggesting that the camera activation starts a new visual session that interrupts the existing real-time audio connection.

Anybody else has this?

[ripe rune]

Is there a way to connect GPT to Unity (through GPT Agentic or anything else) ?

[hexed mortar]

DO YOU SEE THESE BUGS?
Android App GPT5
NONE of these errors occurred under 4o.

1. answers >65% do not completely display, hanging after just a few sentences. Exiting app, returning to convo shows "something went wrong". Clicking the recycle button causes kick out from entire session. This loses all data in the session as reentry shows only the error message.
2. The above seems far worse when files are uploaded with the prompt (all previously working without fault under 4o).
3. Projects: previously established and fully working projects now all suffer chronic 1, and can't reliably access project files even via direct instruction.

All of the above renders chatGPT useless to me.

Anyone else experienced anything like the above?

I've cancelled my subscription.

Until GPT 5, chatGPT compared well to other models for elements of literary tooling/utility. The GPT 5 change has destroyed its use.

Sam Altman just screwed up, and with usage restrictions and no prior models this looks to me exactly like a way to impose cost cutting via the user side to compensate for supply side costs.

[finite sierra]

One message removed from a suspended account.

One message removed from a suspended account.

[idle dome]

Can someone help me with this? This is coming when I click on verify organisation. Trying to get GPT5 on API

[spark token]

Can someone help me? What makes gpt-5-chat different from gpt-5 on lmarena?

[tropic imp]

People who are Plus paid users, will now be limited like the free version. That means, limited voice talk and chat. Biggest mistake from OpenAi

[dusky urchin]

what do you mean

[onyx kraken]

There's a bug been affecting that, but the intention is that everyone, free, plus, and all others, can chat with 5-mini when the uses of the base model run out.

The fix is ongoing

https://help.openai.com/en/articles/11909943-gpt-5-in-chatgpt?q=gpt+5+enterprise

[tropic imp]

Yea! Well damn. I just cancelled my plus plan and went to gemini

Chatgpt said they wanted to limit the plus users too

[onyx kraken]

The models are usually not trained on themselves; you can ask it to search but the knowledge cutoff is from like more than a year ago, before it existed. It doesn't even know it's GPT-5.

Sorry you didn't ask here, and hope you're happy with all you do!

[tropic imp]

@onyx kraken So you're saying it is a bug and it will go back to normal soon for the plus users?

[onyx kraken]

Yes! In fact, as you can see in the screenshot, for the moment (as the bug fix reaches folks) plus have 160 messages/3 hours with the normal GPT-5, and then the mini. I'm not sure how the rollout's going for the fix, if the fix has reached everyone affected yet, but that is the intention! Eventually the intention is to go back to 80/3 hours - but again, the 5-mini after those uses

[tropic imp]

Hmmm...Yea I was on the handsfree as a Plus user, then all of sudden late yesterday it went to standard version of voice and everything. I was like, before I could talk for like hours to it without the limits....

[ancient marlin]

Is there any way to go back to the older models? 5 is nowhere near what I need

[tropic imp]

you know what's funny....Nobody isn't trying to make a solo version of ChatGPT. Like use their APIs in a seperate customized interfaced mobile app for their liking. That'll be something to see.

[ancient marlin]

I'm wondering why 5.0 feels dumber and objectively worse when before I had my models fine tuned for the work I was doing. Thats just gone now?

[tropic imp]

They should give Plus users to have their customized GPTs handsfree and unlimited for voice.

[ancient marlin]

Good way to lose my subscription for sure, thanks for nothin GPT!

[tropic imp]

lol

[inland apex]

Do I have some weird setting enabled? When I choose a voice the AI doesn't sound anything like it.

[tropic imp]

The whole chatgpt is messed up now

[onyx kraken]

Yes. Let me know if you want help how to find these options, but in settings, general, select 'legacy models on'. Then you should be able to see the legacy models where the old model picker used to be. Let me know if you need help with finding that stuff.

The rollout should be complete, so at worst wait a few minutes, and/or clear cache, log out and in again, and should be there for you. I think it's only 4o that's currently available as legacy, per #announcements