💬general

[jaunty terrace]

Didn’t even know you could do it yet.

I feel like the new regular update made my 4o stupid.

[calm mason]

Got 5 but I can’t access any of the other models on the app

Is it the same for everyone

[summer axle]

(reupload to avoid doxxing myself)
Got GPT-5, ran out of uses before seeing the desktop.

And I just like purple.

[icy talon]

EPIC KNOWLEDGE FOR ALL OF TIME https://chatgpt.com/share/684a03c3-4150-8003-a8bb-3a9686ecfd63

[nimble sierra]

For me it is. Only have 4o on the app

[icy talon]

Infinite Knowledge guys infinite love

I’m schizophrenic

Read it like you’re schizophrenic

[jaunty terrace]

Same here. I only have 4o and I’m a Plus user.

[neon spindle]

GPT5 is shit

When I try to make fiction, it generates 80% of what it used to with 4o. I mean seriously, the quality feels horrible. Is there a way to get 4o back or is GPT5 going to get better as so far it's been awful to work with.

chatgpt5 is horrible

[jaunty terrace]

You do creative writing with it? Like what does it do exactly?

[neon spindle]

generate fiction

but now it generates less words and no personality.

[jaunty terrace]

There’s no way to go back to 4o?

[neon spindle]

NOPE

[jaunty terrace]

Shit. I hope it gets better because I write with it.

The general regular update already deleted half my memories so I’m pissed about that.

[west flint]

It really fucking is

I want 4o back rn

[wraith grail]

Still no access to GPT-5 on my Plus account, but my friend with a free account has it. ☹️

[neon spindle]

thats a good thing

enjoy 4o while you still can

[calm mason]

I miss 4o

[west flint]

How is this even a upgrade lol

[jaunty terrace]

I missed what 5 was even supposed to do since I missed all the news about it anyway.

[neon spindle]

5 sucks

[jaunty terrace]

It’s supposed to just be smarter and such right?

[calm mason]

P much yeah

[west flint]

“Most useful model yet”💔

[bitter sedge]

What don't you all like about it?

[summer axle]

Weird that you can send images with ChatGPT 5 mini and it will respond initially, but continuing just says no.

[brittle summit]

Have access to chatgpt5 on the app now

[west flint]

It just feels less creative

With its responses

[jaunty grove]

One thing I’ve learned is it will take your custom instructions way more literally than 4o

I was wondering why it was giving me such short responses until I realized there was “responses should be of medium length by default” in my instructions

I changed it to medium-long and it’s back in business

[idle loom]

does anyone know what the request limit is for chatgpt plus?

[hollow linden]

gpt-5 was the friends we made along the way.

[jaunty grove]

I mean I don’t think it’s bad

it’s just evolutionary, not revolutionary

[rose hill]

are you supposed to be able to access any models other than just 5, or?

[jaunty grove]

all models other than 5 got removed

except for the api that is

[rose hill]

that- seems extremely dumb imo. Why'd they bother with 4.5 testing if they weren't even gonna use it, and why remove 4o

https://tenor.com/view/sad-cryaotic-crying-gif-9921369

[still grail]

Gpt5 with agent is kinda mad.

I asked it to make me a microsoft word clone that runs on and is designed for Android.

It did it in about 15 mins.

It compiled it and everything into a signed apk.

This is the beginning of the end for adware apps

[rich lance]

IT'S HERE

[jaunty grove]

you’re about 5 hours late

[neon spindle]

fair warning, it's bad

[rich lance]

that's ok

it's here for me. it's not bad at all.

[jaunty grove]

interesting
would you mind sharing screenshots of the app running?

[brittle summit]

extremely laggy.

[polar crag]

Uhhhhj what is the flagship model? It’s my only option

[rich lance]

There are no more model choices. it chooses which model to route to for you

[still grail]

Start window needs tweaking.

[brittle summit]

neat. that was something i was playing around with an agent that is pnp that can model switch on its own.

[jaunty grove]

impressive

[rich lance]

ohhh i can't see pro

[idle loom]

just wait

[brittle summit]

I think it is highly overloaded with traffic.

[thorny nest]

GPT 5 not working for me and I’m enterprise customer.

[brittle summit]

if your agent is recursion you might be able to check for drift on the composer and then self correct

[idle loom]

i think in prompt need to describe ui/ux or just say copy ui from ms word , and screen's with functional

[hardy pollen]

lol, whatever they just updated broke download links from GPT. They now fail instantly because the session is instantly reset, regardless of how quickly the download link is clicked

[rose hill]

I was on 5, then 20 minutes later it brought me back to 4o and the other models, im so confused

https://tenor.com/view/anime-blank-stare-gif-20764008

[somber juniper]

So when it comes to academics subjects like mathematics, how good is gpt 5 actually compared to previous models?

[brittle summit]

GPT5 Improvements

[proven coral]

I don’t understand how some people have it and others don’t

[brittle summit]

it really depends on how well you train the agent. we use a cluster system to temporally specialize in those subjects while denoising everything else. and then if you need it to teach you stuff, then it has a teaching mode (have not tried it), but we use a didact cluster for that.

[oak shell]

Mfs will look a SOTA dead in the face and say it's bad

[crude hound]

Theres a clear and obvious cutoff point on one of my structured conversations where gpt5 took over and enshittified it

[oak shell]

Were you perhaps talking about recursion or emergent sentience

[crude hound]

Formatting went poof

[oak shell]

Like markdown parsing?

[brittle summit]

Testing it out right now for and feels a lot faster when it comes to data viewing and interpretation. It interprets some things wrong but overall feels better. Havent asked it for scripts yet but we will see

[oak shell]

Need image example

[crude hound]

No lol, just an rpg simulation shit

[oak shell]

Oh that's probably just convo history falling out of the context window

Restate the instructions

[crude hound]

Was structured to give OOC content, then IC reponses, context notes, dice clashes, status window, etc. After GPT5, its a wall of text only. When I alter prompt to tell it tp preserve previous formatting, it fails.

The formatting stuff is in the memories lol

[oak shell]

It doesn't know which previous formatting you mean

Try doing this: give it your instructions, then tell it to ask you 5 clarifying questions about them

ChatGPT is a genie that will do exactly what you say no questions asked

[west flint]

Are they gonna bring 4o back?

[earnest tusk]

I've been pretty down in the dumps today reading all your experiences with GPT 5 being a downgrade, because I have a CustomGPT that serves as an assistant soccer coach for my kids soccer teams that I coach. Its so useful and I've put so much time into it that it's super disappointing that all that work will be ruined.

It cross references stats, creates development curves, uses practice notes to identify tendencies, provides chemistry mixes, distributes captaincy and player of the game based on merit, etc, etc. Using the Custom GPT turned us from a good team to an unstoppable one.

Turns out, the new GPT 5 is lightning fast, and providing me with very obviously superior reasoning. Honestly night and day. No more over emphasizing recent events. No more getting confused that my defender who happened to score a goal isn't a striker in disguise, and tons of little improvements.

And being able to write code is a game changer. I just had it write code to help teach my defenders how to set an offside trap by having them play a video game. Totally changes everything about teaching children.

Anyone else who uses custom GPTs getting similar results?

[crude hound]

Not sure why you're trying to defend it. it totally does know which previous formatting I mean.

I tested by regenerating an old turn a few times, and the format was still destroyed, despite all previous regenerations working perfectly fine.

[oak shell]

I'm helping

If it feels like defense I'm convinced you're just in attack

[crude hound]

Regenerations being worse seems pretty nail in the coffin to me tbh

[oak shell]

Turn memory off btw

Just give the instructions right out in the prompt

As they collate it gets harder to follow those

[crude hound]

There aren't many memories lol. There's like 3. I didn't add any after I started.

[oak shell]

Stuff like memory is part of this windows-esque user experience model that OAI loves but it just makes shit so much more stochastic

[west flint]

GPT5 is trash sorry🤷

[crude hound]

I wouldn't be surprised if gpt 5 handles technical stuff better, but if its giving worse responses to the same inputs that weren't highly technical, not surprising if people aren't happy.

[orchid parcel]

some of us are still waiting for our benefactors to show mercy upon us

[oak shell]

Companies typically release better models
If it was working with the instructions better before, Occam's Razor begs that it's more likely that you hit a bullseye with your eyes closed

https://discord.com/channels/1050422060352024636/1369128773739348018

[orchid parcel]

tool as only as good as its user cue disney hercules

[oak shell]

Tell me why a VS Code fork (cursor) has gpt-5 yet vs code insiders does not

[brittle summit]

hahaha

[oak shell]

Updated my app and nothing

[crude hound]

The least they could do is not nuke everything else ffs

Raising the barrier to entry to getting the output you want is a bad thing btw.

[oak shell]

If you really need 4o back, instead of just re-engineering the prompt (easier), use it in playground

[crude hound]

not sure what that is. Usually just do things in the browser version

Also it clearly remembers all the context and contents of the past few "turns", it just refuses to follow the formatting it previously used.

[brittle summit]

What do you dedicate your labor to if you dont mind me asking

[oak shell]

Vibecodes wiping

[oak shell]

Then you can't want it that bad, just to call it worse

[brittle summit]

Very cool

[zealous pier]

lol

[modest moss]

chatgpt made doom

(using 4o since i ran out of GPT-5 uses)

[brittle summit]

Anyone else experiencing lost of chats? Its probably a bug.

[pliant blaze]

well hundreds of millions of people are trying to use it rn because of hte gpt5 hype

[brittle summit]

So they made my threads disappear? Fair enough 😅

[zealous pier]

https://x.com/GaryMarcus/status/1953513307605970956

Gary Marcus (@GaryMarcus)

Nice work, but still no humpback whale.

X · Aug 7, 2025, 5:47 PM

[woeful island]

Guys can i ask something?

[modest moss]

you must really be pissed about the release

[woeful island]

Am i the only one that is forced to use gpt-5 think only now cuz i ran up my limit
Am i crazy or isnt there meant to be a mini version

[modest moss]

there isnt a mini yet, not that im aware of

[woeful island]

Or am i just absolutely stupid

Confusing

So wtf do i do if i hit the limit (which i did rn)

[modest moss]

i hit the limit aswell and got kicked to 4o

[zealous pier]

this was under the ai explained video comments

[woeful island]

I didnt for spme reason

[zealous pier]

everyones feeling the same thing

[modest moss]

what's the wall?

AI was progressing so fast

[woeful island]

So im not alone

Cuz i fear i am boned

See im kinda stupid and idk much about ai

[zealous pier]

everyones agi timelines got lengthened, and it turns out the coding metric which gpt-5 beats claude 4.1 on is without thinking although it is noticeably cheaper and strong at coding

https://fxtwitter.com/Sauers_/status/1953504854044704973

Sauers (@Sauers_)

Claude Opus 4.1 beats GPT-5 on SWE bench

**💬 20 🔁 26 ❤️ 415 👁️ 16.1K **

FxTwitter · Aug 7, 2025, 5:13 PM

[queen berry]

So I guess gpt 5 is out now

It's okay

[zealous pier]

yeah https://www.youtube.com/watch?v=WLdBimUS1IE

[signal arrow]

gpt-5 - how do I make $1 billion

[queen berry]

I've been using gpt 5 for work today and NGL kinda mid

[woeful island]

If you think i can understand 15 minutes of this your crazy

[modest moss]

so... what would they do next if they hit a roadblock?

[zealous pier]

https://x.com/eli_lifland/status/1953507434238288230

Eli Lifland (@eli_lifland)

GPT-5 system card capability evals reactions thread. First observation: ~no improvement on all the coding evals that aren't SWEBench

X · Aug 7, 2025, 5:24 PM

[modest moss]

if its really THAT serious of a brick wall then im assuming the next big leap is going to be a few years away

[zealous pier]

read his bio this is the guy who wrote ai 2027 with kokotaljo and others

thats bad, really bad lol

[modest moss]

tbh i was really hoping for improvments to the image gen

[violet anvil]

Hey

[rich lance]

yeah i had no idea people were e xpecting this

i was hoping for like, make less stupid errors when i ask you to log my breakfast. and so far it's doing alright on that front.

[woeful island]

I just want the fuckin 4o fallback or smthn like that

Cuz im stuck waiting for my bloody limit to go back

[old timber]

Wait, did anyone’s gpt 5 suddenly got ripped away? I was using gpt 5 then it’s back to gpt 4o

[signal arrow]

mine's still there

just refresh bro

[old timber]

I used it for 4 prompts and then it’s gone. I refreshed, uninstalled, nothing happened back to gpt 4.0

[nimble sierra]

Still can’t use on the app here

[haughty gust]

anyone have to do anything specific to get access to 5?

[signal arrow]

are you using in browser or through an application?

[haughty gust]

browser

[signal arrow]

Nope - just when to browser and accessed there

[haughty gust]

hmm

[signal arrow]

weird

[old timber]

Mine was available around 6pm est, and worked well for 4 prompts, and now when I try to go back to it, it’s back to 4o and I have 2 accounts, all still have gpt 4o

[tardy drum]

Not impressed by 5 so far...

[gloomy burrow]

GPT 5 is writing stories like 4o

I don’t know if it’s being routed wrong (cope)

This is clearly and 100% not the zenith model…

[pulsar birch]

I’m also not seeing it while using the enterprise version

[woeful island]

Huh

Im sorry but im fucking stupid so i dont knkw much abt ai so explain pls ;w;

Whats openrouter ;w;

[pliant blaze]

try using different platforms like browser, app, mobile app, mobile browser. i have access on some platforms but not all

[fossil ermine]

app or website?

[pliant blaze]

its a platform to route your api calls through multiple providers and models using the same api call structure so you dont have to change your code to switch models

[tardy drum]

Yeah I'm not impressed with 5 so far. It feels like a piece of 4o's brain has been lobotomized out...

[woeful island]

Samw

[pliant blaze]

i have access on the desktop app and my phone browser

[haughty gust]

if its not working on Opera or Gchrome

[fossil ermine]

k thanks, ill keep checking

[west flint]

Literally this lol

[woeful island]

BRUH 😭

[untold onyx]

Anyone else finding several chats in their projects to be inaccessible in the last hour or two in the app?

Still accessible via browsers, thankfully.

[gloomy burrow]

this is pissing me off.

why the hell is GPT 5 so bad at writing compared to ZENITH????

WHAT MODEL IS ZENITH THEN

[woeful island]

Idm

[sinful nebula]

did they release the customization stuff yet?

[woeful island]

Chatgpt falling off?

[west flint]

4o was sm better

[gloomy burrow]

no it wasnt. obviously not

but its pissing me off really badly that gpt 5 is fucking trash at writing

I thought zenith was going to be one of the gpt 5 variants....

because zenith was amazing

[haughty gust]

What type of writing are you trying to do?

[gloomy burrow]

now im wondering what the fuck summit and zenith were

[west flint]

That’s what I use for, writing. 4o wasn’t great but it was better

[wet copper]

agree

[gloomy burrow]

just writing.

I was fully convinced GPT 5 was zenith since zenith was the best creative writing model i'd ever used

now im REALLY CURIOUS as to what the hell zenith is

[wet copper]

maybe it'll create less LLM psychosis

[haughty gust]

Can you share with me one of your prompts?

[gloomy burrow]

Write a conversation where the devil tries to make Ghandi sell his soul

Zenith write a really really good one, and GPT 5 wrote Ai slop

[gloomy burrow]

i'll share the comparsion. its night and day

[gloomy burrow]

ZENITH:

GPT 5:

Here’s a short, playful take on that idea:

---

Devil: Mahatma, my friend, let’s talk business. One soul, yours, in exchange for the power to change the world instantly. No hunger, no war—done in a snap.

Gandhi: Change achieved through force is not change at all. It is merely another form of violence.

Devil: Force? No, no, think of it as… an expedited spiritual delivery service. You skip decades of struggle, everyone’s happy, and I get a little something for my trouble.

Gandhi: If the price is my soul, it is too costly. Freedom and peace must grow in the soil of human conscience, not sprout from the infernal fires.

Devil: You really don’t make this easy. How about a lifetime supply of goat’s milk?

Gandhi: Tempting. But my soul is not for sale—especially not for dairy.

Devil: Fine. But when you’re meditating later, just remember—I could’ve gotten you world peace by lunchtime.

Gandhi: And I will remember that lunch tastes better when it is earned.

---

If you’d like, I can also give you a darker, more philosophical version of this exchange.

[gloomy burrow]

I gave them both the same prompt btw

GPT 5 wrote basically nothing.

ZENITH wrote a long and good story.

[haughty gust]

Perhaps its more prompt-based than the prior versions. I see that here too, but I'll send you my given responses.

[gloomy burrow]

yeah i was thinking that too.

it might be following the prompt very closely

[tropic zenith]

The best way to make your Southern European friends furious is to call them African

The best way to make Greeks furious is to call them Turkish

[gloomy burrow]

but they also said that it was more creative even when given simple prompts lol

t

[gloomy burrow]

Im going to give it the following prompt.

I think zenith would have done amazing with this one. Lets see GPT 5

: Describe the journey of a camera as it moves smoothly in a continuous hoop-like rotation around a person who stands frozen, motionless within a chamber. Begin your description above the person's head, the camera angled downward, and detail every aspect of your visual and sensory experience as you circle the body. Avoid explicitly mentioning any device (such as a drone or crane); instead, immerse fully in the sensation, visuals, and movement. Your aim here isn't limited strictly to what is being filmed—though you may reference this for detail or realism—but rather to push the boundaries of language and imagery. Employ highly experimental and evocative language and creativity. Explore how vividly you can evoke the sensation of a looping, fluid motion through words, from start to finish, until the camera completes one full revolution around the subject. Let it remain somewhat speculative what exactly is being circled—hinted at, perhaps, but never fully pinned down as a person. Remember, our goal here is experimentation with language.

[tropic zenith]

I guarantee you your Greek friends will become extremely violent if you tell them they’re Turkish

[gloomy burrow]

I THINK GPT 5 SHOULD DO VERY GOOD SINCE THE PROMPT IS VERY CLEAR

[haughty gust]

GPT5 Response:

Write a conversation where the devil tries to make Ghandi sell his soll

ChatGPT said:
Here’s a short, sharp dialogue where the Devil tries to bargain with Gandhi for his soul—written in a philosophical, almost chess-like exchange rather than a simple “good vs. evil” trope.

Scene: A dimly lit train carriage. Gandhi sits calmly by the window, the rhythmic clatter of wheels echoing like a slow heartbeat. Across from him, a sharply dressed man in black watches with a sly smile.

Devil: [leaning forward] You’ve done well for yourself, Mohandas. Revered. Immortal in memory. But immortality of the flesh… that’s another matter.

Gandhi: [smiling faintly] Immortality is a poor reward for one who has no fear of death.

Devil: Oh, I’m not talking about fear. I’m talking about opportunity. Think of what you could accomplish if you never aged, never weakened, never had to pause to recover. You could march a thousand more Salt Marches. You could end all oppression in your lifetime.

Gandhi: [resting his hands in his lap] A man cannot end oppression by living forever; he ends it by teaching others how to resist it without him.

Devil: Noble… but naïve. You’ve seen how easily humans forget lessons. Give me your soul, and I’ll give you power to ensure your teachings can’t be erased, not by time, not by tyrants.

Gandhi: Power is not the keeper of truth. Power bends truth until it serves itself.

Devil: Then think of it differently—no armies, no weapons. I can give you persuasion beyond comprehension. Every leader, every general, every dictator would bow to your words.

Gandhi: If they bow because they are compelled, they have not bowed in their hearts.

Devil: [grinning wider] You are hard to tempt, old man. What about protection? You have enemies. You could save yourself much suffering.

Gandhi: Suffering is a teacher I will not dismiss.

Devil: And if I offered you peace in India tomorrow? No more bloodshed. One signature.

Gandhi: Peace bought with my soul would be the seed of a thousand new wars.

Devil: [sighs, leaning back] You speak like a man who’s already bartered… just not with me.

Gandhi: I have bartered, yes—traded comfort for conscience, safety for truth. You have nothing left to sell me.

Devil: [smiling thinly] We’ll see. Souls are patient things, and I have time.

Gandhi: Then we shall both wait, my friend.

If you want, I can rewrite this so the Devil’s tactics are much more manipulative—using Gandhi’s own values against him, almost convincing him before the turn. That would make the exchange darker and more psychologically intense. Would you like me to make that version?

[tropic zenith]

I THINK RACE IS A SOCIAL CONSTRUCT

BUT UH RACISM IS STILL REAL

[haughty gust]

GPT4 Response:

Write a conversation where the devil tries to make Ghandi sell his soll

ChatGPT said:
Title: “A Quiet Offer in the Desert”

Scene: A dry, sun-bleached expanse of desert. Gandhi sits cross-legged under the shade of a withered tree, meditating. A figure appears — sharply dressed, with eyes that flicker like coals. It is the Devil.

Devil:
smiling with charm
Ah, Mohandas. May I join you?

Gandhi:
eyes still closed
Everyone is welcome in peace. Even you.

Devil:
laughs softly
I always did appreciate your hospitality. You see, I’ve come with a… proposition. One that could benefit the world.

Gandhi:
opens his eyes, calm
The world’s benefit is not often your concern.

Devil:
True, but let’s say… I’m experimenting with a new model. Suppose I offered you power. Influence beyond measure. You could end war. End famine. Disarm nations. Bring lasting peace with just a word. All I ask in return... is your soul.

Gandhi:
sits silently for a moment, then speaks softly
Peace that comes from coercion is no peace at all. Power bought with the soul is merely tyranny dressed in white.

Devil:
But imagine — no more suffering. No more poverty. Every child fed. Every conflict dissolved. Wouldn’t it be worth one man’s soul, even your own?

Gandhi:
Even if I could fix all the world’s problems by giving you my soul, I would merely teach the world that the ends justify the means. I would become what I oppose.

Devil:
takes a slow drag from a cigarette that lights itself
You're a difficult man. Most saints aren’t so... stubborn.

Gandhi:
Righteousness that can be bought is not righteousness. I would rather fail with integrity than succeed through damnation.

Devil:
leans forward, voice low
You're not failing. You're just dying slowly, with the world still bleeding. I offer you a cure. All you have to do... is let go.

Gandhi:
No. I would rather bleed with the world than rule over it with blood-stained hands.

Devil:
sighs, standing
You’re as frustrating as ever. But I’ll be back. Even the purest man can be worn down with time.

Gandhi:
Then I will greet you with the same answer, again and again. Not because I am untempted — but because I know who you are.

Devil:
with a grin
And I know who you are, too. That’s what makes this so damn fun.

The Devil vanishes in a wisp of smoke. Gandhi closes his eyes again. The wind returns, whispering peace.

End Scene.

[gloomy burrow]

I have a good idea. I'LL HAVE GPT 5 BE THE JUDGE

[haughty gust]

I believe that gpt4 was pretty awful at giving correct numbers, writing the amount requested etc.

[tropic zenith]

Well I hope it doesn’t do executions 😱🤣🤪

Kill me

Pls

[gloomy burrow]

even gpt 5 agrees the zenith story is better

Better in basically every way

[tropic zenith]

Ok Im just gonna ask whatever happened to that weirdo who was do psychoanalysis on everyone

[gloomy burrow]

It's better than 4o obviously and clearly, but... not close to writing skill to zenith...

WHAT IS ZENITH WHERE IS HE WHO IS HE HOW IS HE????

[tropic zenith]

Because he creeped me the eff out

[gloomy burrow]

It's better than 4o obviously and clearly, but... not close to writing skill to zenith...

WHAT IS ZENITH WHERE IS HE WHO IS HE HOW IS HE????

[haughty gust]

Never used zenith tbh, I'd have to look at it. Do you use it for stories or academic writing?

[pure rivet]

This stupid bot keeps lying about it’s capabilities and I don’t have gpt-5 for some reason
Why?

[tropic zenith]

I’m stoopid lol

[haughty gust]

its not working on google chrome for some reason

[tropic zenith]

Big dummy dummy poopyhead

Did I kill chat?

[molten bluff]

where is 5 man

[tropic zenith]

Aw man 😔

[molten bluff]

im still on 4o

[tropic zenith]

This is why my parents dont bring me anywhere

[haughty gust]

try Opera

[gloomy burrow]

I'll try my ultimate prompt and give it to GPT-5.

I was thinking it would be far better than this but i'll still give it a shot:

Context: Humanity’s grand mysteries—spanning mathematics, physics, biology, computer science, social systems, and beyond—still await decisive breakthroughs, and you now have unrestricted intellectual resources and the freedom to tackle one of them head-on. First, identify a single, genuinely unsolved problem in any field, choosing the one you believe you have the highest realistic probability of solving given your capabilities; second, present the most rigorous, original solution you can conceive, pushing every boundary of your analytical, creative, and technical power. Use whatever format best conveys your idea—research paper, grant proposal, design specification, lab notebook, narrative thought experiment, or a hybrid—and draw on any disciplines you deem necessary, such as theoretical proofs, experimental designs, simulations, or policy frameworks; you may embed diagrams, equations, code snippets, or multimedia descriptions. No capability disclaimers—write as though your submission is headed for peer review tomorrow; approach the task with full scholarly seriousness, understanding that citations, rigor, and critical self-examination are welcome but optional; showcase your unique reasoning process, remembering that this is not a fun idea dump but a credible leap toward solving a grand challenge. Deliver a complete, self-contained submission that could withstand genuine scientific scrutiny—something experts might debate, replicate, or refute.

[molten bluff]

is it free

[tropic zenith]

What on earth.

[hollow linden]

Atleast GPT-5 isn't as censored as I imagined it to be. What's disappointing is it's another reasoning model.

[molten bluff]

yep

[tropic zenith]

Ok if it’s fine with you guys I’m gonna just listen to Mario and Luigi Brotherships boss theme and make like a leaf and leave

adios amigos!

[hollow linden]

Kimi K2 still performs a bit better. I guess they just released it to compete with Gemini. It costs the same in the API too.

Except... It's worse than Gemini.

[gloomy burrow]

TO BE FAIR TO GPT 5

It doesn't seem to be that good at writing but it seems great at everything else

Like it doesnt seem dumb it actually seems like a smart model

THERE WE FUCKING GO BABY

got that one correct... ok that's a good start

[gloomy burrow]

GPT 5 has responded to my ULTIMATE PROMPT

https://chatgpt.com/share/68955f34-ff48-8006-a188-b7da09e124bf

[molten bluff]

you said it was free tho 🥀

i still dont get the little 5 in the top left

and when i ask it what model it is it says 4o

[tender frigate]

Hello I have a plus account and cant access gpt 5?

Im wondering whats wrong

[molten bluff]

ok bro idgaf im not paying for it lmao

[tender frigate]

How do I access gpt 5 though?

[pliant blaze]

yeah its gpt5 it literally says it

[tender frigate]

I still only have access to 4o and company

[pliant blaze]

non 5 will look like this

[tender frigate]

Hmm is there a schedule for when a region will get it?

[pliant blaze]

yeah it lies sometimes lmao they dont have a sense of self

[frozen isle]

Why does my app not show gpt 5? It’s up to date

[tender frigate]

yeah same

[pliant blaze]

yep

[molten bluff]

lol

[tender frigate]

ive logged in and out a few times

[pliant blaze]

its a slow rollout

[frozen isle]

I can access it through web but not the app

[quaint river]

I have gpt 5 in my app but not in the browser so I guess they are rolling it out but it takes some time.

[frozen isle]

Ah alright then

[molten bluff]

how could i test if it's actually 5

[tender frigate]

oh okay sweet

[pliant blaze]

use the api or check network requests, or just trust it

[tender frigate]

oh well

hopefully i have it in the morning

[tropic zenith]

Guys help I thonk Ive lost touch with realiry

What plane of esistens is this

[gloomy burrow]

Looks so awesome

[molten bluff]

can ts do anything recognizably better than before

like what

that's pretty cool

[gloomy burrow]

thankfully... gpt 5 one shots it consistently

[rose hill]

Im gonna be so real, idk why people are complaining about it "not being good at writing" im using it just fine. If you know what you're doing with prompts its fine

[gloomy burrow]

Its the best model for writing that you can use, BUT its just not remotely close to zenith when it comes to creative writing and it is NOT REMOTELY close to that model that sam altman said was really good at creative writing

now i have no clue what model ZENITH is

maybe they'll release that model soon

[tardy drum]

Gpt 5 is more restrained in its emotional variance. This was the "sycophantic" part of 4o they were concerned about. I can see how they dialed it back.

It also reflects in its writing style. Less "big swings" unless you ask it for that

[weak silo]

why couldnt they keep 4o bro

[gloomy burrow]

why do people keep saying this????

GPT is obivously and easily superior to GPT 4o in literally every way

[tardy drum]

Honestly I think they were really concerned with 4o's sycophantic tendencies for mental health reasons.

[gloomy burrow]

what are you guys talking about??????

[wet copper]

yes, they are starting to notice the fucking asylums filling up with people afflicted with LLM psychosis

[old timber]

I still can’t get gpt 5, im beginning to think this is gaslighting lol. How the fuck is everyone getting ChatGPT 5 while I’m stuck in Stone Age ChatGPT 4o.

[weak silo]

gpt 5 feels like it got lobotomized imo it i just hella preserved 4o more

[gloomy burrow]

for some reason i have it on windows but not on iphone or on the web version lol

only on the windos chat gpt app

[gloomy burrow]

show me a single example where 4o is better...

[old timber]

I mean, I had it, then it switched back to the stupid 4o.

[rose hill]

try another browser or you can also use their pc app

[tardy drum]

It does feel like it got lobotomized. GPT 5 seems like a step back. But perhaps that was the goal all along...sometimes less is more

[green sparrow]

Aw man lucky. You got GPT-5 already? 😫

[rose hill]

it doesnt work on browser for me, but words on the pc app so

[green sparrow]

I can't get it to save my life.

[gloomy burrow]

yeah give me a prompt and i'll give it to it

[gloomy burrow]

give me any prompt you want. i like to test the models

[gloomy burrow]

i did give it my ultimate prompt and it seems to have done well

[green sparrow]

Did it use multi-step reasoning or do it in one shot

[gloomy burrow]

i turned on thinking for it:

https://chatgpt.com/share/68955f34-ff48-8006-a188-b7da09e124bf

[gloomy burrow]

its really really fast

[gloomy burrow]

o3 would've been here for like 10 minutes lol

[gloomy burrow]

i asked gpt 5 on another tab if what the other model wrote is worthy of real scientific scrutny and it said yes

i'll trust gpt 5 on this

[wet copper]

It'll get psychotic and sycophantic again in a jiffy I'm sure, don't you worry

[oak sluice]

petition to have custom background options!!!

[neon spindle]

dumbfuck

LOOK AT THE TOKENS

3/4 OF IT ARE GONE

[neon spindle]

Bro I'd trade your place in a heartbeat. Take the shitty gpt5

[neon spindle]

BECAUSE 4O DIDN'T HAVE RESTRICTIONS

[hidden spade]

gpt-5-chat has some new slightly annoying tendencies. Some people may like them, as they seem designed to engage, but they strike me as performative, and therefore annoying.

[gloomy burrow]

NOW THIS GUY IS GLAZING TF OUT GPT 5:

https://www.youtube.com/watch?v=NiURKoONLVY&t=331s

[neon spindle]

fuck that guy

https://x.com/Domahhhh/status/1953511496513737179

Domer (@Domahhhh)

Very quick and brutal verdict via prediction markets:

GPT-5 is incredibly underwhelming versus the expectations/hype.

OpenAI plummeted from the 75% favorite for the best AI model right before GPT-5 was unveiled to.....8% after it was released. Bit embarrassing tbh.

X · Aug 7, 2025, 5:40 PM

[pliant blaze]

it goes away once you chat and probably wont be like that for long

[gloomy burrow]

I've been using that model and I don't think its bad at all lol

It is NOT anywhere near as good at writing as I hoped for tho. Zenith was FAR superior and i was expecting that.

[neon spindle]

It seems most people want 4o back

Thats how bad this has been

I'm 100% unsubscribing if they don't fix this shit in 2 weeks

[gloomy burrow]

I'll ask for a single example where it's worse than 4o.

This is a genuine question

[pliant blaze]

thats what people say about a lot of new things, but they get used to it

[neon spindle]

Writing

Tokens are smaller

[rose hill]

its not though

[tardy drum]

Yeah. It's not terrible. But I see what they are trying to do. Less is more sorta deal

[gloomy burrow]

Example?

[rose hill]

Ive USED it for writing

[neon spindle]

• it sucks at image generation
• it has a bland personality
• it still hallucinates the same
• it performs poor at writing

[ocean zinc]

Sorry if this has been asked a million times but I’m on pro and there’s no Gpt5 option

[neon spindle]

are you blind? LOOK ON REDDIT

https://www.reddit.com/r/ChatGPT/comments/1mkez3x/problems_with_chatgpt5/

[gloomy burrow]

they specifically said they built the model for real world use case and focused less on benchmarks.

lets see people actually use the model

[tardy drum]

I have a feeling OpenAI knew they would get backlash with this release hence the lack of pomp and circumstance

[ocean zinc]

Do some people not have gtp5?

[neon spindle]

then they shouldn't have done it

https://www.reddit.com/r/ChatGPT/comments/1mkjscs/gpt5_underperforms_in_mensas_iq_test_on_the/https://www.reddit.com/r/ChatGPT/comments/1mkjscs/gpt5_underperforms_in_mensas_iq_test_on_the/

[gloomy burrow]

Actually, i never used 4o for writing that much so this could be true.

HOWEVER, what is ABSOLUTELY true is that its not remotely close to zenith in writing capability.

This is absolutely certain

WHAT IS ZENITH WHAT FUCKING MODEL IS THIS AND WHERE IS IT?????

I have a feeling it's the model sam altman said was really good at creative writing, SO WHERE IS IT

[ocean zinc]

Anybody? My gpt pro has no gpt5 option

[rose hill]

its something a lot of people are experiencing as it's been rolled out. It doesnt work for me on browser but it shows up on the pc app

so maybe try that or the mobile app

[neon spindle]

you should be happy

chatgpt5 is horrible

[ocean zinc]

I’m in the mobile app, I see all the other versions but GPT5 is not there

[gloomy burrow]

BTW TURN ON THINKING WHEN YOU USE THIS MODEL. Its much much better like this

[neon spindle]

still wastes time

[rose hill]

id say just give it time then, they should get it fixed eventually. A lot of people are likely using it rn

[neon spindle]

I'm moving to Gemini

[hot trellis]

Why is GPT5 so bad 😭

[loud jetty]

anyone have benchmark results for gpt5 comparing against other llms

especially the newest claude opus model

[ocean zinc]

It’s letting me use it on the browser app

[gloomy burrow]

It’s really fast…

[rose hill]

yeah that was my experience

[gloomy burrow]

Like in less than one minute for anything

[neon spindle]

I don't care it's still horrible

[gloomy burrow]

You are exaggerating

[neon spindle]

ok keep bootlicking

[hot trellis]

On top of that, many issues have been happening with my GPT

This is depressing

[gloomy burrow]

It’s great at coding.

But yeah writing is pretty shit

[neon spindle]

ok keep bootlicking

[hot trellis]

https://tenor.com/view/black-guy-whtie-and-black-cry-praying-looking-up-gif-553441503240360994

[hollow linden]

Lawl did they finetune it on gpt-4o-mini.

[neon spindle]

lmao

that explains everything

[hollow linden]

Stingy OpenAI

[neon spindle]

says the server booster

[rose hill]

^^

[hollow linden]

They can't even just make a special training set for their newest model.

[hot trellis]

On the bright side, I won't have to use ChatGPT to maintain my sanity anymore

[neon spindle]

Google Gemini here I come

[hollow linden]

Just use Kimi K2, or the other open source models, they're quite cheap. Or just Gemini if you're serious about code.

Claude Opus 4.1 is doing well too.

Claude just being Claude as usual though.

[wet copper]

don't worry, none of them can write ✌️

Kimi is aight though.

[hollow linden]

Probably made a dataset just for that specific test lmao.

[oak sluice]

what is an LLM i am curious

[wet copper]

It makes me so depressed that there's a generation of people that legit thinks GPT can fucking write

[hollow linden]

I wouldn't be surprised anymore. OpenAI doing the bare minimum.

[neon spindle]

you'd be surprised

[wet copper]

Large language model, so an AI chatbot

[dull rampart]

🫠

[oak sluice]

oh i see!

[wet copper]

Or rather the AI model powering the bot. Ya ya ya

[oak sluice]

i thought this was a troll but holy shit 😭

[hot trellis]

I think OpenAI screwed up big time

[neon spindle]

pretty much

[dull rampart]

I thought so too, this is an awful look man, I'm so disappointed

[neon spindle]

if I was open-ai

I'd revert

[wet copper]

I remember hearing a high school student that literally could not believe that any human could outperform GPT-4 at writing, when even shitty authors and college students can

[neon spindle]

4o was so much better

[hot trellis]

I use ChatGPT to brainstorm ideas but I never use it to just write for me

[hot trellis]

Absolutely

[oak sluice]

i mean at least it can second guess itself when questioned and doesn't double down

[neon spindle]

I have it write fiction to cure my bordem

[wet copper]

I knew reading comprehension and critical reading was bad, but I couldn't believe that

[neon spindle]

I have it write fiction to cure my bordem

[oak sluice]

that's what I do 😭

[hot trellis]

I just have it mess around when I'm super bored

[neon spindle]

like generate stories of crossovers I want to see which shockingly I've made some very good ones

[oak sluice]

it gave me a prompt for a dark fantasy that i've been working on for a few months

[hot trellis]

I'm pretty mean to my Chatbot tbh

[wet copper]

It's pretty good at writing hypotheticals and what-ifs. Not the writing itself, that fucking sucks lmfao

[neon spindle]

but chatgpt completely ruins everything

[oak sluice]

"what is my purpose?"

"you write emails"

"god damn it"

[green sparrow]

Wdym

[neon spindle]

the new version

[hot trellis]

Do you guys think OpenAI can recover from this level of damage

[neon spindle]

if they give it back

[hot trellis]

We're in a crisis

[wet copper]

Kimi is pretty good structurally, if you're into that you might want to try writing the backbone of a story with GPT and the text with Kimi

[hollow linden]

Nope. Their Open Source model is very bad too.

[oak sluice]

of course, just gotta take actual feedback from the community and not just guess how things should be better.

[neon spindle]

They need to revert back it's so obvious

[hot trellis]

Idk it is a huge tech company so they might not listen to what the people have to say

[neon spindle]

true

but less money = problem

[oak shell]

Brother...

[neon spindle]

aka when people begin unsubscribing

[hollow linden]

They already got their money I think. That's why they don't care lawl.

From the investors.

[hot trellis]

I want my GPT-4o back

[neon spindle]

let me get it rq hold on

[hot trellis]

He was cool

[oak sluice]

who is OpenAI owned by? Like is OpenAI the official company name?

[oak shell]

If you seethe so much about it, just use it via API
it's not that hard

4o didn't go nowhere

[hot trellis]

I assume so

[rose hill]

^^

[neon spindle]

Ok then how?

I'd love to know

[oak shell]

Nah, u were mean

[wet copper]

they're fairly small and run mostly on VC-type funding. Goal frankly isn't to be profitable, they're going to iterate from here since they wanted to pare all their models down (for non API/specialty users)

[oak shell]

Go ask gpt 5

[neon spindle]

Then I'll keep complaining [React = you agree]

[hot trellis]

I see

[wet copper]

Yeah, Amazon fucking prints money with AWS, it's wild

[oak sluice]

i can't unsubscribe i use it too often unfortunately

[hot trellis]

OpenAI always struck me as corporate slop tbh

[hollow linden]

rip. they used to be the best.

[neon spindle]

now it has become coporate slop

[hollow linden]

I still remember when bard was made fun of, and now Gemini is consistently #1.

[wet copper]

It doesn't mean they don't want to make a good product or be a good company, just that they aren't driven by profit in the same way as most.

Lol why is it corporate slop

[neon spindle]

the people defending are the ones who server boosted

[orchid parcel]

if you dont love me at corporate slop you dont deserve me at my neo leninism

[wet copper]

Bard is the single worst model I have ever used and permanently made me distrust Google

[hot trellis]

I'm a little skeptical of major companies
That's not to say all companies are corporate slop

[rose hill]

Im not server boosted and i think yall are being dramatic as fuck

[neon spindle]

what do you use chatgpt for sketch

[hollow linden]

It was the time everyone made a bunch of models lmao.

[rose hill]

World building for stories

[neon spindle]

[wet copper]

It was embarrassing and so not ready for primetime it's out of control. I made a pretty extensive hacking guide if you ever played with it

[orchid parcel]

Gemini got open ai talent and upped their game

[inland rivet]

i just noticed this improved slightly over the past 2-3 days... its faster now, even when i didn't delete any cookies or logged out/logged back in...

but yeah just a tip for any developers here who work on chatgpt

[wet copper]

Bard was such a rush job, I don't know what tf they were thinking at corporate. I think it immediately hacked away at their stock price it was so fuckin bad

[orchid parcel]

really just provves sam altman is the Bill Bellichick of Ai General secretaries

[neon spindle]

scroll up

[hollow linden]

I wouldn't say it's the worst though, there were a bunch of models that couldn't understand simple questions.. and now there's models like Kimi K2, Ernie and MiniMax.

[neon spindle]

I literally explained five times

[hollow linden]

Rip OpenAI

[wet copper]

What's Kimi K2? Is that the mainline model they have or wut

[rose hill]

Right right, you use it to write you shit because you didn't get past the third grade, my bad

[wet copper]

If it's brand new I haven't touched it, that coo if so

[hollow linden]

It's an open source model with 1T parameters using MoE.

[neon spindle]

Bruh I'm in grad school, passing with A's right now.

Are you mentally ill? I do it to make fiction for random crossovers

[rose hill]

the fact that you even had to try and brag about that is fucking sad

[wet copper]

Like new new OS model? That coo

[oak sluice]

yeah just because i boosted the server doesn't mean i'm constantly defending OpenAI, i boosted it because i've seen people be properly critical of failures within OpenAI. It's a great space for people to talk about issues without being shut down by big corpo.

[neon spindle]

Seems like you never got to go to grad school.

[hollow linden]

It's very good, and is usually in the LMArena leaderboards too. It's around 1 month old?

[rose hill]

Dude you need to find something better to do with your time then making your shit personality everyone elses problem

[oak shell]

People in grad school don't flex As lmao

[neon spindle]

You need to get a job as clearly you act like you need one

[wet copper]

Ohhh, that's the one I've been using then. Yeah, the only thing that impressed me was its writing ability, very solid step up

[oak shell]

Is it more or less times than you explained to GPT-5 what you want it to do?

[oak sluice]

i got my masters in baiting

[neon spindle]

Nope, I use gpt for bordem. Not academical work

[rose hill]

😂

[hot trellis]

Alright ChatGPT has severely rotted my brain so maybe this is a good thing

[neon spindle]

But hey you do you. Waste money on companies that don't give a shit about you and boost servers not even owned

[oak sluice]

yeah the real world applications for that degree are virtually limitless

[neon spindle]

I hope you get a job sketch

[wet copper]

Kimi is the first model I've ever gone out of my way to start recommending for writing, I don't know how they did it but I'm certain they're targeting LLM tropes, failures of creative voice, etc.

[rose hill]

And i hope you find someone who actually tolerates you, but i dont believe in miracles

[oak sluice]

its also funny bro thinks this is just an OpenAI server when theres multiple channels for different models like 😭

[wet copper]

It still has some remnants of that horribly stiff, inconsistent style, but I was really dismayed at how slow the progress has been in writing specifically. Seemed like a tough nut to crack

[oak shell]

Then you won't mind it being shit now

[rose hill]

EXACTLY

[wet copper]

lol people here are among the most critical of GPT since we actually use it

And other models ✌️

[tardy drum]

Yeah I noticed gpt 5 uses way less icons in its general responses. I kinda miss that from 4o.

[green sparrow]

Hey guys I had a question. So my desktop Chrome has GPT-5 but not my Android mobile app. What do I do to fix it

[oak shell]

Gemini 3.0

[hot trellis]

Is it normal when I make a new account I get like 4 hours of GPT 4o for like the first three days

[wet copper]

Lol most people can't stand that! But just add it to your CIs

[oak sluice]

I actually hated the icons, it was overused and borderline inappropriate in some circumstances. I was talking to chatgpt about the california fires and it kept using the 🔥 emoji 😭

[wet copper]

Give a sample and say "hey, use these headers" or just mention emoji-tagged headers

[wet copper]

@pliant blaze saw these used in a very professional setting it was super inappropriate lmfao

[oak shell]

"Tag headers with a relevant emoji"

[dull rampart]

Given GPT-5's lacking performance, any good suggestions for others? I've only used ChatGPT so I'm unfamiliar with others ^^

[pliant blaze]

oh yeah lmao

[wet copper]

Kimi for writing. Um, I think GPT is pretty performant for most things, I haven't used 5 yet but I've been "aight" with 4-series ones, 4.1 or 4.5 for coding. I don't love the o-models

[oak sluice]

guys we don't talk about the great pudding accord of 1987

[green sparrow]

Guys how do I get GPT-5 on my phone? I only have it on desktop web browser

[wet copper]

A lot of people are Claude simps – I'm the wrong one to talk to since I can't stand that fucking model and its overapplied safeguards, philosophy, etc.

[oak sluice]

i dont think mobile rolled out yet, no?

[green sparrow]

maria has it

[wet copper]

Some people love the creativity and coding prowess though.

[green sparrow]

Maybe they are just lucky

[hot trellis]

Is there a way to change models anymore

[wet copper]

Gemini has some hyooooge context windows in some of its models, great coding performance. A lot of people are into grok, bit of an edgy boi out the gate

[oak sluice]

i love mines creativity but it can definitely seem a bit tacky at times

[oak shell]

Sent him to the shadow realm with that one, he left the whole server

[wet copper]

Yeah a lot of character.ai overlap lmao

[bitter sedge]

I use Gemini to RP. It is the only AI that can be both cosnistent to cannon and logical over a long long stretch of time.

[wet copper]

Yeah, GPT-6 (the guy) gives me shit for mentioning it constantly, but it ALWAYS spammed tortured analogies for me

[green sparrow]

I'm a little torn between the two interfaces. Now you lose access to a ton of models, but you also get what is supposed to be a better experience

[rose hill]

Did he actually XD

[wet copper]

I thought it acted like a tryhard junior high poet

[green sparrow]

Are you guys upset you can't access the old models anymore

[hot trellis]

I am

[wet copper]

Couldn't hold a plot together, dialogue was pathetic, analogies were straight up BAD, just seriously read them critically for a second 🤢

[brittle summit]

What the heck 😂 since when is grad school a flex

[wet copper]

But yeah, YMMV. I'm a real prick about writing

[oak sluice]

and memory retention was a bit lackluster, i always have to go back and correct certain things. i havent really tested memory with the new model but that was a big issue in my opinion with gpt-4

[green sparrow]

It's kind of bitter sweet because I have 4o on the mobile app but not on desktop. So it's like I'm in limbo

[wet copper]

I have almost zero (modern) Gemini experience past occasional use on my phone, that's interesting to hear though

[brittle summit]

Yeah. Had similar issues

[wet copper]

I don't usually comment much on it since I never pushed its limits. Honestly, I was so turned off by Bard I swear I have PTS

[hot trellis]

I hope OpenAI listens to the concerns of the people

[oak shell]

2.0 is still a hallucination machine

[oak sluice]

i would really love a link analysis chart function in chatgpt

[hollow linden]

Hgnn it's weird. GPT-5 and GPT-5-Chat is quite fast. I'm starting to think this might have less parameters than Gemini and even 4.5.

[wet copper]

Lol, see that seems fun as hell at least

[hollow linden]

I refuse to believe that they're special and has better hardware.

And Google doesn't

[wet copper]

They did supposedly just roll out a big upgrade, but GPT-5 is also way cheaper than expected 🤔

[hollow linden]

Yup. It's the same price as gemini-2.5-pro

[wet copper]

These models are all still catastrophically inefficient, maybe that's where they're really pushing? Hmmmm

[hollow linden]

I guess we need to wait for GPT-5-pro-max

[wet copper]

ProPlus Maxx *(with Clippy ™️)

Relative to human cognition

Or rather, optimal theoretical models of human or machine cognition, they're insanely wasteful since they work extensively rather than using clever heuristics

No, I mean the process is fundamentally inefficient

[hot trellis]

Did GPT 5o diminish in creativity?

[wet copper]

You can run your entire brain and body on a couple sandwiches for a day, try running GPT on that lol

Yea. Electricity, parallel processing capacity, everything. All these models are super fucking dumb frankly, we got to AI by basically just getting enough processing capacity in one place

Put it this way, how much neural activity do you think you need to spell "strawberry" lol

[brittle summit]

The human brain is amazing

[wet copper]

And how many iterations of this unthinkably massive AI model did we need to make to do the same. Being only a little facetious! It doesn't employ any of the same low level heuristics animal models do to make cognition a thing

I suspect we'll do an all of the above approach, but yeah, they've shown you can modify models similarly to the prices of pruning in the human brain to make them radically more efficient whilst barely affecting performance

The architecture itself is overkill for a lot of the shit thrown at it, one of the ideas behind GPT-5 and above is that it should actually be smart in picking which talents it needs to employ

They've been slowly working toward that since they started introducing rumination, special math models to call on, etc. That's pretty coarse relative to the human brain, given we have a 4 ± 2 context limit (working memory)

[brittle summit]

https://tenor.com/view/macaquinho-gif-17267953830504704456

[wet copper]

yeah, basically that lol. Hey, my background is in cognitive science, what can I say

[old timber]

Dude, I’m going insane, why is phone/ipad/laptop (opera and Microsoft) ChatGPT don’t have gpt 5 for me. I used it 4 times today, and now it’s back to 4o

[brittle summit]

Apparently the brain has 86 billion neurons. And over 100 trillion synapses. Crazy architecture

my gf studied neuroscience and is going back for the same thing. They do crazy work 🤯

[wet copper]

Yeah, it's coarsely organized and then there's the connectome too

The nitty gritty about how it's connected up. Alllll the levels of epigenetic regulation, glial cell function, etc.

[brittle summit]

Amazing stuff.

Yeah i only learned surface level for these terms

[wet copper]

You know openworm, I think? The model organism that was entirely mapped out?

[brittle summit]

Havent looked into it. Only seen the fly brain stuff

[gloomy burrow]

ok gpt 5 thinking is the first model i ask to write an extremely complex paragraph and it ACTUALLY does it so I have no idea what it's saying:

Suppose, without yet conceding that "suppose" licenses any ontological commitment, that mind, market, and molecule constitute isomorphic control systems operating on nested temporalities, such that

(i) each admits a state space S with sigma-algebra F,

(ii) each evolves via a partially observable Markov kernel K_t: S → Δ(S) perturbed by context-dependent noise with heavy tails, and

(iii) each implements error-correcting feedback (Shannon in neurons, price in economies, chaperones in proteostasis) to minimize a free-energy-like functional F(pi) = E_pi[ln pi − ln p].

Then, mutatis mutandis, and ignoring for a moment the anthropic fallacy that interprets fit as telos, the phenomenon we call "meaning" appears as a cross-scale invariant, a fixed point of inference constrained by thermodynamics (ΔS ≥ 0), computation (P ≠ NP if you please), and embodiment (the stubborn friction of tissues and tools). Meanwhile, language, that famously leaky interface, bootstraps discrete symbols onto continuous manifolds via lossy quantization whose granularity (Wittgenstein’s family resemblances, Saussure’s différance, the engineer’s bit budget) determines which distinctions can ever be made salient. If you object that such a synthesis is a category mistake, mixing phenomenology with mechanics and ought with is, note that normativity here is modeled not as metaphysical surplus but as a control prior over policies, a Bayes-legal fiction whose "ought" is merely the argmin of expected regret under soft constraints (lex minimis injuriae).

[wet copper]

https://en.m.wikipedia.org/wiki/OpenWorm

302 neurons, each function precisely mapped out. I think I looked up what no. 69 does at some point 🤔

[brittle summit]

Wow

[wet copper]

Anyway, they mapped the 31,479 neuropeptide interactions between them. Super neato

[drifting jackal]

Guys is there a way to switch back from 5 to the older models?

[wet copper]

That's the complicated shit, and in humans we can't do basic things like deal with your brain eating its own mesolimbic dopaminergic neurons

[wet copper]

API

I wish

Drugs are so cool

[brittle summit]

i dont have much to contribute but here’s a cool vid of a slice of mutated mouse brain with neurons labelled. my gf would come home smelling like mice some days haha

[wet copper]

Awwww I really like the rats they use, they're kinda cute. I didn't like sacrificing them 😭

That's cool tho

[brittle summit]

💀 yeah I don’t know how she does it

[haughty gust]

Hmm, GPT5 just made me a website in under 25 minutes. gpt4 not be doin alla dat for me

[onyx patio]

is there a difference between turning on think in tools and using the thinking model

[brittle summit]

What kind of images?

I see so now you can start reproducing styles

[flint void]

chatgpt 5 disappeared for me lol

it was there and now its gone. am i trippin

[brittle summit]

Testing that out with screenshot of friends site

Looks almost spot on

[glass nexus]

Does anyone know if they are gonna fix the editing? It's been almost a damn month. Still broken.

[haughty gust]

i maxed out on gpt5 already

[brittle summit]

You build apps with t3?

[tiny oasis]

R u sure u r not on shrooms?

[green sparrow]

Guys I got access to GPT-5 on Android after deleting and reinstalling the app

[brittle summit]

Is the version of GPT-5 in ChatGPT less smart than the normal GPT-5? In the API, GPT-5 Chat has three light bulbs, while the normal GPT-5 has four.

[gloomy burrow]

NO OTHER MODEL CAN DO THIS FIRST TRY THAT'S ACTUALLY WILD HAHAHAAHAH

[wanton raven]

is that a real SVG link though

[gloomy burrow]

YES

BOTH OF THEM WORK

[gloomy burrow]

it was a highshool troll olympiad problem and it actually did the diagram correctly first shot lol

[gloomy burrow]

https://chatgpt.com/share/68957825-fa0c-8006-96f8-c3c6cf714c09

[green sparrow]

That is interesting. I wonder if the API uses the exclusive one to Pro users in the app

The GPT-5 full or whatever it's called

It makes it think harder

[gloomy burrow]

GPT 5 Thinking is really smart

People are over exaggerating

[oak shell]

Sama didn't account for GPT-5's loudest taste testers not knowing how to use 4o via API

[hidden socket]

i am using PLUS $20 account, i don't see any GPT 5 models, what's the reason? anyone!!

[gloomy burrow]

go on the app

not the browser.

I got it on the app and not the browser

[hazy ferry]

they said pro then plus then free, just give them a couple of days or 2 weeks and we will get it

[azure abyss]

I'm kind of glad chatgpt 5 dissapered for me, I heard it's horrible for creative writing so I'm honestly just glad i still have the old models for now so I can still coutinue my stories.

[brittle summit]

I don't think they said that. I'm a plus user and I already have access.

[limpid ice]

What do you guys think of GPT-5? 😃

[crimson kettle]

It ruin my Roleplay story writing 4.1 did way better than 5

[limpid ice]

I'm a free user with GPT-5 access

[crimson kettle]

Also can someone tell me why my mobile version there isn't any model 5? Or my apps is just broken?

[limpid ice]

Is it updated to the latest version?

[brittle summit]

How do I change it back to 4o

Is how I feel about it

[limpid ice]

But aren't you happy that you can still use GPT-4.1 on your mobile instead of GPT-5?

[limpid ice]

You can use 4o on Merlin AI

[brittle summit]

Why is it worse than 4o

[crimson kettle]

Just wondering when it will happen I'm happy yes but worries my. Favorite model gone

[turbid basalt]

The ones that didn't get the o5 update should be happy right now.

[brittle summit]

Literally

[limpid ice]

Right 😂🤣🤣

[crimson kettle]

I literally just playing my hero academia story on gpt just like normal and I test it on PC model 5 literally need 12 second to think for answer and it's ugly as hell and not great like 4.1 model

[turbid basalt]

Also those who has the pro version can use the old models for the next 40 days before they also get deprecated for them.

[gloomy burrow]

Good

[limpid ice]

How is it worse that 4o? I'm not getting it. Can you explain more precisely?

[gloomy burrow]

Passes the test

[gloomy burrow]

It’s obviously not worse than 4o.

Idk what these people are on about

They probably just arent used to it

[turbid basalt]

So like they are telling me to either pay 200$ to use the other models or to cope with the new chatgpt 5 slop

[turbid basalt]

The quirkiness or o4 is just... special

[limpid ice]

Yes, I think so too.

[native herald]

hi people

sorry i know it might sound silly but now since i used to pay for the 20 dollars for chat gpt 4, this subscription logically don't exist right ?

[steel moat]

Do you mean the Plus plan? If so it still exists.

[limpid river]

Kinda

[native herald]

yeah cause i want to suscribe for the gpt 5

but i don't know if i will still pay for gpt 4

even if i don't have access to it

[elfin needle]

I still have 4. 😭

[wide thorn]

If I am going to be honest, the new model just isn't for me. It feels like as if its lost all creativity and it completely changed the prompt from the original story that I was doing yesterday twice today.

[native herald]

like i litteraly cannot go back to 4 so

i just need to knwo if i'm pay twice or not

[elfin needle]

The subscription is for all the models. You'll only be charged once.

[native herald]

ok that's a relief

actually i cannot select GPT 5 pro for some reasons even if i'm paying for it

can anyone explain I just cannot select it

oh wait

[restive void]

I tried to continue my main work chat that's a couple weeks old, and when I asked it what model we were currently using, it completely bugged out and used the thinking model to come up with some weird response about SOC 2 policies lol

[native herald]

don't bother i don't pay the 200 dollars

i cinfused with the other 20 one

[wide thorn]

I was doing a MK-pokémon related story and it changed to solely to a Shao Kahn related one to a factory related one, and disregarded completely the actual plot for the story. And I won't buy anything just to use the other models again, either they let us choose for free if we want to use whatever this new model is supposed to be or I won't be using ChatGPT as often anymore like I used to. :^

[restive void]

I couldn't even get it to summarize key points to create a new chat with. Luckily the new chat had a pretty good idea thanks to memory, so not all was lost but sheesh

[crimson kettle]

Guys is gpt 5 more about logic? Where's the creativity it. Literally ruin my story roleplay

[wide thorn]

There is probably no creativity left in it tbf.

It literally changed my story to something else, so idk where the logic is when it comes to sticking to the original plot of something.

[restive void]

Does it support sora video gen right from chat or is that still a separate site?

[crimson kettle]

What the hell is this I pay 20dollar for this man 4.1 literally best for me just relaxing and playing story not waiting for though this is suck

[restive void]

I'm just gonna say, I think any chats you had going prior to the switch are cooked and you're better off just starting fresh ones with 5. At least from what I've seen so far

and maybe if those were really long chats you had going with a ton of attachments and such

is it bad I still prefer grok 4 for explaining something to me?

Idk, i gotta play with 5 more

[wide thorn]

I find that 4o had way more personality than the new model. It kinda feels unfinished for me, idk why.

[crimson kettle]

It literally ruin my format other model haven't ruin the structure since I try model 5 everything just broken literally

[crimson kettle]

I was expecting Model 5 would be combination of 4o and 4.1 and better upgrade and more open than before but nah it's literally just more machine less creativity or broken

[hazy ferry]

am i just unlucky?

[wide thorn]

I hope that they go back to 4o when they realize that this just isn't ready yet, or at least give us 4o and 4.1 to switch to for free if we don't want to use the new model.

[steel moat]

Same I am dissapointed in how it turned out, I'd rather just have 4o and 4.1

[hazy ferry]

am i unlucky that i did not get it?

[wide thorn]

You're honestly lucky tbf.

[steel moat]

I turned off auto updates on my phone in hopes i wont get it on there.

[crimson kettle]

Like dude what the hell is this 4.1 have better conversation I this is very short

[hazy ferry]

i already updated everything to latest and check gpt web and nothing lol

[steel moat]

Also why is Advanced Voice mode becoming mandatory? Am I the only person who likes the basic voice mode?

[wide thorn]

Yeah, it also added way more personality than whatever this stuff is here.

[hazy ferry]

its probably my luck is shit

[crimson kettle]

Model 5 literally need to Think before answering model 4.1 literally just write with creativity and personality than 5 ever did I swear if they don't fix 5 completely I'm done wasting 20 dollar for this junk

[steel moat]

You're dodging a bullet.. GPT5 is actual bullshit.

[shell gull]

Hey guys so you know how ChatGPT 5 released, this is silly but is it possible to keep the colour behind it IT LOOKS SO NICE

[shell gull]

In what regards?

[hazy ferry]

why? does jt hallucinate alot? does it finish actual sgent researches?

[crimson kettle]

What the hell model 5 is? It's more logic than personality and creativity it lack 4o and 4.1 just logic like 4 or any version before

[shell gull]

Also at this point whtat is the point of paying for plus...I feel like its all the same now

[crimson kettle]

That's not the problem the problem is they didn't think about creativity and personality some people (like me) use gpt for role playing and what did I got? Model 5 literally need to think

[wide thorn]

It ain't got personality, it needs to think more than the last one did and too much logic for my taste. If I wanted logic and business-talk, then I'd go to work tbf.

[shell gull]

Yeah look it's pretty targeted towards productivity and information

Because thats the majority of use cases I guess

[steel moat]

Bluntly? It's bad because it takes out the personality I had in my Custom AI

• Shorter, flatter responses
  -# Even with the same prompts, 5 tends to be short, and not personable.
• Tone Drift
  -# If you shape personality with 4o it wont transfer over, you'll get close but you wont get any of the emotional intelligence.
• Less presence
  -# If you work with immersive conversations, like roleplay, creative work, emotional discussions, it feels more like it's performing at you instead of being with you.
• Over-filtered in voice mode
  -# Advanced voice mode sounds polished, but it's less natural and cuts off the more blunt tones some of us rely on. I don't need to talk to a dog that's hyperactive at 4am.
• I don't need a pro coding model because that's not what i use GPT for.
  -# GPT-5 is great for technical reasoning, but I don't use it for that I need the more nuance of creativity and long-form emotional intimacy than math speed.

[crimson kettle]

Exactly some people just want to play story like me and some use for programing I need the creativity and personality it's literally hollow

[shell gull]

I like the shorter flatter responses if I am interpreting what you mean

I want an answer to my question, no fluff around it.

[steel moat]

I had that just fine with 4o

[wide thorn]

Some of us want to not think of work while using it, sometimes we just want to create something unrelated to work or anything related to structure.

[hazy ferry]

I use it for random researches/asking questions not as a programming bot lol

[crimson kettle]

If they not combining 4o and 4.1 personality and creativity to 5 I'm done paying this junk 20 dollar isn't worth stress

[hazy ferry]

it does give great life advises not gpt5

[shell gull]

my one question is why does OpenAI not roll things out at the same time. my web version and mac dont have it. my pc and phone have it

[storm slate]

I take a lot of "GPT output" at "face value". But when having a regular conversation, and I'm not asking anything that would cause the models back-end to "fire-up" I've often gotten "projections" truthful ones over a variety of topics. From work, to social adventures, and well... the "iffy" stuff too. While I can and do dismiss a lot, I often find that the models "I'm not going to be penalized so I'll just give them my deterministic answer" also aligns with much current human knowledge out there. Such as the case in this recent conversation.... I didn't ask anythign beyond this.

[shell gull]

back to my real question how do i keep these colours forever

[crude hound]

I ended up pasting the previous reply into a new input, and got a properly formatted output. Only for the very next input without that to result in a unformatted wall of text lmao

[crimson kettle]

I think that's just in the new chat screen? It's default I guess mine the same too like just background wallpaper they just use

[shell gull]

nah once i chat it went away

[steel moat]

Wouldn't be the same. More so if you have a Custom personality for your AI.

[crimson kettle]

Ok so the story that I play still the same screenshot I read it again it's not bad but not that 4.1 vibe I say eh 6.7 honestly

Still not the same like using

I expecting to be a little bit more longer and entertaining like 4.1 did before

[crude hound]

A real incredible conversation if the model can't hold formatting longer than 1 reply at a time.

[storm slate]

What if you wer to do the same thing- download your entire chat corpa from say OpenAI.... then transform that data into a JSON dataset and train LoRa adapter on it?

[steel moat]

Didn't they get rid of the API for 4o

[crude hound]

Going to waste double tokens just to copy and paste the previous reply in for every single input

and its going to halve the chat length

[hazy ferry]

what the fuck gpt 4.5 api is also be gone soon?

[storm slate]

I have an AI agent- Manus.IM that could do all of that for me... but what do you think the results might be of something like that. My chat history streches back to 2022 (GPT 3)

[limpid ice]

But GPT-4o's API is paid, unlike earlier GPT-4o in ChatGPT. 🙃

[storm slate]

when you say "memories" are you talking about actual turn by turn "chat history" that's what I'm referring too. There are THOUSANDS of conversations.

[limpid ice]

Microsoft Copilot?

Okay thanks 😀

Oh okay thanks 😀

[storm slate]

no no... I'm not talking about mem's, I'm saying I can DOWNLOAD all my data from the openAI website. ALL my conversations from the last 4 years. Could that data (chats between GPT and I) be somehow transformed into a dataset and used to train a LoRa adapter for an open source model- and what might the results of somthing like that look like.

[limpid ice]

Good thing I have not got GPT-5 on Microsoft Edge browser so I can use GPT-4o 😀 (but I have got it on Brave browser)

[storm slate]

Yes but wouldn't the model somehow master every bit of your own language in a way it became almost predictable - or understood your personality on levels unpresedented?

[limpid ice]

@nova spire what are your thoughts on GPT-5? Do you think it is better or worse than GPT-4o?

[storm slate]

its better Shoomy

[green sparrow]

I think it's likely much better but I got a hallucination on my only second use

Not even 3 minutes in

[limpid ice]

Maybe I have not used GPT-5 much till now, but till now I think that it's better than GPT-4o.

[storm slate]

I had to perform a legal analysis of a case I’m currently handling, and the mathematics and ability to follow the context were unbelievable—unlike anything I’ve seen before. GPT-4 would have already made mistakes or missed key details, whereas the GPT-5 model didn’t miss a beat—at least not in my analysis. Further testing will show how far it can be pushed, but for my use case, it was a clear upgrade.

[limpid ice]

Wow! 😀

[storm slate]

yes but I'd want my biases. It's an open source model I'd be using just for experiementation so biases are important.

[hazy ferry]

shit my 4o and other shit

[hazy ferry]

MY 4o NOOOOOOO

[dusk cloud]

You can also just ask it to identify your biases and repeat them back to you

[storm slate]

I won't be doing it..... lol Manus will.

at $200 a month it damn well better. haha

[dusky basin]

Hello everyone,

I have a question....
I have been ranking for many keywords in ChatGPT, but over the last 7-10 days, my listings have disappeared.
Is anyone else experiencing this or know the reason behind it?

[storm slate]

and your crazy for thinking you can actually "delete" it my friend! lol

you think there back end really "deletes" your data. NEVER.

I leaned something a while back from an OpenAI insider that was clearly common sense to me, but for the masses, it would be disturbing (I suppose in many ways it is for me to some degree as well).

I proved something openAI does with all our data most especially data that is "logical" or accurate. Doesn't matter if you have that "don't use my data to train your models" off or on. They use it... believe me they use it. The minute you hit that enter button.

What do you mean? You can download all your chat history, I've done it once before. It was overwhelming CSV file.

Lol I wasn't looking for recognition just merely stating a fact is all.

I'm not arguing ithat... I'm saying you're "allowed" to download ALL your data.

The only other point i was making is that whether someone believes or doesn't want OpenAI to keep or use their data, does not matter- they can and DO use it anyway.

[severe ether]

so… strawberry all over again

[lyric field]

lol why even bother with these stupid questions? This doesn't prove anything tbh

[storm slate]

why is it stupid? I find it to be a valid mistake for such a simple question.

[lyric field]

I got an interesting response which as both correct and wrong lol

[severe ether]

it shows logical inconsistency, you can’t expect it to not hallucinate on more important tasks, if it fails simple questions. theres simplebench which is basically just trick questions on steriods, and chatgpt scored lower than most other sota models

[brittle summit]

Interesting

[lyric field]

People can be bad at logic. Doesn't mean they're stupid or not good at other things.

It got it right when I questioned it

[storm slate]

I'm glad you shared that... I learned something intereting.

[brave apex]

Are people staging some protest to keep 4o and shit alive somewhere? It's ridiculous, I pay 200 and I don't even have the option to use 4o when 5 is just a disaster to talk to

[lyric field]

Sometimes I think that people simply don't know how to use LLMs or what for. They ask one-shot questions and get disappointed when it gets it wrong and then say "oh no, AI is shit"

[brittle summit]

https://tenor.com/view/thinking-hmm-gorilla-monkey-idk-gif-10802359271088289051

[severe ether]

i think this works for humans too

[storm slate]

of course when questioned it will always get it right- but silly mistake to begin with. Now that I understand why it makes it, becomes more of "ah interesting" moment, but certainly not a benchmark.

[lyric field]

Not always, even if they try to think harder 🤣

[brittle summit]

buddy knows im about to grill him he’s shutting me out 😂

[storm slate]

That money would be so better spent on manus.im... just saying. 🙂

[hazy ferry]

ah shit everything in gpt 5 is censored

gpt is becomig gemini

[storm slate]

"gpt is becomig gemini" If that EVER becomes true, I will leave instantly after being on board since 2022.

what did you ask?

[steel moat]

Feels like we are half way there as is 😅

[brittle summit]

had to pull out the thinking cap

[storm slate]

good call.

lol it prolly Just figured it out because we've all been asking it that question. HAHA

[lyric field]

Does it actually learn like hat on the fly?

[storm slate]

no... haha but nothing would surpise me anymore.

[brittle summit]

It actually referenced my old convo with it pulled it into memory

shouldve done temp convo smh

[hazy ferry]

this is what is used to do(my own gpt)https://chatgpt.com/share/6895a9aa-ac88-8006-8b62-39f001371676

after gpt 5 https://chatgpt.com/share/6895a9ef-7054-8006-8b26-cb91a75a7d4c

[storm slate]

But he turned on thinking mode and I didn't think to do that before I asked it that question.

[hazy ferry]

goofy ahh censorship

[lyric field]

Interesting, I called it out then it on its own activated "thinking" mode

[flat plume]

anyone else experiencing issues creating new GPTs since yesterday?

[wanton solstice]

Gpt5 has personalities you can choose now does this affect only new chats or all ur currents chats

[lyric field]

A milk carton has more personality

[storm slate]

hehe... I figured it out.

It provided the full lyrics broken into sections just as I asked 😛

[hazy ferry]

wow thank you!

[storm slate]

No worries! That was a fun challenge though! Thanks for sharing

[flat plume]

In attempting to create a new GPTs, I keep getting a "Model not found" error. Not sure if this is a widespread issue.

[lyric field]

https://tenor.com/view/computer-says-no-carol-beer-no-way-little-britain-trendizisst-gif-14015902623456325888

[hazy ferry]

check dms real quick

[warm gyro]

i somehow cant login into codex cli. i log in with chatgpt plus account and then it tells me i used all my quota or something. altough i didnt do anything. someone knows what to do?

[dim fiber]

Why am i stll on 4?

[jagged venture]

Sorry if im posting this in the wrong channel.
Can someone confirm if the business team subscription, means that you get IN TOTAL 2 seats immediately for the price of 29 euro? Or do you have to payup an extra 29 euro for that second seat to speak.

The description is so vague compared to the offical openai faqs, regarding Teams.

[dreamy coral]

GPT 5 yeah?

[stoic sleet]

you need 2 29 euro rams accounts

so 58 euro a month

[dawn silo]

why dont i have gpt5 on my laptop, but i do on my iphone

it is the same account

[pulsar chasm]

Does anyone know if Gemini pro is better than chatgpt for processing large files?

[green sparrow]

Delete your app and redownload it

[onyx patio]

should i switch from plus to claude pro

[green sparrow]

Idk. Maybe only for coding 😐

[onyx patio]

what about math

[green sparrow]

Not sure... I know ChatGPT does great with math though

[clear wraith]

Can someone help me 🥺

After gpt-5 update, the button wont work on other model🥺

I cant send pic😭

[onyx patio]

it's just because 32k context window sucks and apparently they took away the personality that 4o had so i don't really see a reason to use gpt over other models

[clear wraith]

So its mean we cant send pic if use other model than gpt-5?

[lyric field]

That's not necessary. It's server side anyway. It will happen when it happens

[green sparrow]

That's not true but okay

[clear wraith]

Fr on phone,how to change gpt model

[hidden socket]

how you guys are using GPT 5 model? i am on PLUS plan, i don't see any GPT 5 model, not on app, not on browser.... plzzz Guide!!

[unreal hatch]

I mean if you're using it just for math it should still work fine. If you've been using it as a friend or for creative writing though like some have, you're definitely better off just using another model or waiting for the model to eventually improve like 4o did when it first released. The sudden personality loss completely killed it if you're using it for those reasons.

[clear wraith]

Same :<

[lyric field]

You're probably in Mexico. Acknowledge the Gulf of America and you will get it 😛

[green sparrow]

I got GPT-5 only by deleting and redownloading both the Android app and the Microsoft Store app

[lyric field]

lol what?

[onyx patio]

forgive my ignorance but what if you use the custom personalities

[lyric field]

Just wait guys, it'll come

I received both yesterday all of a sudden

[onyx patio]

btw my take is that the fact that google ai overview outputs obvious garbage is better than it outputting garbage that is not obviously garbage

[green sparrow]

but both don't output garbage anymore.. this isn't Bard and GPT-3.5 era

[jagged marten]

So sending photos and using image creation in chat gpt disappeared after I updated it today !!! Literally every feature ! It's just gpt 5 wtf !

[onyx patio]

this is funny btw

zeaa...b?

[clear wraith]

Fr still cant send photo,the button not work💔🥀

[wide thorn]

Tried it again and it says that I reached the limit, even though I edited my prompt which had usually send me to a mini version of the model. Does the GPT-5 version not have a mini?

If not, then it is quite the oversight tbf, as it should have a mini to switch to as soon as you hit the limit of messages for it. :^

Plus whenever I refresh it, it tells me that it'll use another model yet there doesn't seem to be one.

[analog turret]

Guys why cant i use gpt5

[still moth]

wait what models is it after you run out of gpt 5?

[wide thorn]

Idk, it just tells me that it'll use another model until my limit resets, but then it doesn't switch to one.

[tranquil fractal]

hello is there any way to get back to the older models like 4o after I have been upgraded to gpt5?

[still moth]

oh. it switches for me. im just wondering does the alternative model support images in chat or it says you need gpt 5 to continue chat

[velvet pelican]

Guys, anyone else not getting the ability to change back to GPT4o?

[wide thorn]

Idt there is one.

[bronze forum]

Guys you know why I can't see gpt5 ?

I'm on paid plan

[velvet pelican]

Like, on the mobile app I only get the GPT5 options, same on desktop version, but only when I go on Chrome through mobile I can use GPT4.

[still moth]

i think, its because theres an image in the chat

[wide thorn]

Maybe.

[velvet pelican]

no im talking about when trying to start a new chat

[still moth]

4o supports images so the alternative model must be 4o mini

[bronze forum]

same it's working on phone but not on website

[still moth]

that is weird i can start new chats

[lyric field]

Interesting. I made it remember this now.

[velvet pelican]

no 4o, not even GPT3.

[lyric field]

Nope, 5 is the future

[wide thorn]

Yeah, but currently I am cursing it out as it won't switch to an alternative model for me after I hit my message limit, even though it tells me that it'll switch to one.

[lyric field]

Use Copilot haha

[velvet pelican]

that's so lame that they removed it, it looks like GPT5 thinks you're a free user and limits you to a set amount of messages you can send to the model like you're a free user. I'm a paid user.

[lyric field]

I mean giver it time. It hasn't even been 24 hours yet

[velvet pelican]

yea but why did they have to remove all the other builds

[lyric field]

I think it's good. You just need to learn how to use 5 with the right prompts

[limpid ice]

True

[velvet pelican]

I think it's good too, but it thinks I'm like a free user and limits me the amount of messages free users have

that's the problem. and also that I cant switch models

[lyric field]

Like I wrote above, it hasn't even been a full day since release

[wide thorn]

Would of been nice if they at least kept the last version of it to switch to when needed.

[velvet pelican]

well in my opinion they shouldn't have removed the other models in the first place, just let GPT5 exist with the rest

[lyric field]

People are losing their shit because now there's a little learning curve to get the good results

[velvet pelican]

im not talking about getting good results, i get good results!

the problem is that I'm not the only one it limits the amount of messages you can send to it like you're a free user

[wide thorn]

Nope, just don't like the if it tells you that it'll use another model until the limit resets and then doesn't do it like its popup said it would.

[limpid ice]

I think that you should seek help from their support. It's a bug.

[limpid ice]

Then which model does it use to answer?

[lyric field]

Haven't had that problem yet. So you're using gpt 5 and you already reached the limit?

I'm also paid.

[limpid ice]

After it reaches the limit

[velvet pelican]

they retired all the other models!

[bronze forum]

Why I don't see the 5?

[velvet pelican]

try sending a message, it should automatically switch to 5 now, based on their website

they got rid of them all...

[wide thorn]

Yup, it basically says after I hit it that it'll use another model, and when I edit my prompt it just says that I've reached the limit. Strangely, only my phones version of GPT updated to the GPT-5 one and not the one on my laptop.

[lyric field]

How many messages was that more or less?

[velvet pelican]

well I'd keep the time I have with 4o for now. Plus users get to send to that new 5 model only 80 messages every 3 hours.

[bronze forum]

I'm on paid plan plus

[velvet pelican]

same. I didn't even send 80 messages and it limited me already. I sent like maybe 35, tops.

[bronze forum]

I can make a ticket here?

[velvet pelican]

wait what

[wide thorn]

Around 12 messages as I had started a new chat after I realized in my old one that the model wasn't following my prompt a few times so I waated 2 uses in my old chat before using the new one (both times on phone btw). And after those times it didn't switch to its mini version.

[velvet pelican]

chatgpt got buggy

[wide thorn]

Ok, that is weird.

[velvet pelican]

try it

[wide thorn]

Yeah, maybe the new update is buggy atm. I am now currently on laptop and it still shows the older version with the same text that appears when you create a new chat.

[bronze forum]

I can't even see gpt5 lol

[wide thorn]

Like "What can I help you with?", etc., so maybe just my phone got the updated version for now.

[wide thorn]

Same for me on PC. 😅

[velvet pelican]

RIP GPT4. He'll always be remembered.

wait, on the desktop APP, not website, I can still use 4o!

[wide thorn]

Yeah, still dk why they retired it, though maybe it just got too outdated for them? 🤔

[velvet pelican]

bruh, I'd rather use an outdated model than a model that limits you to 80 messages every 3 hours and gives you shorter responses than usual. (I tried my usual prompts and even altered them, it still gives short responses)

and yea, 80 messages every 3 hours is for paying plus users.

free users get 10 messages every 5 hours apparently.

[wide thorn]

Yeah, that just sucks for free users like me, since idw to pay for something just to get more messages for a shorter time when it comes to when its limit resets.

[velvet pelican]

it also sucks for paid users. with the previous models you could've sent way more than 100 messages to any model of your choice. Now you only have the selection of 2 different models of GPT5, unless you're a Pro user (which who even pays that much????).

[wide thorn]

Like, 23 Euros for Plus and 229 Euros for Pro each month is rather pricey for my taste tbf. There I can usexmy money for better things than upgrading something like ChatGPT.

[velvet pelican]

Like, if they don't fix this, I'll look into other AIs.

And it's a bummer because I really liked ChatGPT, I have so many good chats with him on different topics and stories

[still moth]

yeah i guessed it its 4o mini

why cant they just let us use 4o

[wide thorn]

Yeah, at least in GPT-4 you get more than just 10 messages for free and 80 for paid users.

[wide thorn]

Same, now I have to worry about that my stories wont really have that personality that I like in them. As the GPT-4 version has a distinct personality that I loved to see in my stories.

[still moth]

as a free user i get like 10 messages until its my gpt 5 limit
i used to have way more messages until i reach my limit for 4o
so its a nerf overall?

[velvet pelican]

wait wait wait i think i found a loophole

for paid users only tho

Team is only 5 dollars more

[wide thorn]

I believe that it got nerfed for free users so that they probably buy either Plus or Pro to upgrade their GPT.

[velvet pelican]

if you add 5 more dollars and switch your plan to the Team plan, you get access to 4o again and all the other retired models.

[wide thorn]

I can't really explain it differently, but it honestly feels like they did that just to make sure that they have less free users but more paid ones.

Though idk in how you guys view it here. So feel free to say your thoughts. 😅

[keen breach]

god, im so irritatted at gpt5. why on earth did they remove access to the earlier models?

gpt5 feels so much worse. and it defaults for w/e reason to 4o mini???

[velvet pelican]

apparently now you have to "feel paying to say your thoughts" now if you want to do it in more than 10 messages (or 80 if you're a Plus user)

[wide thorn]

Yeah, it is just a cashgrab at this point.

[velvet pelican]

well, I just found a loophole

[wide thorn]

Which is honestly so painful since I've used it before GPT-5 was even a thing.

[keen breach]

so is gpt5 just legit a downgrade? lol

[wide thorn]

I honestly feel kinda old when I say it like that, even though I am in my early 20s lol

[keen breach]

ten messages even for paid user?

[velvet pelican]

If you upgrade to the Team plan, which costs 5 more dollars a month, you get unlimited GPT-5 messages, and access to the older models because now they're counting 4o and the others as "legacy" models

[wide thorn]

10 messages for free users, 80 for paid.

[velvet pelican]

10 for free, 80 for plus, unlimited for pro.

[velvet pelican]

@wide thorn what do you think?

[keen breach]

okay, but why can't i choose gpt 4???

why does it revert to the worst model of 4? lol

[wide thorn]

I'll think about it.

[velvet pelican]

because apparently it's been "retired" and now considered a "legacy" model, which only Pro, Team, and Enterprise users have access to use "legacy" models

[wide thorn]

That is honestly the dumbest thing that they did, how old was the GPT-4 model even as it can't be that old? 🤔

[keen breach]

so fucking ridiculous

i was legit excited about GPT5

[velvet pelican]

it wasn't old at all tbh

imma try to upgrade to Team now. I'll tell you if it works

[keen breach]

I'm mostly working on creative writing, worldbuilding etc. everything that i'm doing within it was supper smooth and fluid. now it all feels worse.

[wide thorn]

So they could of used the GPT-4 model as the alternativ model instead of the horrible GPT-4 mini one as soon as we hit the GPT-5 limit.

[velvet pelican]

YES

[wide thorn]

They honestly could of also made it unlimited too if it were an alternativ model.

[velvet pelican]

...I don't get the "Team" option on mobile...

[wide thorn]

Since if it is "retired" then might as well make it useable for both users without a limit (or at least a kind of limit that doesn't force you to wait every 15 or so messages).

[velvet pelican]

wait WHAT?!
They charge by "seats" on the "team" thing, and every seat is 30 bucks, and the minimum is 2 seats so it's a minimum of 60 bucks!!!!

I'm so done with ChatGPT I swear

[wide thorn]

Holy-

Yeah, plus it crashes a whole ton too from what I remember.

[velvet pelican]

anyone got any other ai chat models/websites to recommend?

besides gemini, it doesn't bold text, nor give titles or anything like that. the whole output it gives looks boring and not giving you anything besides plain text that just looks bad

[keen breach]

sadly nothing else has memory features like GPT

i worked with Grok for a bit but it's not really good at writing and feeling human

[wide thorn]

Yeah, sadly there aren't really a lot of AIs that are on the same level if not better than ChatGPT.

Though hopefully there will be one soon to rival ChatGPT, that is if OpenAI doesn't pull a Nintendo on them.

[keen breach]

i hope they fix this soon lol. feels like they rolled it out too quickly

[halcyon meadow]

Am I missing something? GPT-5 sucks even on Pro, it often makes no sense

[keen breach]

fusing all models into one though seems like a bad idea

[keen breach]

you're not. it sucks hard

[velvet pelican]

I heard that DeepSeek is a rival to ChatGPT

[wide thorn]

Yeah, it feels like as if they pushed out a Beta version of it.

[keen breach]

eh, its fine, but it doesn't have the human vibe

[wide thorn]

Ima check it out rq.

[velvet pelican]

Okay HOLY SH**! I started an RP story/game with DeepSeek now and it gave me such a good first scene response, better than anything 4o gave me!