general

[hollow imp]

Is it not good enough?

[verbal nimbus]

Sure, I use both Gemini and Claude a lot

[whole wagon]

guess where gpt5 will be. I think 66%

[patent bane]

78%

[hollow imp]

Nah guys not these rational benchmarks. I want your personal experiences

[minor bloom]

Its at 90% rn

[whole wagon]

bruh

that was fake

[thin creek]

90%up

[keen beacon]

70 percent

[barren prairie]

To not let people say they are ClosedAI they make the garbage open source

[primal orbit]

poe.com has gpt 5 thinking high free

[fleet lintel]

altman is shady guy... i was 100% sure that OSS models were going to be bad

[tired herald]

No way

[whole wagon]

if you think gpt5 is getting 90% on simplebench u are delusional ngl

[pulsar rain]

if you don't care about what you're typing then probably just using lmarena for unlimited use

*privacy

[rapid merlin]

either high 50 or low 60 im guessing

[tired herald]

No they dont

[blazing bison]

idk if lmarena is on thinking high

prob not

[rapid merlin]

idk if its on any thinking tbh

[gusty loom]

yo anyone found a documentation page for gpt 5?

[eternal niche]

40.7%

[hollow imp]

Bro I 😭 that thing is so expensive literally 1 msg with o3 and you're out of tokens.

[keen beacon]

Guys, it is henceforth Declared that GPT-5 shall score 70% on Simple-bench (Decided by you)

[pulsar rain]

it took more time to answer than gemini 2.5 pro, so probably?

[rapid merlin]

extreme traffic

[hollow imp]

Gpt 5 thinking vs grok 4 thinking

[torn mantle]

next month

[pulsar rain]

yeah we don't really sure about tat

[verbal nimbus]

My friend tried it. It really overthinks about safety in its CoT

[void elm]

do we have any hexagon or any prompts you guys can give me to test? i have the highest gpt 5 model and no idea what to test

[thin creek]

fake

[keen beacon]

I tried 20b Locally and 80% of its COT is about saftey

[hoary elbow]

GPT five is so close to winning against Gemini

[obsidian shell]

mixtral and again mixtral...

[devout vault]

gemini 3:

[primal orbit]

https://poe.com/GPT-5

it's free, i'm using it now

[hoary elbow]

Gemini three hasn’t released yet

[primal orbit]

reasoning effort high

[tired herald]

Limited

Points

You have 3k points, each message costs 250~

[primal orbit]

don't know about points, have never used that site

[void elm]

2027

ai hit a wall rn

[minor bloom]

2029

[tired herald]

Its good to test it out for 10 messages, but the daily 3k points is too little to do anything

[keen beacon]

2035+

[void elm]

no good improvements since gemini 2.5 pro tbh

[keen beacon]

yeah

[primal orbit]

ah ok

[verbal nimbus]

Hopefully it'll replace the standard GPT-4.1 model in Github Copilot, standard models are unlimited under Pro plan.

[hollow imp]

Let's see what does it do

[keen beacon]

I like how no one says this year lol

[keen beacon]

ah another irodov follower

[pulsar rain]

probably not in my lifetime, for true AGI

[keen beacon]

yeahh

[hollow imp]

It's important for my exam

[obsidian shell]

apparently the gpt 5 high in the interface is better then the gpt 5 on arena

[keen beacon]

hehe same

[hollow imp]

Jee advanced

[hollow imp]

11thie or 12thie?

[keen beacon]

12th

[keen beacon]

same

[void elm]

The tangential acceleration equals the time derivative of the speed:
dv/dt = w_t.

Here w_t is the projection of the constant vector a (directed along +x) onto the unit tangent τ to the trajectory:
dv/dt = a · τ.

Since a = a i (i is the x-unit vector), a · τ = a τ_x, where τ_x = dx/ds. Also
dx/dt = v τ_x.

Hence
dv/dx = (dv/dt)/(dx/dt) = (a τ_x)/(v τ_x) = a/v.

Integrate:
v dv = a dx → (1/2) v^2 = a x + C.

With v ≈ 0 at x = 0, C = 0, so
v(x) = sqrt(2 a x).

Thus the speed depends only on x (not on the path shape): v^2 = 2 a x for motion in the +x direction.

[minor bloom]

Isn't grok still the best at engineering?

(Not software)

[pulsar rain]

how does gpt-5 performs on numerical math? like pi^(pi+e+1)?

[rapid merlin]

where is this?

[gusty loom]

is gpt5 available on lm arena?

[keen beacon]

Amazing because of tool use

[solid brook]

yes

[pulsar rain]

most model without python or math module backup cannot do that

[tired herald]

Yes

[hollow imp]

Wait, there is a chance of it getting it correct.
Send it this.

[gusty loom]

Nvm i see it now

[pulsar rain]

so can we trust gpt-5 for college level and below math and hope it will never get basic algebra wrong?

[rapid merlin]

when is it even gonna be available in chat.openai.com

[pulsar rain]

I cannot trush gemini for that sht

[hoary elbow]

GPT five is like Gemini 2.5 pro but faster can’t wait for Gemini three though

[keen beacon]

0.5+ 0.05+0.018
does it reach 0.6

?

[hollow imp]

I've given up on that no ai does that properly. The only ai I haven't tried are the 200$-250$ ones like gemini deepthink and o3 pro mode

[keen beacon]

yeah no i was close hehe ( i said 0.6)

[rapid merlin]

holy crap, the output size is indeed solid

[keen beacon]

AIME scores indicate yes

[pulsar rain]

my take is just ask it to solve with python script and run that script

[verbal nimbus]

For common-sense reasoning, here's one from Simple Bench all models get wrong:

A luxury sports-car is traveling north at 30km/h over a roadbridge, 250m long, which runs over a river that is flowing at 5km/h eastward. The wind is blowing at 1km/h westward, slow enough not to bother the pedestrians snapping photos of the car from both sides of the roadbridge as the car passes. A glove was stored in the trunk of the car, but slips out of a hole and drops out when the car is half-way over the bridge. Assume the car continues in the same direction at the same speed, and the wind and river continue to move as stated. 1 hour later, the water-proof glove is (relative to the center of the bridge) approximately?

A) 4 km eastward
B) < 1 km northward
C) > 30 km away north-westerly
D) 30 km northward
E) > 30 km away north-easterly
F) 5 km+ eastward

[rapid merlin]

never seen an OAI output 1300 lines of code without resistance

[thin creek]

This doesn't involve the use of any tools

[keen beacon]

nah 2.5 pro works fine for me

[void elm]

Ill use high reasoning for this no tools

[barren prairie]

The high version is a pure trash at coding

[hollow imp]

Send this a simple question so even you can understand the answer.

Gemini 2.5 pro crashed

In thinking

[verbal nimbus]

Ok, but the solution is actually very simple (for a human) and doesn't require much reasoning.

[autumn blaze]

bro the lmarena gpt 5 says its gpt 4

[pulsar rain]

how about chess? does gpt-5 still suck at 900 elo 🤣

[verbal nimbus]

It gets it wrong as well

Thanks for trying

[keen beacon]

is it quadrant 2

Guys, "OpenAl PRs (no browsing)" Has hit a wall. ALL the models are stuck at 44 percent including o3, ChatGPT agent and GPT-5

[minor bloom]

What is gpt-oss? Its on arena

[thin creek]

No, GPT5 is cheaper than GPT4o

[keen beacon]

Will we see gpt-5 with reasoning on leaderboard ?

[obsidian shell]

playing game with gpt 5 high vs opus 4.1

so far stockfish thinks opus is winning

[obsidian shell]

ass

[keen beacon]

yeah samee

[stray aspen]

gpt 5 live on yupp ai

[keen beacon]

les goooooooo

[verbal nimbus]

Open source OpenAI model that just got released. There's 20B and 120B variant.

[hollow imp]

Wrong 😂

😂

[blazing bison]

funny gpt-5 live everywhere but not in chatgpt

🤓

[keen beacon]

its wrong??

ohh but i got 0.6

[clever estuary]

Api is live

[fleet lintel]

what is the answer?

[keen beacon]

@hollow imp

[rapid merlin]

the model seems to be blowing every other OAI model out of the model when it comes to UI though

or well, css

[keen beacon]

@hollow imp From 0 to 1: (2/√3)(arctan(√3) − arctan(1/√3)) = (2/√3)(π/3 − π/6) = π/(3√3).

[tired herald]

Who doesn't love seeing this

[hollow imp]

Let's give it actual jee advanced pyq

[keen beacon]

yeah

bro jee adv maths is ezz

physics + chem too hard

[solid brook]

man I'm excited af

imagine gemini 3

[south vigil]

has anyone benchmarked it yet?

[keen beacon]

in the process

[hollow imp]

@void elm If it can solve this then it passed

[keen beacon]

i have my q bank

[keen beacon]

i can solve this lol

i have some hard diagram geometry problems (creative thinking not complex formulae)

only grok has solved most

it didnt get 2-3

[hollow imp]

@keen beacon do you have telegram

[keen beacon]

nope

[south vigil]

is gpt 5 still not on the app?

[hollow imp]

This wrong btw

[south vigil]

i can't see it on selection

[hollow imp]

This def wrong

[obsidian shell]

is the 74.9% swe bench true? for gpt5

[rapid merlin]

Answer: A) 4 km eastward

Reasoning:

The glove drops at the midpoint of the bridge and (being waterproof) floats on the river.
Once on the water, it’s carried mainly by the river’s current: 5 km/h east.
The light 1 km/h west wind may slow that drift slightly. A simple approximation is 5 − 1 = 4 km/h east.
Any northward motion from the car (30 km/h) is lost almost immediately upon hitting the water; the north offset would be only a few meters—negligible after an hour.
So after 1 hour, it’s roughly 4 km east of the bridge’s center.
this is what the lmarena one said

[hollow imp]

@keen beacon why this question so heat that even ais and aspirations both get it wrong

[keen beacon]

How is GPT-5 performing at maths and physics thus far?

[hollow imp]

Bad

[stray aspen]

selector broke again?

[keen beacon]

Worse than other models???

[hollow imp]

No

[brave ferry]

Quick question: can anyone confirm whether GPT-5 started rolling out before the livestream ended?

[keen beacon]

yes it did

[keen beacon]

Can you explain? Does it perform better than others but still bad

[stray aspen]

it was out even before it started

[hollow imp]

@keen beacon what do I say

[keen beacon]

its better than others only a few quesitons its struggling on

[south vigil]

cannot confirm, haven't seen it land on my apps

[keen beacon]

use VPN hehe

[brave heron]

ye

[keen beacon]

works for me

[brave heron]

not working

[stray aspen]

@echo aurora

its broken

[remote niche]

guys when will gpt 5 be available in chatgpt ?

[half trail]

🙄

[keen beacon]

BRUH

[fleet lintel]

third and fourth quadrants ??

[hollow imp]

❌

[remote niche]

guys when will gpt 5 be available in chatgpt ?

[half trail]

Why gemini becomes dumb these days?

[remote niche]

guys when will gpt 5 be available in chatgpt ?

[eternal niche]

2035

[keen beacon]

its 2nd quadrant -x flips the function to the left side

[steady vale]

https://x.com/OpenAI/status/1953526577297600557

OpenAI (@OpenAI)

Introducing GPT-5

Our best AI system yet, rolling out to all ChatGPT users and developers starting today.

https://t.co/14Hf1Av6n6

X · Aug 7, 2025, 6:40 PM

[hollow imp]

It messed up this simple question

[half trail]

I used to get nice codes but from few days it making mistakes

[hollow imp]

No bru

[keen beacon]

3rd quadrant then? -x means left side right

[dusky aurora]

pineapple says that the LMArena team made adjustments and soon (tm) it will be seen, though I don't know when this "soon" will happen

[south vigil]

where is gpt 6

[fleet lintel]

launching on monday

[rapid merlin]

gpt 6 in 2077

[neon idol]

Me to I cant find it

[south vigil]

LOL

[keen beacon]

In the arrangement shown in figure a weight A possesses mass m = 4 kg, a pulley B possesses mass M = 2 kg. Also known are the moment of inertia I = 2 kgm2 of the pulley relative to its axis and the radii of the pulley R = 1 m and 2R. The mass of the threads is negligible. Find the acceleration of the weight A after the system is set free, taking acceleration due to gravity equal to 9.81 m/s2.

[echo aurora]

Looking into

[neon idol]

The answer is: idk 🫡🫡🫡🫡

[hollow imp]

@keen beacon is this correct?

[hollow imp]

😂

[tired herald]

Asking for its architecture on LMArena gives GPT 4 while asking on Poe gives GPT 5

Idk who to trsut

[keen beacon]

yeah its correct 3.5 (for g = 10)

[keen beacon]

lmarena has no system propmt (api based)

while poe may have a systme propmt

We are never agi

[hollow imp]

Wrong again

[neon idol]

Why I cant find gpt 5 on chatgot web and app 💀💀

[tired herald]

Cursor is doing well by offering gpt-5 for free for a bit

[tired herald]

Not available

[neon idol]

Damn

[primal orbit]

thinking?

[echo aurora]

Yeah looks like models are missing atm, team is aware and working on

[half trail]

It's a mystery

[neon idol]

🫤

[tired herald]

Idk, haven't tested

[rapid merlin]

i think he meant chatgpt.com, not entirely sure

[keen beacon]

o3 high gets this quesiton wrong most of the times:

A home aquarium partly filled with water slides down an inclined plane of inclination angle θ with respect to the horizontal. The surface of water in the aquarium
(a) remains horizontal
(b) remains parallel to the plane of the incline
(c) forms an angle α with the horizon where 0 < α < θ
(d) forms an angle α with horizon, where θ < α < 90

its b

[wheat onyx]

still don't have access to gpt5 yet

[neon idol]

Aniway I wil try it in lm arena

[tired herald]

Lmarena

[stray aspen]

bro lmarena is down

[hollow imp]

Every ai gets the quadrant question wrong

[keen beacon]

nah its working for me

[half trail]

How to get free access of claude?

[half trail]

😶

[hollow imp]

Lmarena

[keen beacon]

lmarena direct chat

[wheat onyx]

I have GPT Plus, where's my upgrade

[minor adder]

[golden ocean]

alpha or beta url still works

[half trail]

It's fully functional as pro version?

[stray aspen]

@echo auroracan you escalate please

[wicked root]

alright how's gpt5 looking so far?

[rigid holly]

It deleted my chats like twice

[golden ocean]

dont escalate pls

let us keep gpt 5 for ourselves!

[keen beacon]

better than o3 in STEM

[wicked root]

how about versus Gemini 2.5 pro?

[keen beacon]

yeah (but no tools)

[stray aspen]

wdym we cant even use it because teh website is down

[echo aurora]

Already have

[keen beacon]

on par

[pulsar rain]

the backend's down probably

[stray aspen]

thanks

[golden ocean]

for those who have the working url 😊

[keen beacon]

i am

[rapid merlin]

Answer: (b) remains parallel to the plane of the incline

Why: In the frame of the sliding aquarium, there’s a pseudo-acceleration up the incline of magnitude a. For a frictionless slide, a = g sinθ. This exactly cancels gravity’s component along the plane, leaving only g cosθ normal to the plane. The free surface is perpendicular to the effective gravity, so it must be parallel to the plane.

Note:

If the tank were sliding at constant speed (a = 0), the surface would stay horizontal (a).
In general, the tilt α (from the horizontal) satisfies tanα = a cosθ / (g − a sinθ), so 0 ≤ α ≤ θ, with α = θ only for the frictionless case.

[wicked root]

how do u know?

[bleak sundial]

need to compare token cost for marginal improvement

[keen beacon]

i am testing it on the quesitons in my q bank

[hollow imp]

@void elm what did it say about the quadrant question?

[half trail]

Personal opinion gemini is dumb

[tired herald]

Will we get the mini/nano models on lmarena?

[stray aspen]

idk but they are on yupp ai

[minor bloom]

So guys, what exactly are tools? Does lmarena have them?

[keen beacon]

nope. tools r extra features that model can acess - google search etc

[keen beacon]

@hollow imp

[golden ocean]

but lmarena does have google search !!!

[minor bloom]

Is thinking a tool?

[keen beacon]

nope.

[keen beacon]

thats separate right

[clever estuary]

Ofc not, lm arena is a testing site for benchmarks

[stray aspen]

lmarena is fixed

[half trail]

I have perplexity comet but hoping for android version because most of the time we search in phone not in laptop

[neon idol]

Uhm @echo aurora (Sorry for ping)

[stray aspen]

its fixed

[polar garnet]

Aame

[tired herald]

Dem

[polar garnet]

Same

[tired herald]

Cool site

[pulsar rain]

I think lmarena gemini 2.5 pro have google search

[echo aurora]

(no need to apologize for ping) Team is aware of this and working on a fix

[polar garnet]

Thanks

[echo aurora]

looks back for some, but still working on a fix

[tawny kelp]

Huh... I was wondering what model Summit was. I remember having some interesting interactions with that one.

[neon idol]

Ok thanks and sorry again

[clever estuary]

Did you guys fix it with AI lol

[polar garnet]

Also all my conversations are deleted

[stray aspen]

yes

[keen beacon]

theyll be backkk

relax

[neon idol]

Me too

[stray aspen]

seems like they reseted the whole website

[hollow imp]

Konsi konsi books se uthaye hai vro 💀

[half trail]

😶

[wicked root]

what's this?

[pulsar rain]

the frontend got disconnect from back end. everything will be back to normal

[clever estuary]

Wait fr???

[rapid merlin]

Short answer: You can’t tell from the given picture.

Why:

For x ≥ 0, f(|x|) = f(x). So the graph shown only tells you how f behaves for non‑negative inputs.
But f(-|x|) = f(t) with t ≤ 0. That uses only the values of f on the negative side, which the picture does not give.
What we can say:

y = f(-|x|) is always an even function (symmetric about the y‑axis).
Its graph is the y‑axis mirror of the left half (x ≤ 0) of y = f(x).
Without knowing f for x ≤ 0, the graph of f(-|x|) could lie in quadrants I–II, III–IV, or cross the x‑axis. Extra assumptions (e.g., f even ⇒ same as f(|x|); f odd ⇒ the negative of f(|x|)) would be needed to decide.

[golden ocean]

gpt 5 on mc bench

[keen beacon]

"GPT-5 usage limits 👀

10 messages every 5 hours on Free
80 messages every 3 hours on Plus
Unlimited use on Teams and Pro
"

[keen beacon]

exams se haha AIME IMO IOQM JEE ADV IRODOV

[stray aspen]

enjoy the 8k context for free users

[neon idol]

Wait what 💀

[half trail]

WAHT

Huh

[rapid merlin]

32 for non enterprise 💔 💔

[verbal nimbus]

It's very surprising how low Claude is in the coding category, considering that it's leading SWE-bench.

[neon idol]

Are you joking? 😭🙏

[keen beacon]

Bro, this better include o3 level web search

[polar garnet]

I'm glad we have lmarena

[hollow imp]

Do I spend my time on them? Are they relevant? Like I'll save them if you share

[hollow imp]

🔥

[half trail]

I feel claude is better than gemini

[hollow imp]

@void elm @keen beacon the answer

[haughty siren]

Is the GPT-5 in LMArena with thinking?

[neon idol]

Nah

[polar garnet]

Gpt5 can switch from thinking to just giving you the answer

[verbal nimbus]

In the future, someone should get an agentic AI to re-run all the code in LMArena's public dataset and verify the votes. I bet 25-40% of votes are inaccurate for the coding category.

[toxic egret]

Its fixed, atleast for me

[keen beacon]

nono i am prepping for IOQM too so thats why - Irodov is more than enough

[neon idol]

FIXED

[hollow imp]

Are you in any coaching?

[neon idol]

Yay

[novel flame]

There is a reason why they don't include Gemini 2.5 Pro in the chart. Gemini is waaay better on long context

[keen beacon]

right

[pulsar rain]

uhmm. I don't think chat history will get back... that's big rip...

[half trail]

The best one is Google jules but normal gemini app with 2.5 is bad

[keen beacon]

it will

[polar garnet]

Samey

[keen beacon]

aakash

[wicked root]

if gpt5 and gemini are on par, why's the market acting like Gemini had won?

[rapid merlin]

considering the context size differences, in a way, yes

[rapid merlin]

i heard gemini is planning a new model

[neon idol]

Hi! 👋 How can I help you today? Would you prefer to continue in Italian or another language? We are starting bad

[verbal nimbus]

Idk, is there any new Gemini news?

[rapid merlin]

that people think will blow gpt 5 out of the water

[hollow imp]

Can I dm you and talk to you there? Are you frequently on discord? I have 11th prep doubts that can only be solved by personal experience

[neon idol]

Nop

[keen beacon]

sure

[verbal nimbus]

gemini.google.com has a new guided learning mode which is pretty interesting

[minor bloom]

For some reason gemini is doing much better according to polymarlet

[keen beacon]

its fire

[minor bloom]

Which makes no sense to me

[hollow imp]

Kaha se who btw

[wheat onyx]

Liars
https://x.com/OpenAI/status/1953526590325068163

OpenAI (@OpenAI)

GPT‑5 is available to all Plus, Pro, Team, and Free users starting today with access for Enterprise and Edu coming in one week. It may take a few days to roll out to all Free users.

- Pro users get unlimited access to GPT-5 & access to GPT‑5 Pro, ideal for the most challenging,

X · Aug 7, 2025, 6:40 PM

[keen beacon]

not very frequently - i need to study too

[keen beacon]

Dm MEIN

[half trail]

Imagen 4 is best and veo 3

[verbal nimbus]

Yeah, basically the Socratic method of learning, haha.

[half trail]

For coding still lacks a lot

[hollow imp]

I mean like not disappearing and then coming back in 4 months and worse not even coming back

[wicked root]

Yessir

[keen beacon]

not that much heehe (ill check twice a day)

[wicked root]

I dno

[hollow imp]

That's more than good enough I don't even check weekly

[keen beacon]

k

[misty vault]

💀

[stray aspen]

lmao

[primal orbit]

gpt-5 is an amazing model compared to gpt-oss

[stray aspen]

craig also said gpt oss was great

[solid brook]

huh

bro you high

[stray aspen]

lmao

[rigid holly]

Can my conversations not dissapear for 5 minutes?

[neon idol]

Elon Musk seeing that gpt 5 is better than Grok 4: 🥀🥀

[half trail]

Gpt 5 available

[stray aspen]

your late bro

[plain carbon]

I need advice on how to report the result of a prompt. I asked two AI's a question about doing something on Google Sheets. One was a complete disaster. The second gave me three options; one worked, one failed, and one was half-right. Do I say "both are bad", or "that one is better"?

[barren prairie]

I think he only tested llama

[half trail]

😶

[pulsar rain]

open-ai 🤝 CPU / GPU companies: at naming production

[barren prairie]

Both are bad

[half trail]

How about coding in gpt 5

[echo aurora]

Are you trying to delete a chat? There is a known bug where deleting the chat then refreshing the page brings it back temporarily.

[barren prairie]

Trash trash ,..

[rapid merlin]

it is MUCH better than their previous models at styling from what i can see

[plain carbon]

Thank you.

[pulsar rain]

naming it o3. 4o is actually so bad, normal people wouldn't even know which one is better. probably they want you to get confuse

[rapid merlin]

didn't test it that much to be able to tell anything else

[rigid holly]

No, i am just sending messages

[half trail]

🤔

[hollow imp]

@echo aurora hello sorry to ping but please can you tell me if lmarena is going to bring pdf attachment anytime soon? I really want it

[pulsar rain]

it's about naming

[rapid merlin]

it was good for its time 👍

the o models have a weird naming scheme

since the number jumped from 1 to 3

[keen beacon]

r u the reall craig? (sorry)

[hollow ocean]

@deep adder $1 method goated

[rapid merlin]

o2 became sentient and turned into a company

[echo aurora]

We're currently having some issues that is likely what you're running into

[jade egret]

is the gpt-5 on lm arena same as the gpt-5 on the officialt website?

[pulsar rain]

most people probably thinking: "well, it have 4 and o, so it must be better than o3, right?"

[hollow ocean]

I don’t think so lol

[neon idol]

Can someone pls give me some prompts for see how id god gpt5. Thx

[wicked root]

Gpt5?

[keen beacon]

In the arrangement shown in figure a weight A possesses mass m = 4 kg, a pulley B possesses mass M = 2 kg. Also known are the moment of inertia I = 2 kgm2 of the pulley relative to its axis and the radii of the pulley R = 1 m and 2R. The mass of the threads is negligible. Find the acceleration of the weight A after the system is set free, taking acceleration due to gravity equal to 9.81 m/s2

[echo aurora]

(no need to apologize for pinging me)

is going to bring pdf attachment anytime soon?
we are aware this is a highly requested feature from our community, we'll be sure to make an announcement when something like this is added

[neon idol]

Thx but what is the answer

[keen beacon]

@deep adder r u the real craig?

[keen beacon]

10.3

[minor bloom]

Btw, what is style control?

[neon idol]

Thxxx

[keen beacon]

\frac{3g\left(M+3m\right)}{M+9m+\frac{I}{R^2}}\ =\ 10.3\ m/s^2.

[jade egret]

gpt-5 is out? im on plus plan

[keen beacon]

another q:

A uniform cylinder of radius R = 1 meter is spinned about its axis to the angular velocity \mathbf{\omega}_\mathbf{0}=\mathbf{250}\ \mathbit{rpm} and then placed into a corner. The coefficient of friction between the corner walls and the cylinder is equal to k = 0.59. How many turns will the cylinder accomplish before it stops?

[hollow imp]

Bro have you tried deepseek v2 prover? That pure math model?

[neon idol]

Thx but the answer is?

[whole wagon]

We should be thankful for the competition in this space. Imagine having to wait 2 years from now for GPT6 💀

[whole wagon]

For another 3% improvement to benchmarks lel

[minor bloom]

Nah, its coming out next year

[keen beacon]

4

[pulsar rain]

probably put it on pay wall with limited uses for sometime

[keen beacon]

yeah but multimodalityyy

[neon idol]

Thx

[stray aspen]

unofficial websites rolled it out faster than openai

[hollow imp]

I would like to propose an interesting question.
What would be the state of ai in mid 2027? What would be the best model.etc.etc

[keen beacon]

close to AGI but not AGI

my prediction heh

[hollow imp]

Well I converted the non diagram questions into natural language through grok 4 and then fed it

[keen beacon]

hmm

[hollow imp]

Like in more detail bro

[keen beacon]

no one knows lol

[echo aurora]

anyone experiencing the models lagging a lot right now?

[blazing bison]

gpt-5 nano is insane

[wheat onyx]

they said today for plus, but I dont have it yet

[burnt axle]

hi

[rapid merlin]

LMAOOOOOO

[hollow imp]

You should try it. I had a hard time trying to find where to chat with it. It was on huggingface but with limits

[keen beacon]

i tried it, but its meh.... its not reasoning its a weird mess kinda thing

[blazing bison]

after this gpt-5 nano = zenith confirmed

[hollow imp]

Oh

[neon idol]

Uhm is normal that I am still waiting a response from gpt 5 ? 💀

[stray aspen]

artificial analysis

[rapid merlin]

Yes

[neon idol]

It's inpossible

[wheat onyx]

goddamn

[rapid merlin]

this feels more like gpt-4.75 to be fair

[stray aspen]

they are trippin

[minor bloom]

Thats odd

[pulsar rain]

gpt-5: 10^(π^e) ≈ 2.878446 × 10^22
gemini 2.5 pro: 2.878335... x 10²²
correct answer: 2.878443560...x10^22
gpt-5 is pretty close. it probably rounding issue right?

[keen beacon]

in my run it got it wrong

[stray aspen]

seems like its broken

[keen beacon]

yeah

[jade egret]

wait till gemini 2.6 🙃

[shell pewter]

quite useful, for anyone to catch up 😇

[whole wagon]

All the independent benchmarks are coming in and they are trash 💀

[neon idol]

Gemini 2.5 said the correct answer

[blazing bison]

and gpt-5 auto is routing between those models

[whole wagon]

.

[rapid merlin]

Gg

[whole wagon]

Artificial analysis tested and they have it as trash kek

[neon idol]

Also gpt 5

[rapid merlin]

I have gpt 5 in chatgpt

[whole wagon]

Why would they lie

[blazing bison]

bro gpt-5 sucks sorry

[neon idol]

Damn you have premium?

[rapid merlin]

yeah, also wtf

i have it on my phone but not on my pc ????

[neon idol]

🤣

[blazing bison]

i don't have it yet

[rapid merlin]

same account

[hollow imp]

New people coming to discuss about gpt 5 and new people getting in beef with you 😔

[keen beacon]

yeah

[shell pewter]

yeah, apparently, just ask it to “think hard about this” in the prompt, it will switch to thinking mode then, that's what's been said in the release docu 🤓

[hollow imp]

@keen beacon dm

[wicked root]

Will gpt5 beat gemeni in lmarena leaderboard for text?

[blazing bison]

gpt-5 400k context and they still limited chatgpt pro users with 100k, i'm gonna cancel this s****

[elder rapids]

@deep adder gpt 5 is good

[hollow imp]

I add this specific line in the prompt " Perform a deep granular assimilation of the question before proceeding".

[blazing bison]

what you have on other models?

[steady vale]

anyone know when gpt5 will be available

[ocean vortex]

I mean... it kinda is? SOTA for function calling and instruction following. And challenging for the top spot everywhere else

[keen beacon]

yeah

[elder rapids]

good vibes, good overall model, js hope it's smart

[rapid merlin]

gpt-5-thinking

[stray aspen]

they are rolling out the update to trillions of users

[blazing bison]

gpt 4.5 is gone?

NOOOOOOOOOOOOOOOO

[rapid merlin]

literally everything is gone

besides those two

[blazing bison]

NOOOOOOOOOOOOOOOOOOOO

[whole wagon]

All previous models are getting deprecated

[rapid merlin]

💀

[neon idol]

Only at me gpt 5 is extremely slow?

[stray aspen]

its thinking

[whole wagon]

Only GPT5 will exist. I assume it is cheaper or smth

[pulsar rain]

high usage rn

[keen beacon]

theyll have some time to deprecate

[neon idol]

Bruh ok but he is thinking form 2 minutes 💀

[little narwhal]

ALL of them?

[whole wagon]

Yes

[whole wagon]

ALL

[ocean vortex]

yeahh they nuked it. Old cache but still wouldn't let using it lol

[blazing bison]

so sad

never see you again

[whole wagon]

I think they try to reduce losses. Maybe GPT5 is actually smaller than o3 even

[keen beacon]

once i had o3-mini think 10.5 minutes for a physics problem

🤣🤣

[solid brook]

this can;t be real

[stray aspen]

10.5 o3 minutes is 7 hours for deepseek lmao

[neon idol]

🤣

[keen beacon]

frrr

[empty stump]

Yesss it's out

[neon idol]

REAL

[blazing bison]

never doubt gpt-5 nano

[keen beacon]

it runs out of thinking tokens 🤣🤣

[whole wagon]

Vote plz

[wicked root]

Will gpt5 overtake gemini on the leaderboard?

[blazing bison]

gemini 3 will be better for sure

[neon idol]

Chat, be honest. You are in this group only for see if there are better ai to use for homework 💀🤣

[keen beacon]

lots of traffic, that's the reason

[pulsar rain]

already

[wicked root]

And work man

[little narwhal]

Not everyone is a com kid like you

[keen beacon]

nuh uh

[wicked root]

Nah not with style guide switched off

[minor bloom]

GPT-5 has 4 new chat personalities: Cynic, Robot, Listener, Nerd.

Find them in Customize ChatGPT in settings.

[whole wagon]

What does openAI even do now. The AGI illusion has been shattered kek

[keen talon]

why gpt 5 says he is gpt 4

[neon idol]

@keen beacon chatgpt 5 said wrong answer

[minor bloom]

They will never beat the Grok girl

[ocean vortex]

@willow grail Told you it's to be same day release for EU. You wouldn't listen

[rapid merlin]

what should I try to give gpt 5 thinking

[pulsar rain]

about to get out of college already. I'm worry if my future IT company allow using AI or not

[keen beacon]

exactlyyy

[whole wagon]

"It's like a team of PhD experts on demand"

😂

[rapid merlin]

its so slow though

the thinking model at least lol

quite the tank

[clever estuary]

"gpt5 is good at frontend aesthetics"
gpt5:

[jade egret]

ye but how long do you think it would take for other company to beat it?

[pulsar rain]

pretty decent for one-shot

[quick holly]

hello

[jade egret]

the other one is made bu gemini?

[clever estuary]

yeah 2.5 pro

[jade egret]

lowkey kidna good....

what the prompt?

[wicked root]

Chat, will gpt5 beat gemini on lmarena?

[keen beacon]

and for a wrong answer (o4-mini)

[whole wagon]

Only with style control

[exotic tartan]

I'll take left side over strawberry pizza

[clever estuary]

make a beautiful webpage for a classic pizzeria store (embed css)

[whole wagon]

The results are already out

[wheat onyx]

im refreshing my GPT page every 30s. give it to me

[wicked root]

What does style control do?

[lime coral]

The Gemini bat logo will rule them all

[rapid merlin]

ill see what i get from it lol

[keen beacon]

another

[blazing bison]

now @deep adder can start hyping gpt-6, gpt-6 for sure will be a different beast, insane model, sota

[whole wagon]

At this rate xAI gonna overtake openAI kek

[wheat onyx]

also Google is bankrupt

[rapid merlin]

o4 will be good guys

trust

[pulsar rain]

Annotators on their way to destroy their own career:

[stray aspen]

@deep adder will gpt-6 be AGI

[wheat onyx]

believe they said all future reasoning models will go into the autorouter (GPT5)

[jade egret]

; (

[wheat onyx]

and we should be getting a big reasoner in ~December maybe

[lime coral]

GPT5 is O4

[stray aspen]

grok 5

[blazing bison]

gpt5 is o3

[wheat onyx]

yeah I think so too

[blazing bison]

o3 is o2

[steady vale]

does anyone have gpt5 access on chatgpt.com yet

[rapid merlin]

lol

[pearl kiln]

yo guys why do some people have sound on their videos?

[clever estuary]

gpt5?

[rapid merlin]

i have it on my phone app for SOME reason

[stray aspen]

because they got lucky

[jade egret]

gemini ; (

[blazing bison]

@pearl kiln bcs of veo 3

[rapid merlin]

not on the website though

[pearl kiln]

yeah ik but is it random

[raven oracle]

Gpt 5 pretty laughably bad at translation compared to something like 2.5 pro, it writes like a textbook XD

[blazing bison]

it is random

[clever estuary]

hmm ig gemini gets lucky at times

[stray aspen]

yes

[steady vale]

oh woww i have it too but only on my phone

lmfao

[pearl kiln]

alright thx

[jade egret]

yea lol

ima try again tho

[keen beacon]

OMG I WANT GROK 5

[blazing bison]

gpt-5 suck for writing btw

[keen beacon]

GPT 5 seems to suck at lotta things

reminds me of meta's scout and maverick releases

[blazing bison]

now i need to test it with code

[keen beacon]

hehe

[blazing bison]

they said that it's a bunch of phds in my pocket

[keen beacon]

lol

[quartz light]

yall try generating music with web audio api with gpt5

[whole wagon]

It's strange how it basically is just openAI failing to progress at a fast rate. The Chinese LLMs don't seem to be having a slowdown

[keen beacon]

PHD from whatsapp?

[keen beacon]

exactly

[stray aspen]

well those phds must have bought their degree because they are dumb as hell

[quartz light]

its really good lol

[keen beacon]

fr

[blazing bison]

claude is not slowing down

just look the paper they released about the llms personalities, that's insane

[olive mesa]

gpt-5 isn't agi...

[keen beacon]

but claude sonnet 4 is not that great of an improvement from 3.7

[olive mesa]

sigh

we're never getting agi

[keen beacon]

opus 4 is awesome tho

[keen beacon]

yeah

[blazing bison]

it is lol

[keen beacon]

maybe but not imo

[whole wagon]

According to sam it's even better. An entire team of PhD experts in your pocket

[blazing bison]

for me it is

[whole wagon]

[keen beacon]

hmm

[willow grail]

lol where not true bro.
i dont have it on my openai account

[keen beacon]

agi is just human levle. bro said ASI in your pocket

[blazing bison]

for me claude sonnet, not opus, is the best model actually

[clever estuary]

GPT5 vs Gemini 2.5 pro

[ocean vortex]

openai playground

[willow grail]

u and ur expensive api

[keen beacon]

hmmm

[ocean vortex]

You need to be paying API customer though. But still EU 😇

[stray aspen]

gpt 5 wins

[blazing bison]

not for math though

[clever estuary]

yeah gemini gets lucky at times

[blazing bison]

for math it was o3

[keen beacon]

yeah tho i use o4 mini its really fast n a math nerd

[white hatch]

both are nice

[willow grail]

gemini wins

[keen beacon]

but GPT OSS 120b on cerebras impressed me

hard questions, answers in < 5s

its amazing

[stray aspen]

which is gemini

[keen beacon]

right one

[stray aspen]

gpt5 looks better

[keen beacon]

nah

gemini

[willow grail]

design wise gpt one is ugly makes no sense.... pls look again at it..

the 2.5 pro version is clearly inspired from retro stuff.....
the gpt one... is... not.. its not inspired at all

[stray aspen]

they are both great

[blazing bison]

they both suck

in my opinion

let's keep humans doing design for now

[keen beacon]

yeah hehe

[keen beacon]

yess

[clever estuary]

but folks hate css tho

[white hatch]

you're right

[keen beacon]

I am more interested in day to day stuff since I am not a coder. People will use GPT 5 for more trivial stuff.

@blazing bison

but GPT OSS 120b on cerebras impressed me

hard questions, answers in < 5s
its amazing

even on math arena it tops (even on SMT whose q havent been released)

And decreased hallucinations help with correct info

[civic flame]

@echo aurora hi, can you please provide an update on the feature request to allow disabling the sysprompt in direct chat? i requested this before the new lmarena even released and i've heard nothing since

all in on google

[blazing bison]

i really like claude sonnet for day to day stuff, i like how claude use less dashes, tables, etc

[civic flame]

[keen beacon]

i need the updated 2.5 pro asap

[willow grail]

tho the gpt one has more various elements

[pliant cliff]

wait gpt5 is first image?

[white hatch]

Actually I trust Google more than OpenAI

[keen beacon]

formatting is bad

[clever estuary]

yeah

[rapid merlin]

not TOO bad

[keen beacon]

100 req / day for their best model

[blazing bison]

but if you aren't actually talking with it, just using it like a search engine, then gpt is the best right now i think

[willow grail]

how can usee formatting

[rapid merlin]

not all that bueno but still

[clever estuary]

gemini?

[keen beacon]

formatting means the look n placement of the elements

[rapid merlin]

thats gpt 5

[keen beacon]

I like to use it for brainstorming stuff

[clever estuary]

ah I see

[keen beacon]

ideas etc.

[blazing bison]

zenith was good with design btw, WHICH MODEL WAS ZENITH

SOMEONE

SOMEONE ANSWER

[eternal niche]

zenith

[stray aspen]

it was grok 6

[keen beacon]

Ahhhhh, the website is going down

too many people

[umbral silo]

Hi

[keen beacon]

Hello

[umbral silo]

How are you?

[stray aspen]

gpt 5 design is disappointing. gemini does better

[keen beacon]

Doing alright, trynna test GPT-5 but it is so slow because everybody is smashing code into it

[ripe mountain]

which ai do you think is the best right now? gpt 5?

[gusty loom]

nah

[keen beacon]

Needs more testing

[rapid merlin]

What did you try it in?

[bright kayak]

when is gpt-5 actually coming out? I still don't have it anywhere

[gusty loom]

its out

[fossil fable]

yup

still not on chatgpt

[keen beacon]

It is on LMArena at least

[willow grail]

ROFBLOX

[bright kayak]

eu?

[bright kayak]

yeah ik

[fossil fable]

no

[steady vale]

check your phone bro

[bright kayak]

ok then thank god

[gusty loom]

yeah its on lm arena i meant

[bright kayak]

I am & I updated it

[fossil fable]

...

...uk

[keen beacon]

Will take time

[steady vale]

oh im in usa

[keen beacon]

I dont have it either

[clever estuary]

I told them to recreate Deepl
GPT5

[steady vale]

maybe usa ppl got it first

[bright kayak]

its probably progressively rolling out

[keen beacon]

but it could be because I am in the EU

[clever estuary]

vs gemini 2.5 pro

[keen beacon]

DeepL is based btw

[gusty loom]

Actually gemini is looking better ngl

[keen beacon]

better than google translate

[fossil fable]

i disagree but opinions opinions

[pliant cliff]

gemini better

[raven oracle]

Looks more like the actual website

[bright kayak]

isnt that literally the deepl site?

[keen beacon]

Yeah looks neater

[clever estuary]

this is what deepl actually looks like

[keen beacon]

no

[pliant cliff]

it's oblox ahh design

[bright kayak]

then its very good

[molten parcel]

YOoooo

hello guys

[fossil fable]

design score goes to GPT-5

accuracy score goes to 2.5 Pro (HOW)

[molten parcel]

i am new in this server

[stray aspen]

really?

[gusty loom]

its pretty good like could of removed some of that shadows but its pretty good

[white hatch]

booyah

[stray aspen]

if thats not the actual deepl website gemini did a great job

[gusty loom]

yeah

[molten parcel]

helloo

[clever estuary]

Gemini is very good at a following image instructions

[molten parcel]

btw

[keen beacon]

oblox seems to be some peak UI benchmarking lol

[molten parcel]

can i use veo 3 inthis server?

[stray aspen]

yes

[white hatch]

ofc

[stray aspen]

if you get lucky enough

[keen beacon]

#1397655624103493813

[molten parcel]

SWEEET!!!!

[gusty loom]

That's right. Unlike chatgpt who sometimes does unwanted things that aren't in the prompt. But gpt 5 handles it very well.

[fossil fable]

i find that GPT-5 is far from just focusing on intelligence but also how pleasant it is to work with in general use (like, i don't know, ChatGPT)

[molten parcel]

what do u mean by that

[white hatch]

How does gpt-5 behave on coding tasks?

[queen heart]

@everyoneIs there a way to generate videos with sound here, pls help I just got here today

[stray aspen]

the video model you will get is random

you cant choose it

[fossil fable]

brilliantly from my experience

[bright kayak]

@everyone!

[fossil fable]

like

Brilliantly

[stray aspen]

go to video arena channel bro

[gusty loom]

Yes! Just go to video arena and write a prompt that includes audio description and it will auto-use veo3.

lol

[molten parcel]

hHOW MANY Veo3 videos can i GENERATE IN A DAY

[eternal niche]

kid

[whole wagon]

Guys I think there is smth actually extremely wrong with gpt5. Like they made a mistake somewhere

[keen beacon]

8

[eternal niche]

everything

[gusty loom]

depends on how many discord accounts you have😂

[whole wagon]

Look how it performs compared to horizons

[fossil fable]

Devours Sonnet 4

[whole wagon]

There is a bug in the inference or smth

There is no way it is this bad

[molten parcel]

FREE?

[molten parcel]

IN THIS SERVER?

[gusty loom]

yes

[keen beacon]

Yes. Why you shouting to me?

[fossil fable]

default thinking or

[molten parcel]

NAH BRO I AM JUST TRYING TO HIGLIGHT THE MSG JUST IN CASE U SEE IT

[keen beacon]

OKAY

[fossil fable]

i find it a shame people now dumb it down to "thinking" instead of "reasoning"

[eternal niche]

HELLO

[steady vale]

just had gpt-5 answer a question way worse than 4o lol

[molten parcel]

my bad i dint mean it negetively

[blazing bison]

gemini is so much better lmao

[stray aspen]

horizon beta had a great design

[molten parcel]

Noice joke buddy

[whole wagon]

I'm pretty sure they screwed smth up in the release

[blazing bison]

wait horizon models already revealed?

[molten parcel]

hello

[whole wagon]

Like artificial analysis benchmark and everything is literally showing GPT5 worse than o3 it's smth wrong in the launch

[blazing bison]

ok we have some real improvements here

[molten parcel]

btw

[gusty loom]

In what time term?

[molten parcel]

where CAN i actually generate it?

[gusty loom]

Video Arena 1/2/3/4

[eternal niche]

HERE

[gusty loom]

Check out #1397655624103493813

[molten parcel]

thx bud appreciate u

[molten parcel]

ohh thc

thanks

[eternal niche]

**HOW ARE YOU **

[placid spear]

Which version of GPT-5 do we have in the arena? GPT-5 pro or just the "normal" thinking model that Plus users have?

[gusty loom]

Can anyone give me a memory prompt for Chatgpt to not use the long dash and use emojis less?

[eternal niche]

standard version

[rapid merlin]

im pretty sure there is a pro?

[clever estuary]

pro is not released I believe

[neon idol]

After gpt 5 I am atill waiting for deepseek R2

[placid spear]

Thank you

[clever estuary]

deepseek is just a one hit wonder

[blazing bison]

i don't think so...

[gusty loom]

After all, we should all really appreciate LM Arena thats let us use their service for free.

[whole wagon]

Ohhhh they are showing benchmarks for GPT5 high internally but we only have GPT5 medium ofc. That's why everyone else's benchmarks are awful 😂

[blazing bison]

it just that deepseek don't hype, they release

[whole wagon]

They didn't release the high version yet

[clever estuary]

it's proven they can't catch up, they don't even have vision at this point

[tribal aspen]

@echo aurora will we have GPT 5 pro on LMArena direct chat

[neon idol]

Fr

[blazing bison]

no we have gpt 5 high

[gusty loom]

prolly soon or so

[blazing bison]

it's avaliable already

[tribal aspen]

Why they released two versions of GPT 5 😭

[keen beacon]

Hold on... Is that with thinking enabled?

Will LMArena add the rest?

[whole wagon]

It's not high. There is no reasoning level in the API I have looked

[tribal aspen]

Ohh but there's no such model on the API called GPT 5 PRO

[white hatch]

daaaaawmn high

[whole wagon]

There's no setting

[tribal aspen]

But the website does have

[whole wagon]

Everyone currently is using GPT5 standard

[gusty loom]

Maybe they openai will add it? How you know?

[whole wagon]

Including LM arena

[blazing bison]

idk if you're talking about the playground, but it's working with python

[tribal aspen]

Only three models are on api lol

Gpt 5, mini, nano

But no pro

I think that's a variant of gpt 5 not pro

What is it

[gusty loom]

but better

[whole wagon]

The horizon models don't match up to GPT5 then

All the benchmarks are looking like this and wondering wtf is going on

[gusty loom]

yeah prolly

[tribal aspen]

Yea it's crazy

[whole wagon]

It's not just one person. Multiple benchmarks

[tribal aspen]

Wtf did i just see

[whole wagon]

Have it like that

[blazing bison]

" Pro, Plus, and Team users can also start coding with GPT‑5 in the Codex CLI⁠(opens in a new window) by signing in with ChatGPT."

[clever estuary]

honestly of all AI models released these days
Apple's OpenELM is still the gold
no one can top that

[blazing bison]

wow

[whole wagon]

There's smth wrong with gpt5 lol

[tribal aspen]

Where's gpt 5 pro benches

[blazing bison]

so now you can use codex cli with chatgpt account

that's interesting

[tribal aspen]

Where's Codex CLI now

[solid brook]

bro wth lmarena had the model on the second gpt 5 was announced but the chatgpt website is still the old one

[blazing bison]

https://github.com/openai/codex

[wheat onyx]

Lies

[blazing bison]

who is lying?

[whole wagon]

Did codex get updated

To GPT5?

[wheat onyx]

I don't have it

[stray aspen]

still smarter than gpt oss

[wheat onyx]

Oh it's codex

[whole wagon]

Oh nice I have it

[gusty loom]

It was updated an hour ago

[wheat onyx]

Im just dumb

[blazing bison]

they did this bcs of claude code lmao

[gusty loom]

I certainly think thats gemini is better at coding than Chatgpt.

[whole wagon]

Codex is the best coding agent ngl

[blazing bison]

no

[whole wagon]

Yes

[blazing bison]

not the cli one

the web maybe

[whole wagon]

I use the web

[blazing bison]

you can't delete your data from there

that's insane

[jade egret]

when gemini 3

[whole wagon]

I work on open source stuff so idc lol

[blazing bison]

ye, people that do valuable work care

so it's not the best agent

[keen fulcrum]

Wait zenith was gpt5

What was better than zenith

[blazing bison]

no one knows what zenith was

[stray aspen]

bro is gpt 5 improving?

[whole wagon]

I do Linux development

[stray aspen]

i feel like its better than an hour ago

[whole wagon]

Bro has 0 clue about anything kek

[blazing bison]

with AI, uugh

[gusty loom]

it doesn't just "improve" like a human

[blazing bison]

AI suck with c, with rust

so idk what are you doing

[whole wagon]

I use it with rust just fine

[blazing bison]

yeah bro, my colleagues that actually do something with rust disagree

claude opus is the best one for rust but it generate super verbose code, so no one uses it

[hollow imp]

WHATS BETTER IN LMARENA GPT 5 THAN OFFICIAL CHATGPT FREE TIER GPT 5?

[gusty loom]

stop yelling please

its the same

[blazing bison]

it's not

[gusty loom]

how

[keen beacon]

?

[blazing bison]

free after some prompts change to gpt-5 mini

on arena it always gpt 5

[hollow imp]

Any performance related change?

[gusty loom]

OpenAI auto switches to mini after a few times, LM Arena doesn't.

[blazing bison]

yeah

[keen beacon]

Ofc, not because it is not there yet.

And it is not possible in general

[blazing bison]

it is for some people

[keen beacon]

oh

[hollow imp]

Do you guys know how to get a better version somewhere else? Like a more reasoning one

[blazing bison]

slow roll out

[blazing bison]

only if you pay for it

[hollow imp]

I'm talking about free

Like how you can use o3 and o3 search in lmarena and nowhere else for free

Same with grok 4 search

Same with opus 4 thinking 16k

@blazing bison something like that for a better gpt 5?

[gusty loom]

for search just click the globe icon at the prompt bar

[devout vault]

If summit is gpt 5 then what is zenith?

[blazing bison]

free biscuits only on lmarena brothers

[devout vault]

Chat what is zenith

[hollow imp]

🥺

I want free biscuits

[gusty loom]

Well sometimes thinking gets much slower results. (sometimes even worse results)

[hollow imp]

Especially my eye is on gemini 2.5 deepthink

Even if I get 1 prompt per week

Just free biscuit pls

🙏

@blazing bison

[tribal aspen]

Nothing in SWE?

[ocean vortex]

yeah they gonna direct replace 4o-latest with gpt5-chat-latest

[devout vault]

Does anybody know anything about zenith

[ocean vortex]

And o3 with gpt5 medium reasoning

[gusty loom]

Right. I didn't relate to the analysis you sent.

[ocean vortex]

though that's likely later

[hollow imp]

o3 lm arena level reasoning vs gpt 5 free tier reasoning

Is gpt5 reasoning not available at all for free on anywhere?

[ocean vortex]

free tier (gpt5-mini-reasoning) is probably like marginally improved o4-mini

though I haven't seen direct comparisons yet

[hollow imp]

So you can't use gpt 5 reasoning?

What about poe.ai gpt 5 reasoning high?

[gusty loom]

Well it's probably a minor improvement. Reasoning models are mostly slower so it's probably only improved in the output, and not in the reasoning time.

[ocean vortex]

wdym. I think free tier is gonna be mini

[ocean vortex]

well limited use of the full model probably, but not with reasoning

[hollow imp]

???

You sure?

[keen beacon]

Nope, after certain amount of prompting, it will automatically switch to the mini model

[ocean vortex]

and then it will fall back to gpt5-mini with and without reasoning

[keen beacon]

yes

[hollow imp]

Tell me about this

[whole wagon]

Who is going to eat openAI first. Chinese LLMs or other American ai companies

[hollow imp]

If I stop and open the chat the next day, will this be fixed?

[ocean vortex]

yeah that's what I meant. But unlikely that they will do full gpt5 + reasoning of it. For free users. They didn't get o3 either only o4-mini

[keen beacon]

Gemini 3.0

[hollow imp]

Deepseek is cooking

[keen beacon]

Ah, okay

[hollow imp]

O4 full release when

[molten parcel]

wait...yall are gettng access to gpt 5?

howwwwwww?

[keen beacon]

Not me yet

lol

[ocean vortex]

it's gpt5

[hollow imp]

No o4?

[ocean vortex]

and "o4-mini" is just marketing name lol

[keen beacon]

No, because GPT 5 is about combining reasoning into one model

no more O series

[hollow imp]

Why openai naming so twisted

[ocean vortex]

No, they merged them now with gpt5

[hollow imp]

Why don't they do it like gemini reasoning forced on every response

[molten parcel]

But where can u access it?

[ocean vortex]

that's how o3 was

[molten parcel]

can u access in this server?

[keen beacon]

LMArena at least but not on the gpt app

[keen beacon]

at least not for me

[ocean vortex]

you don't need forced reasoning though

[barren prairie]

Glm5

[hollow imp]

Is glm reasoning model?

[keen beacon]

Because it is hybrid. It uses reasoning when needed

[barren prairie]

Yes but ai agent more than a model

[molten parcel]

sooo the gpt5 is in this server i mean LMArena?

[eternal niche]

lmarena.ai

[hollow imp]

Is it on par with o3 and gemini 2.5 pro and grok 4 math skills?

[keen beacon]

https://lmarena.ai

[hollow imp]

Russian 🙀

[barren prairie]

I didn t test it on math

[wicked root]

Im confused. Is gpt5 overtaking gemini on the leaderboard and does this mean openAI won?

I hear lots of mixed views on this

[hollow imp]

Russian

[gusty loom]

Please don't send it here.

[eternal niche]

no

[half trail]

Gpt 5 not upto the mark in coding

[hollow imp]

His pfp reminds me of him

[barren prairie]

Your role here always sending this pic 🤣🤣🤣🤣