general

[silent tree]

hes lost lmao

fr and then they'll accuse me of worshipping sam altman like i dont even do that

[surreal zephyr]

i worship tibo

[silent tree]

tuff

LMAO

SO IT WAS A DEEPSLEEP USER

[surreal zephyr]

5.5 thinking lambo via plain code

[silent tree]

ALL ALONG 💀

[next trout]

deepsleep LOL

[surreal zephyr]

ok this is funny but

deepseek passes the riddle in 30 mins

claude fails after 18 mins

gpt passes after 3 mins

[silent tree]

well yeah

but he cant be defending claude

like he uses it

[next trout]

which riddle is this?

[surreal zephyr]

math task

[next trout]

https://tenor.com/view/dear-autocorrect-bigger-gif-21687522

[next trout]

what's the prompt

[surreal zephyr]

holy based gif i got it starred too

[surreal zephyr]

too long to paste here

<@&1349916362595635286> @echo aurora

execute bro

antisemitism

!!!

[silent tree]

pineapple missile boom boom

[next trout]

XD

[silent tree]

I already took ss

[surreal zephyr]

another claude user dead

[silent tree]

he csnt delete

[surreal zephyr]

o7

[silent tree]

• theres logs soo it doesnt matter

[next trout]

no it's a deepsleep user

[surreal zephyr]

dead

W

[silent tree]

claude lover

[next trout]

WHAT

he actually got banned 😭

[silent tree]

rest in piss

onv

obv

why would someone post those

[echo aurora]

Thanks for flag

[silent tree]

Np pineapple

[surreal zephyr]

tahnks pinapple

[silent tree]

and yeah thanks

[surreal zephyr]

last thing he saw

before being abducted

[silent tree]

I wonder how many of these guys prob said "I HOPE THEY DIE!!! 😡"

we keep winning bro

no competition ig

[silent tree]

rip

[surreal zephyr]

i should rename to 5.5

@echo aurora come back

[next trout]

+rep mrbeast

does anyone have that segment of where sam altman talks about mythos?

[surreal zephyr]

codex

[silent tree]

PINEAPPLE MISSILE INCOMING

[surreal zephyr]

wheres that anthropic outage gif 😭

[next trout]

sam is such a cute eboy

[modern lily]

When will opus be back??😭

[silent tree]

Cant find 😭 😭

[silent tree]

when they release new usage system

but you'll eventually fall in love with gpt 5.5

you should

[next trout]

or otherwise...

[silent tree]

bro

[surreal zephyr]

he happy

[silent tree]

😭 😭

codex cute

[next trout]

not u obv

[silent tree]

Im not sam 😡

ampro is tibo btw

[surreal zephyr]

lies

im uhh

[silent tree]

u da real tibo

[surreal zephyr]

mario

openai hire me when

[surreal zephyr]

me good pr maker

[next trout]

is this the modern bonzi buddy?

[surreal zephyr]

🥀

[surreal zephyr]

clippy

[silent tree]

wait 30 years more

ok but why did gpt image 2 give him a good jawline

I didnt even prompt that

[next trout]

because that's how his jawline looks like

good model.

[surreal zephyr]

i need a good jawline

[silent tree]

img2 can generate any faces well but not darios lol

[surreal zephyr]

my pfp

[next trout]

u don't even need to generate him as ugly, just pull an image off google

[silent tree]

double chin

😔

start working out bro

[surreal zephyr]

quadtruple chin

[silent tree]

Infinite chins

[surreal zephyr]

look at my fih

[next trout]

https://tenor.com/view/im-a-fihh-fihh-miata-on-fihh-mode-im-wet-come-swim-with-me-gif-1581551123058247416

[surreal zephyr]

they grow up so fast 😍

[silent tree]

Prompt: Dario Amodei selfie

IT KEEPS GENERATING THIS RANDOM GUY IN EVERY DARIO PROMPT

maybe dario isnt famous enough

[next trout]

excluded from the training data, too irrelevant

[surreal zephyr]

dario so cringe bro

[next trout]

not good of a data

[surreal zephyr]

he has 0 following

like literally

ego

[next trout]

seriously tho, he's mogged by sam

like facemogged

[next trout]

fake

[silent tree]

AHH

WHO'S THAT

[next trout]

AI slop

[silent tree]

Dam Satman

[surreal zephyr]

HOLY name change

[surreal zephyr]

im going to /hatch this

[next trout]

he's such a mogger

[silent tree]

wait is sam in

openai?

[cedar citrus]

bye guys i know you guys dont want my presence but bye i wish you guys a great day im moving over to claude and salt server goodbye

[silent tree]

Is SAM IN OPENAI

DISCORD

SERVER

[surreal zephyr]

what

bro hes on twitter

[next trout]

u guys just bullied him into leaving LOL

[silent tree]

I meant discord serv

but there is a sama

people say it was him but he doesnt chat anymore

[next trout]

dm him and ask for free spud.

[silent tree]

LMAO

[surreal zephyr]

6o realtime will be peak trust me

[next trout]

i had an alt who was banned in openai server xD because i was sharing prompts of jailbreaking it in early days (2022)

[silent tree]

what if a mod saw @light sleet messages

he was keep on writing sam mogman

in openai 💀

[surreal zephyr]

Still easy to jailbreak

[next trout]

i don't think so, like, sometimes it refuses to even interact with my codebase

because it's "close to malware"

[light sleet]

yeah guys someone definitely told sam

I hope

sammy poo poo

was told

to change his name to mogman

with my reference image 😭

but that's imaginary ig

[next trout]

he changed it fr?

it is real XD

[light sleet]

yes he did

lol

i always posted sam mogman before xD

[next trout]

he's in this server./!?/1?!/1

[silent tree]

WHAT DOES MINE AND VINCE SERVER HAVE TO DO WITH THIS BRO

[surreal zephyr]

Theres other way, works consistently but its funny

[silent tree]

😭 😭

Oh yeah we forgot to invite Ampro

[surreal zephyr]

😭

[silent tree]

ampro pls no heartbreak

[surreal zephyr]

@silent tree we are braking up

[silent tree]

we'll give you VIP

[silent tree]

NOOOOOO

[surreal zephyr]

Im going to marry codex now!!!

🤔

[silent tree]

Im gonna give u a cool intro

join quick

check yo dms

[surreal zephyr]

Ok

[silent tree]

hey mods theres no unauthorized joining here

he wanted to join its his fault 😡

[next trout]

what model is #ask-here using?

[surreal zephyr]

Small one

[next trout]

which

apparently gpt4

[silent tree]

yo tibo give me gpt 4o 😡

dead chat rip

[surreal zephyr]

[silent tree]

arenanan

[silent tree]

are arenas models accurate

[light sleet]

guys atleast elon aint evil

If he was, Sam wouldn't be in X 😭

[surreal zephyr]

Real but bad harness

[silent tree]

bruh 😭

post my pfp in X

and tag sama

and then he'll put it as his pfp

[surreal zephyr]

Did the 4chin

[silent tree]

and it'll be tuff

[silent tree]

NOT THAT 😭

[surreal zephyr]

🥀

[silent tree]

pls sam altman and elon musk become friends 😭

musk has so much potential to create the world's biggest ai model

[surreal zephyr]

They used to

Before scam altman

[silent tree]

Ik

[surreal zephyr]

Also sam doesnt say hes mad

Sam invited him to the event

💀

[silent tree]

😭

[silent tree]

sam has potential to be BETTER

WHY CANT HE

😭 😭 😭

[silent tree]

what do u think about 🍓

strawberry guy

in X

🍓 🍓 🍓

[surreal zephyr]

Real?

[silent tree]

yes

its in yt

[silent tree]

was a year ago tho

[silent tree]

https://youtu.be/pDYZ4lRLilY

Imagine if this happens in the end

[pastel current]

how can I fix this problem that is in the photo? endless generation

[light sleet]

Okay so first

Create a new chat, type a long prompt and quickly switch back to the session where there's endless generation, the stop button will appear

Trust me @pastel current

[light sleet]

Quickly switch back to the endless generation session and the stop button will appear

[sly cedar]

Sam Altman speedrunning question is crazy

Bro is literal AI in human 💀

[light sleet]

😭

[sly cedar]

Why does he looks so wary lmao

[signal eagle]

gm guys

[light sleet]

Gn for me xd

[surreal zephyr]

Vs 0.15% human baseline

[whole mural]

so gpt5.5 might really be smarter thant mythos

OMG

[sweet knot]

When are real models like Claude Opus coming back to Side-by-Side?
I’m tired of these half-baked, useless open-weight LLMs that barely function — the same junk you can already find everywhere and that nobody serious actually uses.

[void shore]

I am training a new (maybe frontier) ai model

right now i am conducting tests to make sure it can match the intelligence of:

3B
7B (same as it)
13B
30B
70B
100B and more
1T

wish me luck

[storm dust]

yo bry what's with your site

are you still working on it?

[void shore]

stadium?

im trying to find better providers

and the link is https://stadium-ai.netlify.app

[storm dust]

yes

[void shore]

i think

[whole mural]

WOW

[void shore]

Currently it’s up to 13B intelligence

[toxic verge]

Is the government unintentionally promoting anthropic?

[unreal lion]

very nice

[whole mural]

hey i just want to ask

have anybody in this discord server used mythos before?

🧐

[loud herald]

Because hes a scamming loser

[loud herald]

Some say they have but its lies

[azure cove]

Can we add Claude 4.7

Oh wait only battle mode

[echo aurora]

Yeah that’s correct

[azure cove]

But why no direct mode

[pallid crypt]

And how would a solo person uh have the uh funds and the uh compute to train a model to match a trillion parameter model

Unless you have like insane unheard of archetecture

[shy isle]

is gpt 5.2 high, gpt 5.3, or gpt 5.4mini-high better

basically what im saying is out of all the gpt options on arena currently which is the best

[faint willow]

bro why is claude opus not on arena

[molten cipher]

Expensive af

[shy isle]

they keep adding and removing models

I miss when they had every best model lol

that was so op

[vivid coral]

how many blind battles can we do until we get limited?

[shy isle]

😡

[hollow comet]

I think the companies didn't like that their models were being used for free, so they removed top models like the Opus and GPT 5.5 from the modes where they could be selected and left them only in the battle mode.

[shy isle]

why is there no retry button

retried it manually and it didn't work again

should i change to gpt5.3?

=help

[oak python]

ModMail Help Menu

ModMail is a feature-rich Discord bot designed to enable your server members to contact staff easily.

Please direct message me if you wish to contact staff. You can also invite me to your server with the link below, or join our support server if you need further help.

To setup the bot, run =setup.

Invite

https://modmail.xyz/invite

Support Server

https://discord.gg/wjWJwJB

Use the reactions to flip pages. (Page 1/8)

[thorny schooner]

Just woke up to this

[loud herald]

So far from the truth lmao

Arena was losing money

Thats why

These companies PAY arena for what they do

For datasets

Thats how arena makes income

[shy isle]

OMG

nah what

is that fake

[tight meadow]

No

[shy isle]

I don't have it

[tight meadow]

Use Direct Chat then.

[shy isle]

bro how

is it only on phone?

[thorny schooner]

It's real just got lucky since it's still being rolled out for testing

[shy isle]

oh

[tight meadow]

I have it tho

[shy isle]

maybe it is a region thing

where do u guys live

[thorny schooner]

Since I think I'm pretty sure it was said before that they will gradually roll it out as they get data for the preview

[whole mural]

nah

[thorny schooner]

Nah it's an experiment so experiment style is with the random chosen accounts

[shy isle]

kk

did u try it out?

[tight meadow]

I ACTUALLY HAVE IT

I just used it and it works

[shy isle]

what does it do

takes control of your device??

[thorny schooner]

What?lol

[whole mural]

i think every body could use direct even guests

[thorny schooner]

I think i'm pretty sure that was changed though

[tight meadow]

It thinks firstly then gives you options.4 to be exact.

I am not kidding

[thorny schooner]

I can vouch for that

[tight meadow]

Ye

Sometimes it thinks

[shy isle]

what does it do

[tight meadow]

Thinks,gives you 4 options and it makes it

[shy isle]

hmm interesting

[sweet knot]

In EU it is not available?

[thorny schooner]

No it is randomized accounts and gradual rollout

[sweet knot]

So, in other words, Arena has turned into collection of weak open weight llms that nobody wants and uses? The end of the project?

I tried to generate simple lessons using all those Kimi 2.6 and all those deepseekish llms, they all failed. Why should one care? Before when latest versions of Claude Opus, ChatGpt were in the Battle mode, everything worked more or less. Now all the results are garbage 🗑️. Downgraded, totally.

[toxic verge]

I love Claude❤️

What’s your guys’s favorite rate limit with Claude?

[sly cedar]

Prompt: Hi, System: Sorry your rate limit reached

[rocky geyser]

The 5 hour rolling window its just so generous that you never reach it ❤️ Also opus 4.7 reaches the limit 35% faster its just so amazing 😍

[shy isle]

why do people glaze claude so much

compared to chatgpt

[prime karma]

@echo aurora hello

[round raft]

How to use image to video?

[prime karma]

@echo aurora Grok 4.3 is available in direct or side by side mode.

[azure cove]

ahh why I cant see

[river pendant]

use it in arena.ai video arena, by selecting this mode on platform, and upload image and type prompt of what you wanted to do with this image, notice, it's available for only battle mode

[midnight shard]

image to video

[median smelt]

Hmm.. (normal Gemini 3 Flash)

HMM.. (Arena Gemini 3 flash)

I dunno it kinda looks better.....

[marsh horizon]

is that direct or battle mode

[median smelt]

Direct

[marsh horizon]

ok

[toxic verge]

That’s so damn hot

[shy isle]

why did bro delete his message

[toxic verge]

Sorry, I got distracted by a really hot message

I love it when people talk dirty about Claude

Makes me wanna spend all my tokens

They’re all getting out of the way for the pharaoh Claude

[hollow cosmos]

@wary nacelle thanks a lot

🥺🥰

[shy isle]

wtf freaky ahh dude

[hollow cosmos]

https://tenor.com/view/ai-ai-chat-ai-chatbot-ai-meme-ai-memes-gif-1036783523923616791

[surreal zephyr]

eww benjamin

[hollow cosmos]

https://tenor.com/view/jjk-modulo-dabura-vs-mahoraga-yuka-okkotsu-gif-16832153206361704367

[median smelt]

https://019ded35-6def-7223-8457-c892b3ee3167.arena.site/ a random test i made

Gemini 3(?) Flash

[spiral surge]

How to make vedios from text to vedio? . I don't remember how to use it last time i made a vedio with veo 3 was November 2025

[vale quest]

I can predict the future

Opus is closer than you think

But another model is also in mind

<@&1349916362595635286>

Fake

Do not listen

Can yalls do smth

<@&1349916362595635286>

@echo aurora

Ok

[grand raft]

https://tenor.com/view/ah-shit-here-we-go-again-ah-shit-cj-gta-gta-san-andreas-gif-13933485

<@&1349916362595635286>

[gray lion]

https://cdn.discordapp.com/attachments/353622923568873473/1455329871416000687/fax.gif

itys happening to so many people

[clever glade]

elon's scam is better than mr beasts

[vale quest]

Since when bro

Since when can it plan

[loud herald]

Since literally always

[vale quest]

I live under a rock

[surreal zephyr]

lmao

geniuely how would anyone fall for that scam when its sent 3 times in row

[gray lion]

yeah ik

why do they even still do it

everyone knows abt

[loud herald]

Because its not the actual person on the account

They downloaded something to their device and now have a trojan

Then the hackers use a bot to automate those messages everywhere

[gray lion]

yes ik but like why do the hackers still do it

[loud herald]

Because its an advertisement scheme to make money

Why would they stop

[gray lion]

not effective

[vale quest]

Why do people just do that

Its obviously fake

Spends up time downloading the images anyway

On the hackers end

[loud herald]

People aren't hacked from doing that

[loud herald]

Copy and paste, its literally a system they run

[loud herald]

yea, claudes not safe in any way

Never has been

[vale quest]

Claudia

[long minnow]

Never ask a man his salary
Never ask a woman her age
Never ask a bot where to generate video

[surreal zephyr]

10000000$ it was opus not gpt

[vale quest]

Ye

[rocky geyser]

Has gpt gotten better at coding? Bought gpt plus once couple of months ago and it sucked, is it better now?

And im not using claude, i have pro, asked opus to code one single thing and my whole usage limit for 5 hours got wasted in one prompt and it didnt even finish because of my usage

[surreal zephyr]

bro gpt is the only good one

[whole mural]

yep

its already better than claude

[whole mural]

how is this possible?

[unkempt juniper]

It makes my brain think something ugly when I ask some of the models to detail something if they can get the details right without making something up

[hollow ivy]

is gpt-5.5-high better than Claude Mythos 5?

[unkempt juniper]

Idk

We don't even have access to Mythos

[nocturne tangle]

guys how to use gpt 5.5 and claude opus 4.7 in arena ai?

[unkempt juniper]

4.7 is not available in side to side or direct

Only in battle

[queen veldt]

If you type this prompt into chatgpt and select create image you get cursed images loll

[shrewd citrus]

bro the 3rd photo 😭😭😭😭

[stray aspen]

I tested the happyhorse

Its so good

[prime karma]

@deep osprey hello

[gray lion]

i tried it lol

[queen veldt]

I love how it's always some unique one

[median smelt]

HappyHorse

[light sleet]

happyhorse 💀

[long minnow]

🐴

[wintry sleet]

hi

[feral kernel]

is there a way to turn off random battle modes in regular direct chat cause its extremley annoying

[golden ocean]

that 4th one with the teletubbies

thats like almost the original image from training data

(this is the real image from internet)

bro got it to generate images fromm training data (idk what im talking about)

[golden ocean]

is this on lmarena🥺

[split topaz]

I was blind testing in artificial analysis video arena battles and given random videos to comment on. You can totally smell HappyHorse from the quality of the outputs. The famous big model smell.

[median smelt]

I think its already there, but im using a discord bot

Cause its only on battlemode

[queen veldt]

Umm

This is what gpt gave me

Legit same image

[nova torrent]

I’ve been working on full-stack systems that deal with real-world scaling, access control, and data-heavy workflows.
One recent example was a Hospital Management System. The core challenge wasn’t the domain -it was making sure the system holds up under actual usage with multiple user roles and sensitive data flowing through it.
A lot of SaaS and internal platforms run into the same issues once they grow.

[remote marsh]

what up with the web site Security Verification
fails

[jolly pumice]

Hello

[karmic silo]

i have to pay to create images now????

[austere barn]

why is this happening

[remote marsh]

same i cant use gemini

[wide smelt]

did arena break again cuz glm 5.1 wont work even if it aint a rate limit

[austere barn]

same

[wide smelt]

i tried starting a new chat and its the same case

[fluid ruin]

GOD DAMN IT

STOP THAT

[fluid ruin]

YEAH AND IT SAYS UNAUTHORIZED IN THE CONSOLE

[remote marsh]

notthing working keep loging me off and read terms of not working cant

[austere barn]

use claude sonnet 4.6 then, but if you keep asking to edit the same website that you've generated in the same chat, it'll say that you've ran out of credits or smth

[limber crag]

omg it just gave me real life skibbidi toilet lol i was not ready for that jumpscare

[wide smelt]

oh sonnets available in direct? awesome

[queen veldt]

Sounds interesting

[remote marsh]

hmm

[queen veldt]

Oh right

[surreal zephyr]

the last one is cursed

the 3rd is, uhh wtf

[limber crag]

yo why does your lmarena look different from mine!

[surreal zephyr]

uhh i got idea where this one might be from

[austere barn]

the

[surreal zephyr]

the the bad

[queen veldt]

Because arena is always doing a/b testing

They tweak some stuff to random amount of people

[austere barn]

wait there's an agent mode as well?

[queen veldt]

Yeah

I can upload files and stuff

[austere barn]

cuz i don't have it for some reason

[queen veldt]

Well it's a/b testing

Some got it some not

[remote marsh]

fix it

[austere barn]

fix what

[remote marsh]

cant use any models web site Security Verification
fails

[queen veldt]

Do you know what a/b testing is?

[austere barn]

i do

[queen veldt]

Mhm

[austere barn]

it's where you choose which model gave better results

[queen veldt]

Ooookayyyy

[austere barn]

did you not know about that or what

[queen veldt]

A/b testing can be done with anything

Some companies make different prices like 1.99 2.99 0.49 to see which group is most likely to pay and conversion rate

Here on arena they're doing a/b testing with features like agent mode to see how likely the user is to use it and to see how he uses it

And they give a certain amount of people the agent mode to see how they will use it to see if it's worth having it here

Also to check for bugs and stuff

A.k.a. partial release

[storm dust]

when is delivery mode

[fluid ruin]

God damn

How

[queen veldt]

How what

Okay it works nicely

[fluid ruin]

Bro how

Tell me damn it

Mb

I mean

Tell me your ways

[queen veldt]

There are no my ways i just got it lol i got lucky

[fluid ruin]

Tell meee

[queen veldt]

It'll release to more people probably later on

[fluid ruin]

agentic flag

[storm dust]

when is agent molde version 2

[mystic maple]

Battle in direct it’s a stupid update. I put “Direct” specifically because I wanf to chat with only one model

[last trench]

How to get agent mode?

[queen veldt]

I legit explained

Above

What is a/b testing

[fluid ruin]

This model is not permitted to handle this type of question. Please choose another model. This model is not permitted to handle this type of question. Please choose another model. This model is not permitted to handle this type of question. Please choose another model.

[queen veldt]

Well it's not a multimodal model

Maybe the model can handle text only

[fluid ruin]

BRO

[fluid ruin]

THAT IS TEXT

FULLY TEXTY

[queen veldt]

What did you ask maybe you added some images above in the conversation

[fluid ruin]

No image bro

[queen veldt]

Try different browser

Mozzila

[fluid ruin]

Ye different browser works but

[queen veldt]

Chrome

Arena is buggy and confusing not even the devs know how it works

I realized that image generation problem that happens a lot was caused by a lot of people using same api key for image generation

But idk what's stopping them from fixing it lol

Add more api keys duh

[fluid ruin]

They need claude mythos

[queen veldt]

Yes they will for sure add it

It's more expensive than o3 pro lmao

We don't even have opus here 😭

[unkempt juniper]

Opus is gone T-T

[hybrid tinsel]

Is glm 5.1 glitching or is it just me?

[surreal zephyr]

5.5t is better than o3 pro

and 5.5 is absurdly cheap because it uses no tokens almost

[tawdry dove]

truee

[narrow elm]

Where to generate video with veo 3 ?

[uneven cypress]

@echo aurora I think the developers should add opus back to direct chat with rate limit 10-20 messages or all the top models with 10-20 messages

[sharp mirage]

Yo hi all

[sharp mirage]

Where is the point

Like there is no point

[fading heath]

anybody need a workflow ?

[sharp mirage]

Idk

Workflow for what ?

[surreal zephyr]

20 opus messages is more than you get in claude max 200$

lol

[sharp mirage]

Yea fr

[surreal zephyr]

pro gets 0.7 request per day
max5 gets 5 times more
max 20 gets 20 times more ~= 15 reqs per day

(with reasoning on)

[sharp mirage]

Bro if they used the api there is no input,output rate like i can send a whole apps and data in 1 msg

[queen veldt]

You mean more than

[sharp mirage]

If they need to make a rate limit they have to do it on the input output tokens

[surreal zephyr]

yeah lol

[queen veldt]

That's still crazy low lmao

Not worth it

Gemini 3.5 pro will come anyday now and it will be on par with opus lol

[surreal zephyr]

me chillin with gpt 5.5 on pro 100$ nearly unlimited

[queen veldt]

Anthropic has gotten really greedy

[sharp mirage]

Me chilling with gpt 5.3 mini with 0$

[surreal zephyr]

i prefer 5.5

[sharp mirage]

With 0$

[uneven cypress]

I didn't know that

[surreal zephyr]

what

[surreal zephyr]

you dont have it

"toyota carola can reach 500 kmh on highway" ahh sentence

[left quartz]

lol

[surreal zephyr]

blank response?

[left quartz]

no

blankies

[remote marsh]

happen again

[storm dust]

hmm

i will guess

[storm dust]

what about this? how do you know what i will have and what i wont

[tranquil holly]

hi

[storm dust]

i change my mind

[narrow dawn]

how to fix that?

[uneven cypress]

Arena ai cooked with this one

[void shore]

im finding a way

i'm cooking with this model

[pine jolt]

fascinating

[wild blade]

Rap fans 👋🏻
Listen this crazy flows all by ai beat by me

@everyone

[narrow dawn]

how to fix that? like is there way without losing the chat

[wild blade]

If you have signed in then the chat will be there it won't vanish

[narrow dawn]

ye but

the generating thing

is forever

[wild blade]

Copy paste the same prompt on a new chat

[wild blade]

And have some patience it will give you the result you can look at other things it will run even if you aren't on that tab

[wild blade]

Use a different model instead

[narrow dawn]

i cant send messages with other model

[narrow dawn]

bro its already like 4-6 hours

[narrow dawn]

. no i need all the old things there

else i need to do my whole preject from scratch

[wild blade]

Oh I can feel you brother

[wild blade]

Must be frustrating

I guess you have to have i guess

[narrow dawn]

😭

i did a new chat

first thing after first message

crazy

this is site is sometimes like broken or smth

[wild blade]

Or else you can try to build everything form scratch i mean 6 hours is a lot of time i think you could re build everything

[narrow dawn]

this project is already more then a 1 month

i will not do everything from scratch

[wild blade]

It's a leader board site where you get all the top ais at one palace for comparison so you can't complain this site ...

[narrow dawn]

ye but

its broken

its almost completly broken

u cant even like press the stop button at the infinite generating bug

[wild blade]

Oh so you are trapped

[narrow dawn]

YES

[wild blade]

What project are you buliding?

[narrow dawn]

an big one

i cant explain it fully

[wild blade]

May be if you disclose it anyone here could help or join in coders server they might help you

[wild blade]

Like development??

[narrow dawn]

.

i dont want to get an other developer for the project

[wild blade]

I'm a software developer too I'm a MERN STACK DEVELOPER

[wild blade]

You don't need to just ask help and get rid of the problem

[wild blade]

There as always step for using ai
You ask ai to give answer in snippets

So that it won't hang or crash out

[narrow dawn]

._.

[wild blade]

Share me your GitHub

I have a look ...

[narrow dawn]

no

i am not using github

[wild blade]

So where you do save all of your code locally??

[narrow dawn]

no

[clever glade]

openai

[wild blade]

Good

[wild blade]

You can't work like this

[tired herald]

Actually saving stuff locally is kinda what one should do 💀

[narrow dawn]

i am using an server hoster

[wild blade]

I would suggest one this take snippets of the bugs you have in your code and paste it in a new chat and see

[narrow dawn]

i dont have any bugs

the chat of the ai is bugged

bro are slow?

[analog blaze]

Hello

ppl

[wild blade]

It's a better practice to push your code on git hub after you save it locally cause GitHub is for backup, version control, and collaboration. Think of it as a safety net and history log for your local files.

[wise flower]

why cant i see the nwer midels on lmarena now? sorry im ysing this after a long time 😭😭😭

[wild blade]

Ohh

[wise flower]

oohhh kese ho khaan sahb

[wild blade]

All good 👍🏻

[wild blade]

Kha se ho ? So you know me ?? Or is it just by my name ??

[tired herald]

I know. I use gh myself, it's just very bad practice not to have something locally that one is working on

[wise flower]

nothing i just called your name 😭😭 im from pakistan

[willow swift]

are gpt 5.5/claude opus models in search mode (battle) thinking variants or non thinking ?

[wild blade]

You automatically save things locally isn't it ?

[wild blade]

Give it a listen

[wild blade]

.

[tired herald]

What? No, especially some people who use ai don't actually save it anywhere locally.

[wild blade]

What ?? If you don't save it locally how would you test it ???

[tired herald]

There are ways to do it online

[rose tendon]

rq best performing chat model is what ?

[primal orbit]

tetra-0429-3 is broken and gives empty answer

been getting it 3 times in a row

[teal forum]

Can we use claude opus 4.7?

[cobalt smelt]

Hola

[void shore]

im trying a second time

i have refined my data further

and have increased training time from 30 minutes to 45 to an hour

[topaz meadow]

hi

what's the limits

[pallid furnace]

hii

[topaz meadow]

how are you

[stray aspen]

All ai videomodels are so cinematic

None of them have the essence or natural feel of sora 2

[remote wharf]

Each ping I get in the next hour I will do 5 pushups

[hollow ivy]

yes, in battle-mode

of course, you would first need to learn how it answers, to recognize it early (before voting)

[pastel mesa]

Just like both

[stray aspen]

what th ehell

wqwen 3.6 max is actually kinda decent

for the first time

[void shore]

Yet another update

the training run has finished

and ill let the results speak for themselves

i just need to get them

it got one of the 3B problems wrong because it was "too smart" and used mammal instead of animal

wait

the benchmark was flawed

im gonna get a better benchmark

and then verify

[tardy veldt]

ask i wanna wash my car and there’s a car wash 10m away do i walk or drive?

[whole mural]

img generated by gpt image 2?

[pure niche]

Hello

[void shore]

might not get that right

but i hope it can figure it out

[languid hamlet]

hey guys

just ran something and got a model called kartoffeln

can't find it anywhere

anyone know what that is

[grand raft]

i think that's chatgpt's spud

[languid hamlet]

5.5?

[whole mural]

nah

maybe 5.6

or codex 5.5

[languid hamlet]

May be

It’s really good

Better than Claude and gem with front end

Which is surprising for gpt

[tired herald]

Potatoes 😭

[short vector]

holas...abrazos

[austere barn]

is glm 5.1 still giving errors

[surreal zephyr]

https://cdn.discordapp.com/attachments/1212420778000126012/1267869077737701461/maximum-tension.gif

[grand cliff]

Claude sonnet is acting up

Wouldn't generate a response on 3 separate chats.

2 long ones (I know still has room). And a new one

Nevermind

[narrow elm]

How to get api of higgsfield ai clone

?

[merry cloud]

Yo what happened to video generation?

[grand cliff]

Wtf happened to the rates. I got rate limited so fast in Sonnet 4.5.

In just like...10 generations?

Used to be like double that

[shell fox]

I have a question about direct chats and LLM's self-awareness. As you can see, those models are "pretending" to be a different model. I asked some Claude models "What model are you?" both in Arena and Claude official website and in Arena they were either lying or hallucinating. Once I asked the same question to Sonnet 4.6 in Russian and he responded that he is Deepseek V3 which is pretty hilarious. Can you explain what's the cause of this thing?

[willow swift]

are gpt 5.5/claude opus models in search mode (battle) thinking variants or non thinking ?

[shell fox]

@echo aurora

[shrewd citrus]

it’s because these models don’t have access to real time data

so they have no knowledge of a new model (their model) being released

if you asked a search model (one which has access to the internet) then it will 100% get it right

[late sigil]

it's just frustrating to not know , i waste 30 mins trying to get any high quality model but i don't think they're thinking

[shell fox]

I checked web-search. Still the same

I mean, I could see the difference in those models when they are writing code, but I'm not sure those models are the models we are reffering to

[shrewd citrus]

ok never mind i think Claude is weird 😭

[compact flame]

It doesn't exactly know what it is

[late sigil]

system prompts

arena.ai system prompt doesn't tell the model who it is

anthropic tells the model who it is

unless quality degradation you notice is very high then maybe something under the table is going on

[shell fox]

So the guarantee that the models Arena provides are real is trust-factor?

[shrewd citrus]

there’s no point in them providing the wrong model

the reason they make money is on data

and they need specific data for specific models

[hybrid tinsel]

Yes !

[shell fox]

you've got point ig

thanks

[golden ocean]

Large Language Model

[compact flame]

Large large model

No language but large

Or

Language language model

[neat island]

Anyone have numinfo app

[fossil socket]

It's been a month since opus and the other high tier models got removed

Haven't really used this website much afterwards

[unkempt juniper]

Eh

I ran with Kimi and whatever models I can use

[serene breach]

Bro how to create a video

[dense moon]

My chat history is suddenly gone... The second time this year. 😢 😭

I really should backup them more often but it is so annoying... Copy a message, switch to the text editor, paste, switch back to the browser, repeat for every single message.

[brazen dagger]

site off ?!

[surreal zephyr]

50% pro weekly remaining, gonna summon subagent swarm 🔥

[south charm]

off topic :)), my AI is rading now

[low stump]

why arena doesn't have hunyuan image 3 plus yet ?

[willow swift]

anyone having issues with glm ?

[surreal zephyr]

you possibly picked two worst models in existence

just use 5.5 low seriously

[indigo knoll]

What? Is Deepseek V4 bad?

[surreal zephyr]

the whole point of deepseek v4 is to think for hours at low price

and 5.3 instant is for chatting only, so its garbage

[stray aspen]

yes

[indigo knoll]

But I'm considering general chat not only thinking tasks

So what's the best free model rn if it's not Deepseek?

[surreal zephyr]

qwen 3.6 27b

or deepseek v4 at HEAVY THINKING

[rocky geyser]

claude pro sucks 😭

Never buying it again

[supple scaffold]

Me, since 2 days already

[surreal zephyr]

claude sucks*

[indigo knoll]

It's the model that we use in Qwen Chat right?

[surreal zephyr]

you can use it locally

[slender ledge]

Is there an website like arena with opus

[indigo knoll]

I don't like to use local lm, only in the websites

Especially because my PC has limited storage space

[carmine shard]

I vote on DS. Don't worry about it. If you just do normal conversation, DS 4 Pro is better. GPT will give you terse and succinct response by default. I use DS 4 Pro for philosophy, analytical reasoning and dialectic exchange. I can tell you it's nearly on par with Sonnet 4.6. GPT 5.3, esp instant can't go anywhere near. Alternatively you can use GLM 5 or MiMo 2.5 Pro

[loud herald]

I am surprised they haven't added Ling 2.6 here yet

[light sleet]

@silent tree u got the alert too in 4:59pm xd

basically 30 mins ago

[loud herald]

The API is free until the 7th too

[light sleet]

Missile Threat 😔

[silent tree]

Yea 💀

[light sleet]

so it starts again

[silent tree]

maybe

[loud herald]

Well anyone who picks a model that is 2 generations behind the other is just braindead, deepseek is SOTA currently, a model 2 generations behind doesn't stand a chance

[surreal zephyr]

both are garbage

deepseek carries hard by self correcting over extremally long COT chains

but its more expensive than 5.5 low by A LOT while dumber

[loud herald]

How expensive is 5.5 low every 1M

[loud herald]

Why even lie

Deepseeks biggest flex is that they're so cheap

[surreal zephyr]

how about check the token usage per task lmao

[loud herald]

Of course a cheaper model will use more tokens per task when its cheaper, its affordable. If anything its a bigger flex that it does use more tokens and is still 7x cheaper

[surreal zephyr]

too bad its more expensive

lmao

[loud herald]

I just showed you its not

[surreal zephyr]

i literally shown you that its logarithmically more expensive while being dumber

[surreal zephyr]

^^^

[loud herald]

You're comparing a Pro high and Max model to a Low model and saying its more expensive when its literally not, you're spewing nonsense

Deepseek is the better choice any day, you could put $20 in and it would last for months

I care more about pricing then a little more intelligence

[polar pendant]

I am sorry DS 4 Pro is... ?

[loud herald]

Deepseek

[polar pendant]

Ok. Thank you

[surreal zephyr]

how hard is it to read

two models at same capability

one over 2x more expensive for that same task

????

[loud herald]

Hence "I care more about pricing then a little more intelligence"

[surreal zephyr]

vertical axis = capablity

horizontal = price

5.5 low is much cheaper while being slightly smarter

[loud herald]

CoT was how V4 was trained, its going to do more per task but overall will be a lower bill than anything GPT releases. Thats why people look at per 1M tokens, not per task because per task doesn't matter at the end of the day, GPT will always be more expensive

What do you not get

[surreal zephyr]

what do you not get

it uses like 50x more tokens per task

so it ends up beign more expensive

as literally proof i shown you says

ragebait?

[loud herald]

By the time it hits 1M tokens its not lmao thats my point

Its still cheaper

[surreal zephyr]

LEFT IS CHEAPER

OMFG

[carmine shard]

Normally, yeah. If the choice was GPT 5.5, I'd say it's on par, but I'll still go with DS, for the same reason. It's no go for my use case. Not to mention on free or lower tier, I'll face limit and it gets rerouted.

Can't really say on coding or math use case, which maybe why some still pick GPT 5.3, idk

[surreal zephyr]

😭

[loud herald]

Per task, overall its more expensive

As a low too

Jeez

[surreal zephyr]

which makes it worse

how hard is it to get

bro

more expensive + dumber

???

[loud herald]

And why would I be surprised if a low model is cheaper than a high model per task

[loud herald]

GPT 5.5 pro must be real dumb then, holy moly!

[surreal zephyr]

because the low model is SMARTER

price per token is meaningless when price per task is 2x worse lmao

[loud herald]

The amount of token usage doesn't determine the cost either

[surreal zephyr]

blocked for ragebait

[loud herald]

It could use "50x more tokens per task" and still be cheaper

[surreal zephyr]

token cost * amount of tokens = price

4 times cheaper per token, 50x more token usage = more expensive

[loud herald]

Ah yes, ragebait because I am correct. If token usage all had the same price, 1M tokens wouldn't be $5 for deepseek and $17 for GPT 5.5 low

It'd be the same price

Thats what you're saying here

[native yarrow]

yeah noire stop ragebaiting this guy bro

[loud herald]

Buddy

I am NOT ragebaiting

He isn't mathing

[native yarrow]

brooo ur js ragebaiting stop brooo

[surreal zephyr]

literally i gave him evidence

and he refuses to read

😭

[loud herald]

The true ragebait:

[loud herald]

Your evidence isn't evidence

[native yarrow]

ur PISSIN me off. im havin to grt the shower goin its 7 in the frickin mornin and youre PISSING me off

[loud herald]

So get the shower going little guy

[surreal zephyr]

https://giphy.com/gifs/mrw-end-state-7rJE4vH3ItKDu

[native yarrow]

dont tell me what to do.,.,

but i will. completely on my own accord.

[late sigil]

why the fk glm 51 is not working for fk sake

[loud herald]

On what?

What are you using it for?

[late sigil]

i keep getting that error while generating response , only when i needed it most

[late sigil]

arena , because is the only reasonable place to use it but not anymore i guess

[willow swift]

bro left the server 💀

[loud herald]

So send the prompt again and before refresh

[loud herald]

Goodness

He won't become an AI enthusiast acting like that

[native yarrow]

and dont even think about becoming an ai champion..,,,

[willow swift]

[silent tree]

💀

5.5 better

[native yarrow]

dont ragebait this guy noire dont do it

[silent tree]

idgaf honestly

idk what you'll get from hating gpt

any rewards

nah

[silent tree]

little guy 💀

how do u even know if hes a little guy

weirdo spotted

[loud herald]

Because hes a close friend of mine

Clueless weirdo spotted

[silent tree]

so your friends with young people

damn

didnt know

my bad

[loud herald]

I am 16, why wouldn't I be

[silent tree]

no right to call him a little guy

if hes your age

💀

unless hes 13

[loud herald]

Little means height not age, what are you trying to achieve here

[silent tree]

maybe in your language

im not trying to achieve anything here

[loud herald]

Right, thats why I spoke it to him in my language

[silent tree]

im just here to remind

everyone

to switch to 5.5

💀

[loud herald]

That scam altman is about to be 134 Billion in debt?

[silent tree]

btw

its not always about sam

its about gpt here

not him

where'd he come from?

[loud herald]

Btw, Scam is indeed the face of OpenAI, it is

[silent tree]

💀

who are you?

[loud herald]

You'd have a lot more people switch to GPT if you weren't so aggressive when saying it

[silent tree]

claude user?

[loud herald]

Oh my lord

[silent tree]

I dont think I was aggressive tho

is that also in your language

whats your language

😭

[loud herald]

Its the way you talk, its very aggressive

[silent tree]

im talking non formal

[loud herald]

You come off as a know it all to begin with

[silent tree]

It doesnt mean

im aggressive

[loud herald]

Being aggressive doesn't mean non formal

[surreal zephyr]

hes a ragebaiter, dont bother

[silent tree]

I can tell

[loud herald]

I'm not formal with anyone here and I don't come off as aggressive most of the time

[silent tree]

sorry I hurt your feelings 💀

thought it was in your language too

😭 😭

[loud herald]

My feelings weren't hurt, I was just letting you know

[silent tree]

no i just thought

it was in your language too

my bad

[loud herald]

You keep saying it as if it makes sense

What are you saying?

[silent tree]

your language

[loud herald]

They're words in my language yes

What about it?

We're speaking my language

[silent tree]

yes

100%

[loud herald]

Lmao

[native yarrow]

guys i totally dont know this guy

[loud herald]

I overstated when I said "close friend"

[native yarrow]

OH COME ON

[silent tree]

so thats in your language too

[surreal zephyr]

https://cdn.discordapp.com/attachments/1138429463927984159/1251460592082882581/togif-1-1.gif

[light sleet]

lmao

[loud herald]

acquaintance at best

[native yarrow]

now thats just MEAN

[silent tree]

In your language too?

nice

[loud herald]

Yes, in every language

[silent tree]

your language is english 0.3

[native yarrow]

im bros acquaintance in every language

[loud herald]

That makes no sense but sure

[native yarrow]

this is devastating

[silent tree]

shakespeare

TotallyShakespeare

[surreal zephyr]

milkshake spear

[silent tree]

shakenoire

[loud herald]

Does he not know what an acquaintance is or something?

[native yarrow]

wait maybe in Congolese are we best friends .,., @loud herald

[loud herald]

Wrong noi

I'm the true noi

[native yarrow]

NOO

OMG

THATS

onenof the worst things thats ever happened to me

[silent tree]

in mine and your language yes

[surreal zephyr]

why did u ping the better noi

[native yarrow]

answer me boy.

[silent tree]

hes right

[loud herald]

Oh so you actually meant to ping me

[surreal zephyr]

no he didnt, he made a typo

[loud herald]

🤣

[silent tree]

@low kayak doesnt need his language

[surreal zephyr]

look, he edited it now

[native yarrow]

yes chud

omggg this guys so stupiddd