general

[obtuse smelt]

oh yes is captcha strikes again ?

[shy isle]

does anyone know why sometimes when you ask a maths question the equations don't display correctly printing stuff like this:

e^{2x},dx
\int xe^{2x},dx
]

instead of the equation formatting

[grand raft]

oh

[obtuse smelt]

nice code

[sullen sable]

👍

[surreal zephyr]

😭

[shy isle]

?

[compact marlin]

Hi

[wary echo]

what you think about this https://discord.com/channels/1340554757349179412/1498271433371943042

[brisk helm]

upvoted

[gleaming wraith]

does anyone have any idea how long it usually takes until a newly added model gets added into the leaderboard (talking about gpt 5.5 in this case)?
Its been in battle mode for quite a bit now

[strong ferry]

Are GPT image 2 and Nano Pro down right now? Keep getting errors.

[toxic verge]

I think they got rid of it in direct mode and the same with a lot of the top models from what I understand

[gleaming wraith]

I just want to see the updated leaderboard, idc about using it here

[toxic verge]

Oh I see apologies sir I thought you were asking the opposite

[spiral trench]

hello

[fickle sable]

Plz how to generate image-video

[stray aspen]

Go to the video arena

[fiery crow]

grok ai video generation

[vital mantle]

I hate it

so u bought 3 gpt 5.5s

[toxic verge]

That sucks bro

[shy isle]

dang

[toxic verge]

If the arena is feeling the crunch. You know it’s real.

[toxic verge]

How much is paying too much for AI in your opinion?

[slender ledge]

40 euros

idk

[shy isle]

ai should be free

[keen beacon]

w.wiki/MHef Until this comes true, the enshittification of absolutely everything, including AI, will keep worsen

Unfortunately

[shy isle]

hmm

[slender ledge]

Holy

I found an website unlimited opus 4.7 for free

[heady kite]

Nothing is free

[slender ledge]

Yes

And gpt 5.5

And 5.4

[versed kelp]

what is it

[slender ledge]

5 Reacts and I slide

[heady kite]

Do any of the open models actually compete with the proprietary ones? GPT 5.5 Thinking etc are agentic and I'm wondering how you replicate that

[slender ledge]

All free there

[versed kelp]

wats da sitee

[slender ledge]

5 reacts lol

Also opus 4.6

[heady kite]

Are you "just seeing this" or can you confirm that they actually perform? Because I can spin up a site that says it has GPT-6.

[slender ledge]

And gpt image 2 unlimited

[slender ledge]

What should I ask

So I know it’s real opus

[slender ledge]

4

Ima tell

19

10

8

7

6

5

4

3

2

1

Genspark.

W Slido

[sly cedar]

How many prompt limit?

I noticed there is subscription

Which is unsure for me

[slender ledge]

For video

And other

But gpt 5.5 and opus 4.7

[sly cedar]

Does Genpark using fake models?

[slender ledge]

All unlimited

[slender ledge]

I don’t know

[sly cedar]

I'm quite unsure about this website, cuz how they provide free unlimited ai models?

[slender ledge]

So u buy the plus

Plus is video too not unlimited

[slender ledge]

Arena also did

[sly cedar]

Well they do provide free, but is it legal they do that or they have better financial?

[keen beacon]

This is fake Opus 4.7

[slender ledge]

No

[sly cedar]

Im suspicious

[slender ledge]

I belive it’s the real

[sly cedar]

How it effects your project

Is it normal?

[keen beacon]

This is how real Anthropic models look like

[sly cedar]

Ima test this

[keen beacon]

Normally there will be a big # at the beginning of replies

[sly cedar]

To make sure

[heady kite]

Tried their "super agent" mode got this

[keen beacon]

This does not have that.

[velvet furnace]

The same idea

[velvet furnace]

where

[slender ledge]

Genspark

[heady kite]

Try sending one message with Opus and you get the pay up popup

Nothing is free

[slender ledge]

Nope

[heady kite]

You have credits in the bottom left. Mine went to 0 after the first message

[slender ledge]

Look

[keen beacon]

Wrong mode. AI Chat is indeed entirely free

[slender ledge]

Nope

[heady kite]

Yes I was using agent mode

Good point

[slender ledge]

It’s real opus

[heady kite]

Tbh I think agent mode is the only one that matters fir complex tasks

[keen beacon]

But they hacked the system prompt of some $#!+ slow model and call it Opus 4.7

[slender ledge]

Nope

[heady kite]

Well I get the pay popup every time now... have fun. I will just use the real models

If you have good agentic results with an open model please let me know

[slender ledge]

Use ai chat

[slender ledge]

it’s real

What should I ask

So u guys know it’s real

[keen beacon]

"Comprehensive detailed list of all models of computation."
Arena Max vs. Genspark AI's Claude Opus 4.7

[slender ledge]

NVM

[keen beacon]

For that I got 913 lines with Arena Max ("Response provided by Anthropic") vs 194 for Genspark AI's Claude Opus 4.7

[slender ledge]

It’s not unlimited

oh then it’s fake

[sterile tartan]

I have to some really complain about Genspark

[slender ledge]

wym

Yea it’s not unlimited

[sterile tartan]

Im just saying Genspark is not Trustable

[rocky geyser]

What do you think about agent mode in arena?

[stable sage]

Do anyone tried arena.ai/agent mode I just saw it and tried to check that preview and gone

[sterile tartan]

Don't buy their subs

[rocky geyser]

I still have it i think its alright

[stable sage]

For me not opening, even the url just starts redirecting to arena.ai

[stray aspen]

what subs

[whole sundial]

i assume these ones: #general message

[pseudo hemlock]

Hi kiri

[low quiver]

Roblox uses the Luau scripting language, not Lua 😕

[echo aurora]

Good morning/afternoon/evening

[echo aurora]

This is an experiment at the moment, so it's going to be random if you see it or not. https://help.arena.ai/articles/1811908126-arena-experiments-agent-mode

saw it and tried to check that preview and gone
Did the generation not work? Or did you leave the page and when you came back it was gone?

[whole sundial]

you should honestly just let users opt into experiments, windows randomly put people into experiments in their beta builds for a long time but now they are allowing people to opt into them

[stable sage]

I gave the prompt it showed me chat screen like generation and suddenly closed the mode and disappeared entirely

[echo aurora]

When it closed, what happened? Back to the home screen and didn't have Agent anymore?

I couldn't give an early heads up sorry to say.

[echo aurora]

Could happen. This is something we've heard before. Problem being we need experiments to acurrately reflect how users respond to them. Having some kind of opt-in opt-out could lead to an inaccurate understanding of an experiment's results.

We need the votes and to validate the data. This process can depend on how long it takes from arena to arena.

[stable sage]

Closed = back to home, Closed and showed battle mode home page and I couldn't see the agent mode even I tried open directly through url /agent doesn't even showing.

Even I couldn't see it in canary too

[echo aurora]

Okay good to know. Checking with the team and this seems like it's a bug. I'm going to start a post in #1343291835845578853 and will followup with you there as we wouldn't want this lost to #general .

[frank spruce]

If you see a suspicious Discord login without the word "Discord" in the search bar, then don't login; this is the chance: it is a scam.

[pseudo hemlock]

PINEAPPLE

do yk how many votes gpt 5.5 has

or when it will be on the leaderboard

[vernal raft]

Ah do that's why I saw it one time and no more?

It's random?

Damn I didn't even had the chance to use it

[obsidian cargo]

gpt-5.5 high below muse spark??????

[half veldt]

damn

[gray lion]

wow

[silent tree]

ofc

[pseudo hemlock]

I was just about to say that

thats so interesting

[silent tree]

chances of getting gpt 5.5 was so low

how would u even vote??

if only direct was there

[upbeat ferry]

how is that possible

GPT 5.5 is the frontier model that rivals opus

[split kayak]

how to come up with good prompt battle

[silent tree]

It's cuz

[echo aurora]

Right now

[silent tree]

the chances of getting gpt in battle

were so low

its literally hard to vote 😔 😔

[gray lion]

i got it enough times

[upbeat ferry]

We need to vote for lord sam

[silent tree]

yeah but

[arctic lark]

Gpt 5.5 is so low on the leaderboards, it makes no sense. It's waaaay better at coding than opus 4.7 in practice but my god is it low on the leaderboard relative to it???

[silent tree]

its not even possible

muse spark is so trash

[upbeat ferry]

I want team red vs team white

[silent tree]

how the hell?

[glacial mulch]

doesn't make any sense to me, i am pretty satisfied with 5.5

[whole swallow]

How it’s lower than muse spark on coding

[upbeat ferry]

I don't want another team specially with a blue color

[silent tree]

@surreal zephyr

[whole swallow]

Is there an api endpoint for muse spark

[silent tree]

yo gpt-ampro 4o

[cinder jasper]

Gpt 5.5 is not that bad wtf?

[silent tree]

It isnt

its rigged ig

😔

[graceful sundial]

It loses even sonnet 4.6 LMAO

[scenic loom]

Lmfao

[cinder jasper]

The leaderboard is rigged as f*

[whole sundial]

no

[upbeat ferry]

GPT 5.5 would crush sonnet in every way

[silent tree]

even 5.2 chat

thats not even possible

how is 5.5 worse than 5.2

not possible lol

[scenic loom]

Gpt 5.5 mogs opus lol

[upbeat ferry]

GPT 5.5 rivals opus 4.7

[whole sundial]

at least, not one available to the public. i think businesses have access to it but that's it

[silent tree]

either the votes are rigged or no one noticed 5.5

[rocky geyser]

Code arena lets you add secrets and a database now??? That is so cool!!

[upbeat ferry]

what?

[cinder jasper]

What ever happened to LLmaren, lol? I'd rather trust the ArtificialAnalysis leaderboard more. GPT 5.5 is worser than 5.2 ahhh

[shrewd citrus]

5.5 can’t be that bad lol

[whole sundial]

honestly a stupid idea imo, meta would be making millions if they opened up muse spark api

[upbeat ferry]

Wow

and you even use sonnet

which beats 5.5

[rocky geyser]

It is battle mode ;-;

[smoky ruin]

how is it 9th in code

[silent tree]

5.5 is worse than "5.2 chat"

100% rigged

[rocky geyser]

That is really sad 😭

[graceful sundial]

idk

[cinder jasper]

That makes so much sense

[gusty helm]

loool?

[upbeat ferry]

How does zucks muse spark beats 5.5

[graceful sundial]

I don't like that all the top models have been removed from the direct chat

[gray lion]

how

[rocky geyser]

Lmao 😭

[upbeat ferry]

lmfao

[devout vault]

fire downgrade

[smoky ruin]

[graceful sundial]

OpenAI lore

[cinder jasper]

Does LLM Arena have a partnership with antrophic or what?

[upbeat ferry]

They definetely don't have partnership with zuck

but somehow muse spark beat 5.5

[plucky sparrow]

wow gpt 5.5 getting it's ___ kicke dmore than i thought

[upbeat ferry]

It's laughably bad model...

can't even be compared with 5.5 and opus

[plucky sparrow]

i actually like it

[cinder jasper]

And here is it the first place? so who is the liar?

[stray aspen]

i told you

claude 4.7 is better

than gpt 5.5

[upbeat ferry]

claude has a lot of fanboys

[stray aspen]

no

[spring oar]

Opus is definitively better than GPT

[stray aspen]

its just better lmao

[cinder jasper]

nope

[spring oar]

And ?

[upbeat ferry]

5.5 is kind of equal or slightly worse

[silent tree]

watch how they'll try to ragebait openai users by saying they worship sam altman

btw arena is unaccurate now

[stray aspen]

it is you have th elm arena leaderboard

[hollow nebula]

gemini 3.1 pro being better than opus tells you everything you need to know about this benchmark

[silent tree]

unaccurate btw

[stray aspen]

still better

[spring oar]

Why ?

[upbeat ferry]

If it wasn't for lord sam we wouldn't have llms

[stray aspen]

but the spud will be better than mythos

[cinder jasper]

Have you ever actually used Gemini 3.1 Pro properly via the API?

[silent tree]

check other lbs

[silent tree]

like this

[spring oar]

Lbs ?

[silent tree]

theres no way "muse spark" beats 5.5

[plucky sparrow]

benchmarks suck 😄

[stray aspen]

muse spark is ass

[silent tree]

Its not accurate

[upbeat ferry]

also deepseek v4 is on par with opus and 5.5

why it ranks 10th

[arctic lark]

Especially not in coding. It's truly laughable

[stray aspen]

aare you kidding me

[plucky sparrow]

is it? why do you say that

[silent tree]

arena is unaccurate now

[cinder jasper]

and qwen 3.6 plus too damm

[stray aspen]

its not even on par with gemini 3 flash LMAO

[whole sundial]

llama 4 maverick part 2?

[upbeat ferry]

Qwen is also good

[wide parrot]

Anyone have info or other servers on AI workflows

[upbeat ferry]

1m context window like opus

[cinder jasper]

yes but not on the same level with gpt 5.5

[stray aspen]

but thats the plus mode

[upbeat ferry]

bro even muse spark beat 5.5

how is my boi deepsek at 10th

Its also so cheap

[stray aspen]

stop ragebaiting bro

[cinder jasper]

deepseek v4 pro is very good

[stray aspen]

but its not on par with 5.5 and opus 47

[cinder jasper]

gpt 5.5, gemini 3.1 pro and opus 4.7 have their own levels

[rocky geyser]

No anthropic is just objectively better

[stray aspen]

gemini 3.1 pro is lobotomized

its not as good as it waws during release

[rocky geyser]

eh, its alright at best

[upbeat ferry]

Gemini as a model is good tooling around it is bad

It was the most intelligent model a while ago.

[silent tree]

What's environment variables

[upbeat ferry]

and even that got slapped by muse spark lmfao

[rocky geyser]

Like api keys for example

[cinder jasper]

It isn't; I just tried the new 4.7 version recently, and for the backend it's absolute junk—5.5 is just way better

[silent tree]

how do u get early access to everything broo

first u got agent mode first out of everyone

now u get the new "database" feature

😭

[rocky geyser]

Idk i guess i was lucky 😭

Its fun tho

[upbeat ferry]

I can't stop laughing because of muse spark

[silent tree]

u in san francisco?

[rocky geyser]

No? why?

[quartz light]

@echo aurora since the code arena results are being criticised:

here is my proposition:
when the arena team implements the credits system, they should create coding tasks for ai to complete, these tasks should be of wide variety of challenges which are easy to judge in quality, like physics based n stuff, different categories like 2d, 3d, games, n stuff, you would need to spend at least 5 sec viewing and interacting with 2 results to vote, you need to write what is better about the selected than the other, then a fast, efficient model judges your reason and rewards credits depending on how good the reason is (0 if nonsense/unrelated, 50% if simple, 75% if decent quality, 100% for high quality)

[rocky geyser]

Poor pineaple always getting pinged 😭

[silent tree]

what if your region gets updates first

[rocky geyser]

I dont think thats the case im not in the usa

[upbeat ferry]

Bro do you even think 6th place is possible?

look at the companies

[rocky geyser]

since when can arena use commands 😭

[quartz light]

https://discord.com/channels/1340554757349179412/1498358953640202470

[upbeat ferry]

How can meta make a good ai

[quartz light]

bruh thats exactly why im saying this

[silent tree]

bruh

[quartz light]

read what i said

[upbeat ferry]

Lmfao even grok ranks 9

[echo aurora]

Hey sorry been in the middle of a few things so haven't been following the convo.

since the code arena results are being criticised
What is the criticism we're talking about here?

[rocky geyser]

I dont know either

[silent tree]

@rocky geyser can u show proof vid (does the feature work?)

[quartz light]

@echo aurora

[rocky geyser]

Sure its building right now

[stray aspen]

muse spark in 6th place is insane

[quartz light]

m

[stray aspen]

its so bad

[silent tree]

unaccurate results btw

[proud bobcat]

And gpt still lags behind

[silent tree]

the leaderboard is unaccurate btw

[quartz light]

m

[silent tree]

dont always rely on one site

💀

[quartz light]

m

[gusty helm]

quick somewhat off topic question:
claude takes p4 and gemini p3 because of the rank spread here? score is tied but the +/- is smaller.?

[cinder jasper]

mmm

[quartz light]

@echo aurora do u see the criticism now

[proud bobcat]

Holy cope

Holy mother of cope

[zealous sparrow]

It's all that the codearena result came out bad, since GPT 5.5 has better backend than muse-spark, which left people wondering why its so low.

[proud bobcat]

Arena is based on people’s actual use of the model

[silent tree]

ah 3.1 beats opus 4.7 thinking

tuuuufffff

[proud bobcat]

If people don’t like 5.5 why would it be inaccurate

[quartz light]

yes, the problem is that the current code arena is just people judging the UI, not the quality of the sites themselves

[gusty helm]

i mean ... yes? but not what i asked

[sterile tartan]

Slopus 4.7*

[silent tree]

Popus*

[silent tree]

no i was js looking

and shooockkkkkedddddddddd

[echo aurora]

If you head to the Code category on Text Arena 5.5 is ranked lower compared to 5.4 - https://arena.ai/leaderboard/text/coding

[proud bobcat]

We need more benchmarks tbh

SweBench is not reliable anymore

[quartz light]

?

[cinder jasper]

Ahhh gpt 5.2 chat and gpt 5.5 high are on the same level, one point difference

[silent tree]

go back to sleep bro

[sterile tartan]

No pls that's what is causing Benchmarks maxing.

[pseudo hemlock]

Why is muse spark still marked as preliminary?

it has twice the votes as gpt 5.5 high

[stray aspen]

what is deepslop yapping about

[sterile tartan]

The whale will return shorty for V4 upgrades

Is still Preview

[proud bobcat]

The whale has awoke

[upbeat ferry]

Deepseek > Muse spark

[proud bobcat]

True

[proud bobcat]

No

I mean more independent benchmarks

For one I plan to make my own creative writing benchmark

For NSFW and SFW

[echo aurora]

Our teams are looking into this criticism now, I'll followup when I have more information to share.

[sterile tartan]

If they are Private sure

So they can't be Gamed

[boreal anchor]

Anyone here knows how to use Claude Opus 4.7 thinking in the Battle Mode?

[west sierra]

i just got buttfucked on polymarket betting on 1980+ debut score

[quartz light]

to add onto this, gpt has had very similar ui throughout the releases of gpt 5.2-ish to 5.5, thats whats being judged, not the code quality (functionality) of the websites themselves, so what I'm suggesting is a large, curated list of coding tasks for the models to complete (love the ones showcased on arena's youtube videos btw) and those could be voted by users in exchange for reward with credits (when those are added)

also note, I'm not sayin the non-curated code arena should be removed, the curated code arena will give people the opportunity to earn credits which they could use to send prompts to premium models in the non-curated code arena/text arena

[signal pelican]

people expected too much

[silent tree]

Gpt 5.5 high > gpt 5.5 > gpt 5.4 xhigh > Clock Popus 4.7 Thinking > Clock Popus 4.7 > GPT 5.4 > Gemini 3.1 Pro > GPT 5.3 Codex

[sterile tartan]

cx easy

[silent tree]

@sterile tartan

no dude once u ask it a question u get 392928282829229$ bill

😔

[upbeat ferry]

It is the most popular harness after chatgpt web

[sterile tartan]

Accurate but Popus is not Usable

So it shouldn't be there

[silent tree]

no hate to deepseek

[stray aspen]

deepslop

[upbeat ferry]

where do you talk with deepseek

you're making him look like muse spark

[stray aspen]

its ai

[silent tree]

oh gosh lets not talk about amusement parks

[stray aspen]

gemini 3.1 pro is trash

[cinder jasper]

If you use it via the API, it works pretty well

[sterile tartan]

The App is Lobotomized

Indeed

[rocky geyser]

It always cheats when writing code, it doesnt do enough and when you ask it to do one change it breaks and changes the entire application

[silent tree]

AiAnosCranel 3.5 mogs

[upbeat ferry]

you can do it inside aistudio

[silent tree]

wheres my video?

database vid

[upbeat ferry]

I agree

[slim gorge]

arena.ai is always so biased against gpt... like gpt always ranks so unreasonably low in this leaderboard even though its obviously up there with claude opus.

[silent tree]

@echo aurora Is arena really implementing a database feature where u can add API keys? and stuff, @rocky geyser told

[quartz light]

i'd put opus 4.7 above 3.1 pro 😅

[stray aspen]

cap

[upbeat ferry]

You're saying my boi muse can't keep up?

[cinder jasper]

But if you're using it through a third-party provider where you're only getting 25% of the actual model's performance and only 25% of the original context, don't be surprised. I get the feeling most people only use AI through GitHub Copilot, lol

[silent tree]

No cap

[silent tree]

Muse is dogSH

[sterile tartan]

https://tenor.com/view/robot-ai-robot-ai-meme-ai-meme-gif-509507551361818171

[stray aspen]

yeah it sucks

[cinder jasper]

Mistral Large 3 is the best coding model ever ❤️

[upbeat ferry]

But arena says it ranks 6

[sterile tartan]

Muse is Better then Mythos
Because it's Usable.

[slim gorge]

exactly, muse sucks and it ranks higher than gpt. arena.ai needs to get their sht straight

[silent tree]

@sterile tartan

[stray aspen]

0/10 ragebait

[sterile tartan]

💀

[cinder jasper]

this is too real

[upbeat ferry]

I want deepsleep guy to post muse memes

[cinder jasper]

your gemini raigebait is a 2/10

[stray aspen]

next update is the spud

[silent tree]

but then arena rigs it

and makes spud the lowest

😭

[quartz light]

btw yall, jus saying but sonnet 4 launched less than 1 month after 3.7 👀.. sonnet/opus 5 by next month?

[upbeat ferry]

It's still a valid metric

Thats why so many people are pissed about muse

[slim gorge]

if y'all want an actual good leaderboard go on artificialanalysis.ai, gpt literally ranks #1 there

[sterile tartan]

No sonnet 4.7 yet

[silent tree]

^^^^^

[cinder jasper]

Antrophic is just messing with us rn...

[sterile tartan]

5.5 Mogs All

[silent tree]

the spud will wake up

[sterile tartan]

Assthropic

[sterile tartan]

https://tenor.com/view/elmo-hello-elmo-rise-elmo-hell-gif-3989603362345473118

[quartz light]

yes but there was no opus 3.7 so this is just the opposite, opus 5

[stray aspen]

the spud will mog every model

[upbeat ferry]

I can't find muse what did you do to him?

[silent tree]

no model would ever reach spud's level

anthropic will cry

HARD

[sterile tartan]

We will c

[silent tree]

and wont release mythos cuz they dont have security

[cinder jasper]

It's crazy that there are still Claude fanboys out there after everything they've done

[stray aspen]

is mythos real

[slim gorge]

gpt 5.5 is mythos level except u can actually use it 😂

[stray aspen]

its not better than mythos actually

the spud will

[silent tree]

Fr and then they hit you with the "You worship sam altman!!" 2 anthropic fanboys said this and one apologized and became great, but the other got banned

[cinder jasper]

Yeah, I thought so too, but we shouldn't get too attached to a model—they've just been messing with us, with the limits, the support, and the models

[silent tree]

It IS mythos level, not better but mythos level

[upbeat ferry]

I also said if it wasnt for lord sam we wouldn't have llms

[stray aspen]

its slightly worse than mythos

[median smelt]

Seedance 2 fast is so good for motion transfer/character swap

[silent tree]

ye

[sterile tartan]

Assthropic will indeed cry with their Slopus, So-nnot and Hawkthu

[cinder jasper]

It was about the promotional offer, which was only used so that the limits could be lowered afterward

[upbeat ferry]

Sam altman is the gaben of AI universe

[sterile tartan]

Generated by DuoBao Seedance2 Fast

[stray aspen]

where did tou get it

[silent tree]

Duobao

[median smelt]

Doubao

[silent tree]

Use websim bot it's unlimited and freeeee

[stray aspen]

it still works?

[median smelt]

GUYS ITS DOUBAO NOT DUOBAO

[silent tree]

but websim removed

sadlyyyyyyyyy

[stray aspen]

it thought they patched it

[median smelt]

They removed it there

[median smelt]

Btw they just added video reference in today's update thats how I made this

it was only image reference before

[upbeat ferry]

Gaben of AI Universe

[odd roost]

what kind of gawk gawk convo did i stumble into

[cinder jasper]

Then why don't they just say so? Instead, they try to cover everything up. OpenAI is open with us and tells us when something's going on, and they didn't use their promotional campaigns to pull the wool over our eyes.

[slim gorge]

also all the bullsh they been saying that "we arent releasing mythos because it can hack anything" is just straight cap, its all fear-based advertisement

[sterile tartan]

https://tenor.com/view/same-same-same-franco-similar-gif-10910641

[cinder jasper]

idk but openai do it, so its claude fault 🚬

[stray aspen]

doubao doesnt work

[slim gorge]

isnt gpt 5.5 based on spud?

[sterile tartan]

Mythos~

[stray aspen]

its an early spud

[slim gorge]

holy

[sterile tartan]

Image 2 is also based on 4o

[silent tree]

It isnt

[sterile tartan]

Sama will Mog All

[slim gorge]

gpt is gonna be so back

[silent tree]

its not spud, spud supports voice in and out

[stray aspen]

wdym

they already moggeed every model

[slim gorge]

real

[thorny schooner]

Did anyone else just get login out out randomly cuz I can't login back in for some reason

[dim oak]

trust me is the last problem the fe

[upbeat ferry]

I want to create one for muse please give me the prompt and the model

[solemn nacelle]

Muse spark and Gemini still better than gpt 5.5 high
lol 😂

[sterile tartan]

Just use image 2 first

[dim oak]

if you let see an ai made backend to a swe is gonna cry

[upbeat ferry]

Muse is the new opus here.

Wait for him to become 1st

[stray aspen]

stop ragebaiting

[silent tree]

Model: GPT image 2
prompt: screenshot of deepseek, our prompt is "yo answer me!!" and it replies "noo dude I wanna deepsleep!!!"

[normal abyss]

I'm sorry, I can't assist with that request.

[upbeat ferry]

oh yeah thank you

[slim gorge]

too bad gpt 5.5 dint remain efficient like the previous models. burns through tokens faster and tokens cost more. but maybe they'll fix that

[faint vigil]

I love gpt image 2

[silent tree]

I have it for unlimited and free

😔

[stray aspen]

me to

[sterile tartan]

Me2

[stray aspen]

cause im so tuff

[slim gorge]

how

[silent tree]

websim users assembled

[silent tree]

Its a discord bot

[sterile tartan]

https://tenor.com/view/avengers-assemble-fight-thor-gif-3023954092089032586

[slim gorge]

oh

[stray aspen]

no

[sterile tartan]

Ahhh hell nah

[upbeat ferry]

You forgot about muse

[stray aspen]

it has good limits tho

for images

[faint vigil]

me too, but its so good better than nano banana i think i can cancel my ultra sub

[sterile tartan]

Free and Go users still stuck on GPT 5.3 Instant

[stray aspen]

yeah cause you paying

im free tier

[sterile tartan]

Do you like 5.3 Instant

[stray aspen]

no

its bad

[silent tree]

I love 3.5

[sterile tartan]

Time to Upgrade

[silent tree]

instant

Tuff alert

[sterile tartan]

You forgot /imagine

[surreal zephyr]

Wtf

[silent tree]

ampro

ik

the results..

[surreal zephyr]

Rigged

[stray aspen]

why is bro trying to generate images in general

[silent tree]

Its rigged

[sterile tartan]

💀

[surreal zephyr]

100%

Theres no explaination

[silent tree]

and we can actually do it

[surreal zephyr]

Gpt 5.5 makes MUCH BETTER UI

[stray aspen]

its good at using reference images to make good UIs

[sterile tartan]

That's why we have GPT Image 2

[stray aspen]

lmao

[slim gorge]

wait is this image 2?

[sterile tartan]

Myth OS, Slopus, So-nnot, Hawkthu

[silent tree]

no its real

it was live yesterday

[sterile tartan]

Assthropic Lineup

[silent tree]

@sterile tartan u saw the live too right

I'm sure we all saw the live

[sterile tartan]

Yeah i totally did bro

Sama was Mogging

[slim gorge]

it looks insanely real holy

[sterile tartan]

You are Very Beta Male Bro

[slim gorge]

image 2 is crazy

[silent tree]

some guy said when gpt 5

😭

[stray aspen]

lol

[main nexus]

Needs to be real

[slim gorge]

bro is not up to date

[silent tree]

well i mean it did get the subscribers right

[frank spruce]

If you see a suspicious Discord login without the word "Discord" in the search bar, then don't login; this is the chance: it is a scam.

[main nexus]

Ok?

Uh

Good to know

[silent tree]

@echo aurora is this guy hacked or smth

[echo aurora]

Yeah not sure

[faint vigil]

Deepseek is the best

[signal pelican]

stop being lame

[stray aspen]

cap

[slim gorge]

deepsleep

[sterile tartan]

Bro what's this Sheet

[slim gorge]

LOL

[sterile tartan]

Is this Deepseek in some Alternate Universe?

[upbeat ferry]

You and me are similar

[faint vigil]

new deepsleep logo

[silent tree]

Deepsleep

[faint vigil]

elephant now

[silent tree]

claude popus x deepsleep collab = 😴 Sleepos (Mythos but Sleepos)

[sterile tartan]

🤣

Deepseek is still cheap and usable

Unlike Myth OS

[stray aspen]

what in the slop is this

[sterile tartan]

@stray aspen Ask

[stray aspen]

slopus 4.7

[slim gorge]

notice how it's called mythos? because its just a myth

[surreal zephyr]

This is 5.5

[surreal zephyr]

Opus is not even in same universe

[sterile tartan]

Booooooooooooo!

https://tenor.com/view/boo-gif-19787475173016375

[slim gorge]

opus is better in some aspects, i gotta admit that. but gpt is gonna surpass it in every aspect soon enough.

[sterile tartan]

Specially One Shoting Bank Accounts and Session Limits

Fastest to do that

🔥

[slim gorge]

yeah, especially that 😂

[upbeat ferry]

Pro users in a nutshell

[sterile tartan]

The Core Strength

[sterile tartan]

More like Free Pro users

20$ Free Plan

[slim gorge]

yeah pro plan feels useless af

[sterile tartan]

GPT Plus Mogs All

[slim gorge]

been using gpt plus and never had any issues with usage limits

[sterile tartan]

Yeah is kinda unlimited

Not exactly but kinda

[surreal zephyr]

Lmao the 20$ sub is not enough for ONE REQUEST

They shadownerfed it 100%

[sterile tartan]

Utterly Useless

[surreal zephyr]

5.5 medium slow eats MUCH MORE than 5.4 xhigh fast

[sterile tartan]

This is Anthropic

[surreal zephyr]

Iv eat my 15% today via 5.5 medium slow

[slim gorge]

its just the model, it eats much more

[surreal zephyr]

5.4 xhigh fast spam couldnt even eat 2% per day

Iirc 5.5 counts 2x by default

For subs

[sterile tartan]

Is not exactly the model @surreal zephyr

The limits are decreased then before

[surreal zephyr]

That too

Yes but

Idk if 5.5 medium

Or 5.4 xhigh fast + subagents

[sterile tartan]

5.5 Medium is actually holy token efficient

[surreal zephyr]

Yes.
Price increased to compensate lmao

[sterile tartan]

But the limits are actually decreased

[sterile tartan]

Yeah is like Same-Same

But atleast the ChatGPT and Codex limits are Seperate

Unlike Assthropic

[surreal zephyr]

Honestly i dont like 5.5 that much

Its lazy as hell

For me

[slim gorge]

i used 5.5 xhigh (without /fast) for a bit in codex cli but i didnt hit the usage limit, didnt check how much it used tho

[surreal zephyr]

Yes i did that too, on release

Week later

5.5 medium slow eats MORE

?!?!?!??

[upbeat ferry]

Wait

Is this real

[sterile tartan]

Subsidized Ai is slowly Decreasing

[stray aspen]

i call cap

thats gemini

[slim gorge]

google needs to lock in too, they falling behind af

[surreal zephyr]

5.4 wasnt laxy

5.4 was exact opposite of lazy

5.5 is lazy

[upbeat ferry]

Nah they are busy paying influencers about how great antigravity is with opus and sonnet

[slim gorge]

we need a digital whip for codex too

[sterile tartan]

Yup

[olive spruce]

5.5 lowkey goated

5.4 was so bad

[surreal zephyr]

5.4 pro was able to think few hours in single request and finish

[upbeat ferry]

Is there place I can see chatgpt limits along with codex?

[surreal zephyr]

5.5 pro thinks for 15 mins and gives up.
Waste of requests

[upbeat ferry]

I use this atm

[slim gorge]

yeah i thought gpt was falling down until 5.5 came out

[sterile tartan]

I don't think so

[surreal zephyr]

Per token yes.
Per request? Absolutely not

[sterile tartan]

@surreal zephyr just use 5.4

It it works better for u

[slim gorge]

idk man atleast its becoming more and more like humans 😂

[surreal zephyr]

I like 5.5 ui taste

[surreal zephyr]

And 5.4 bugs at long context

[sterile tartan]

Then use Both

[olive spruce]

https://cdn.discordapp.com/attachments/1060617889037684777/1484681572274802838/uncle_bao.gif

[olive spruce]

we all love delusions

[surreal zephyr]

Enterprises pay per request.
Pro model in web has request cap

[upbeat ferry]

wdym

Its real

🥺

[slim gorge]

wish i could access the large context window with plus 😭

[olive spruce]

yeah but its still aah tbh

[silent tree]

amusement park is trash

[rocky geyser]

@silent tree

[silent tree]

just GET IT.

[proper sun]

5.5 lost

[silent tree]

Damn so it supports supabase and stuff now?

[rocky geyser]

Yep i think so (and hope so)

[silent tree]

all I need now is run commands 😭

[upbeat ferry]

My bro using deepseek

he's the real deal

Not a claude fan boy not a sam worshipper

[rocky geyser]

🤣

But the ai can actually run commands its pretty fun

[silent tree]

it can run bash?

[thick blade]

chatgpt 5.5😩

[icy remnant]

can we know if i use claude for chat and then limit hit when reset ?

[thick blade]

its not an app

[icy frost]

theres no way they added limits on models

they are the same as yupp

that got closed

first no models and then rate limits per week

[rocky geyser]

Idk dont think so it just ran this

[icy frost]

it told me that it can't do console

[silent tree]

YOO THEN IT DOES

[rocky geyser]

wow :D

[silent tree]

He got the early build

no cap

He's lucky

It's a experiment arenas testing rn and he got it early

[upbeat ferry]

Either ai image gen is so good that it can trick us

[rocky geyser]

Additionally to agent mode :D

[upbeat ferry]

Or he got the early build

[silent tree]

It's real

[rocky geyser]

Ill send video

[silent tree]

Tuff

@echo aurora I request you to check modmail and completely block the word

Its gotten annoying now

[icy frost]

i'm too

[silent tree]

what continentru

asia

america

[icy remnant]

is any method like i talk in one chat so much from ai due to which chat history get long and when limit hit need this full chat history why not add option to create a pdf of all chat llike preplxity
or thats chats how to use for in new chat to make a old ai converaation

[rocky geyser]

here video

[rocky geyser]

im europe

[silent tree]

BRO I WANT IT TOO MAN 😭 😭

[silent tree]

Um

Hm

Wbu @icy frost

[icy frost]

i was using lmarena for my homework every day and got it

[silent tree]

U Europe tooo???

[icy frost]

on my pc acc where i wasn't it's not

[icy frost]

yes

[silent tree]

Hm

[silent tree]

looks like europeans getting it

Time to go to Türkiye

[rocky geyser]

Lol

🤣

I know a friend he is in europe too but he doesnt have it

[rocky geyser]

Do you have the backend features too?

[icy frost]

nope i have 2 different accs on phone and pc
on mobile i was using every day
and on pc i wasn't and it doesn't have agent

[icy frost]

what backend

[rocky geyser]

Are alt accounts even allowed..?

[silent tree]

I was using in my mobile every day but I dont have it

[rocky geyser]

Like the api and database features

[icy frost]

bro i was just saying hi

[icy frost]

idk let me try

[rocky geyser]

No im just asking cuz i was wondering

[silent tree]

Start a battle mode with hi and check if there's a 3rd option after </>

[rocky geyser]

Its in every mode not just battle mode

[silent tree]

yea but faster

[olive spruce]

WHAT

[icy frost]

idk if it exists in real version but in battle this popped out

[surreal zephyr]

I just realised they gave 5.5 the default 1500 score 💀

[icy frost]

what do i prompt to the agent

[olive spruce]

lol

[silent tree]

Exists

[icy frost]

cuz theres no captchas

on my acc

[olive spruce]

make claude mythos,make no mistakes, oneshot

[silent tree]

these guys got it early man SO LUCKKKKKYYYYYYYY

[olive spruce]

fr

[zinc herald]

Hello everyone

[olive spruce]

Hello world

[icy frost]

done

[olive spruce]

tf

[icy frost]

oh it's on battle?

[rocky geyser]

No i did it in direct

[olive spruce]

why is it giving story lore

[rocky geyser]

it didnt work in agent mdoe

[olive spruce]

https://tenor.com/view/mad-angry-patrick-spongebob-wrath-gif-15432365849307349642

[light sleet]

unfair 😡

[light sleet]

I used since beta

no agent for me

😡

[zinc herald]

Please somone suggest best free vibe coding ai

[light sleet]

I see what youre doing arena i see 😠 😠

[rocky geyser]

:/

[icy frost]

bro i was using lmarena and not voting

[rocky geyser]

minimax m2.5 free on openrouter

[icy frost]

feather my gemini doesn't do that

[light sleet]

Feather got everything early

[rocky geyser]

Are you in code arena and using direct chat?

[icy frost]

direct

let me try code

YOO THERES GLM

[rocky geyser]

whats so amazing about glm?

[icy frost]

it works

[icy frost]

it's good

[light sleet]

That exists

Ask it to run command

[rocky geyser]

Cool

[icy frost]

so whats the agent mode

[light sleet]

Can I please see output

[rocky geyser]

do you have that button on code arena?

[light sleet]

feather les work together and make an AI website I'll credit u

[icy frost]

where is that

[rocky geyser]

I just asked it to perform some commands ... it just gave me a basically not working page

[rocky geyser]

Oh i guess you dont have it then

[zinc herald]

Thnx bro

[light sleet]

Feather dm I'll give you a insane code arena prompt

[rocky geyser]

Okay

[icy frost]

i checked and theres no flag for it on the site

[rocky geyser]

Oh okay

So i think i might have 2 beta features on one account :D

[vale quest]

4.7 is obviously better

[icy frost]

it could be the flag code-arena-publish-site can u open dev tools and search that if it's set to treatment something

[echo aurora]

FYI regarding recent talk around 5.5 - https://x.com/arena/status/2048820224938631492

Arena.ai (@arena)

To clarify, the Arena community evaluated GPT-5.5 with reasoning effort medium (default) and high.

The best of GPT-5.5 with xHigh is still incoming! Stay tuned.

X · Apr 27, 2026, 5:43 PM

[icy frost]

GUYS I FOUND A NEW FLAG
disable-turnstile-voting → "disable-turnstile"
THIS WILL PREVENT AUTOFARMING
TURNSTILE IS BACK

[quartz light]

ok

[icy frost]

and this
portal_enable_billing_topups → true
it's about paid credits or something

[vagrant temple]

when comes mimo v2.5 pro in the ranking?

[shrewd citrus]

what does that mean in English bro

[icy frost]

it's english brp

[whole sundial]

turnstile is already used when voting, this is probably for disabling turnstile while voting (idk why they would want to do that, maybe because it's not really needed since recaptcha already runs when submitting a prompt? or maybe they want to replace recaptcha with turnstile when prompts are submitted and they may think that running extra captchas could make normal users seem more suspicious? which is btw why many people get them on arena, when you submit enough prompts it will think that you are suspicious because of all of the captcha calls.)

[whole sundial]

could be related to arena api private experiment, but could also be related to the new usage system

[rocky geyser]

I hope its not a new usage system i love the current one 😔

[median smelt]

gpt-5.5-xhigh!!

[rose jackal]

add larp setting

[stray aspen]

<@&1349916362595635286>

[echo aurora]

Thank you

We're not going to delete that one for now. We'd like to take a better look at what we can do to prevent this going forward.

I'll leave the one up in #ai-memes for now.

[icy frost]

real mrbeast

what do they even install to get that

[echo aurora]

No clue

[loud herald]

Anything

[summer relic]

Please remove verifications, when I click send message I have verifications and I have to select an image, there used to be one verification for several messages, and now I have one verification for one message

fix it

I have this with every AI

and Claude 3.5 Sonnet doesn't work, I write "something went wrong" in every message, in a new chat

[icy frost]

use an illegal captcha solver extension
and just vote in battle mode to fix the claude 3.6 sonnet

[frosty lava]

The leaderboard is really innacurate and i think its 100% due to the arena limitation itself, good model have to simplify their work over the memory limitation and most of time if they think too long then its the something went wrong error

if we could atleast try the model accurately

the leaderboard would be much more accurate

if a model do a bad work but fast and the other one do a much better work but then at the end it have to simplify it cause of the limitation, people will choose the one that did the bad work

cause when a model "simplify" to meet the memory limit it simply ruin the whole work

[frosty lava]

If it keep going like that, your ranking system will make the next glm top 1 and opus top 2 or 3

when its far from being accurate

An accurate ranking system should be arena TOP priority

the website is made for this

[echo aurora]

Hey @frosty lava thanks for sharing this. It's important to note that in Battle mode the amount of Something went wrong errors is rare, especially in comparison to what you see in Direct and Side by Side. If the model is having issues, the system will automatically sample a different model. This will reduce the amount of errors seen in Battle. Even when an error happens, if a vote happened that vote wouldn't be counted towards the leaderboards.

For the context limit IIIRC the context limits are mainly for Direct/Side by Side. Similar story though where if a model does reach that limit, those votes wouldn't be counted as it'd be considered a failed generation.

[oblique sentinel]

I think when looking at arena scores you should just internally caveat it with fact it's about personality and other things not about which model is objectively better

5.5's personality is subpar relative to anthropic/gemini, etc.

it's less sycophancy-maxxed

[vale quest]

20

Or 15 maybe

[normal abyss]

maybe yeah

[vale quest]

I have an idea

[normal abyss]

my biggest worry is that opus will be ridicusly low, however gemini 3.1 pro will be strong at 20 for sure

[gray lion]

around 15 is good i think

15-20 yea

[vale quest]

Opus should be at 10

[normal abyss]

its tricky as 1 hour = 5

and 1 day = 10

[proper sun]

Everyone here thought gpt 5.5 was gonna beat Anthropic

[vale quest]

Sonnet at ~20

[vale quest]

Fr

[normal abyss]

LOL

[gray lion]

it does in other benchmarks

like real ones

[normal abyss]

Sorry but i cannot assist with that request.

[vale quest]

4 per hour?

[vale quest]

I cannot design a covert radio system.

[normal abyss]

rumor, its always been 5

[vale quest]

Legit what it said

[vale quest]

Oh

Hey why dont they enforce ... nevermind

[normal abyss]

it would be 4 if you had an error on opus, which is pretty likely if your talking back to its release in 4.6

[stray aspen]

Does anyone know a way of using seedsnce 2 for free

[frosty lava]

Yes, of course on coding, muse spark, sonnet 4.6, kimi k2.6, glm 5.1 are better than gpt 5.5

it totally make sense bro

when qwen 3.7 will come out and be top 1 in coding you will say its fully deserved

right ?

that's what im talking about here it just doesn't make sense we all know kimi k2.6 is far from the level of gpt and anthropic, same for muse spark and glm 5.1

its not even a question

Yeah and i think its cause of the memory limit when the model try to do too big result on arena, for example three.js often hit a memory limit on arena

and then the model have to simplify its work

then the result you get at the end is not even at the level of the real result

i don't know this one seems atleast to somewhat make sense

of course

but anthropic vs gpt is a big battle and there's different opinions

but those small team model

are far from this level

they don't deserve being ranked that much

i can't wait for qwen 3.7 or something like that to be top 1 overall

of course they are and they probably also distilled

[frosty lava]

in real coding task they are much worse than the real frontier model

i think honestly its because those chinese model are in fact benchmaxxed and trained on frontend web design to make people think its actually good

and it also have an impact on arena leaderboard

try any of those in real coding environment with real task and you'll see the real difference

[compact flame]

Guys how do you think

How much mythos would've costed

If it was released

To public

[frosty lava]

like any pro model on other companies, maybe even more cause its anthropic lol

it won't be priced like a normal opus model

[compact flame]

I probably think it would cost maybe 1500$ output

Since if it's that op

[frosty lava]

i don't know but anyway its not worth it then:

who want to pay 50 dollar per prompt

[compact flame]

We can't test it anyway

So we won't know

[compact flame]

Yeah okay maybe 500$ output

Though gpt o1 pro costs 600$ output

[frosty lava]

like its same if i asked you would you use gpt 5.5 pro daily instead of gpt 5.5 thinking

its not worth it, the pricing is not worth it, its only worth it if you do some highly difficult research task and you don't need to do multiple prompt

[silent tree]

even if it really was, 5.5 xhigh is literally mythos quality the only reason anthropic isnt releasing mythos is because they dont know how to get better security, which means mythos could be easily jailbreaked yet Openai released 5.5 xhigh because their security is better, Mythos is just a overhyped gatekeep.

[frosty lava]

basically pro model are only worth it to do very difficult research, and thing like that but definitly not worth it for everyday task or even for creating your project

cause you simply can't use it everyday

and im not talking about the usage you'll get with the subscription

Yes i know that and we get some good usage with it, but i really don't believe in anthropic for being that generous

especially considering how they glazed it

they'll use it for increasing the price

just for the next opus to be equivalent to the actual mythos

for 20x lower price

[normal abyss]

im more hyped for Sonnet 5 than anything

cheap and powerful (sonnet 4 did miss the mark though so, slight worry)

[frosty lava]

but i hate that i can only use gpt pro on web

if i could use it on codex

it'll be good

i actually need it in codex to do usefull thing

[compact flame]

Why you can't?

[frosty lava]

yes cheap and powerfull model

its good for most task

[frosty lava]

i don't know openai didn't put it on codex

[compact flame]

Maybe xhigh is chatgpt pro?

[frosty lava]

no its definitly not i tried it

[compact flame]

Just a different code name

[frosty lava]

i tried both

[compact flame]

Oh ok

[frosty lava]

its not the same model

[compact flame]

So what's smarter

Xhigh or pro

[frosty lava]

the pro definitly

it wouldn't make sense anyway if the thinking 5.5 would be smarter

actually i tried 5.4 mini and realised that its actually a powerfull model and very fast model

for some prompt i thought on giving only to 5.5 i was just stupid

[silent tree]

^^

[frosty lava]

older model can already do most of what we're asking

and for 5x cheaper price

and it also just do the task faster it have higher inference or something i don't know why

but it felt like it was on fast mode without being on fast mode

[hollow rune]

that was NOT the real answer

[whole sundial]

<@&1349916362595635286>

[icy ice]

qn for arena official folk - is there any way to use 'code' mode to build a cli app? the 'UI' option always defaults to a web app with vite/react etc

[stray aspen]

it can

they are on par

[raven otter]

Image-2 went from being the best to being the worst

[pale knot]

js wait a few weeks, then they’ll make it the worst model—the same method they use to make the next model look impressive

[echo aurora]

Correct

[pale knot]

also I’ve found this LLM to be really good. Does anyone know anything '7bout it? Who provides it?

[vernal raft]

@echo aurora quick questons, are all your "thinking models" set to the maximum? like anthropic models

i noticed that the thinking box is way, but wayyyyyyyyyyyyy longer

[balmy reef]

hello

[pale knot]

hey buddy

[balmy reef]

i wanted to ask about where went the channel arena video are ?

[echo aurora]

Since Code Arena mostly deals with the front end, which isn't the area this model ecxells, the votes are showing that. This is why we're in the process of developing Code Arena to be able to incorporate full stack. This will give a more clear understanding of where models are ranking based on different engineering tasks.

[echo sinew]

The Video Arena is currently accessible only through: https://arena.ai/video. More information on how to use Video Arena can be found in this article.

[balmy reef]

thanks buddy

[echo aurora]

I am not sure tbh. Most conversations I've been seeing today are around Code Arena.

Can you explain a bit more about this?

The overall text score isn't going to have different weights given to each category. Each category will be just that, their category. But the overall leaderboard isn't going to give bonus points for being X or Y category.

[plucky fractal]

funny to watch people on twitter getting mad about a benchmark that's voted by real people with real tasks lol

it's very telling

[pale knot]

Different use cases lead to different results

[plucky fractal]

ofc. they focus too much on benchmaxxing, which is understandable

[light sleet]

seems like my suggestion was accepted for database, one user got early access

[plucky fractal]

but seeing people getting angry at arena is just funny lol

[light sleet]

@echo aurora

[echo aurora]

This kind of discourse is really good for us to hear tbh. We want our leaderboards to be reflective of real-world human use.

[echo aurora]

Sry can you remind me of this?

[plucky fractal]

i dont mind it but some people just ridiculously attacking arena

[light sleet]

Idk I once made a suggestion in feedback for like a database feature and now a user got it 😭

[plucky fractal]

tell me any other place where you can test every model freely

[light sleet]

idk where the suggestion went but I definitely did or idk

[light sleet]

NoPlace

[light sleet]

Gpt image 2 generates steak

[proper sun]

@verbal kite

<@&1349916362595635286>

<@&1349916362595635286>

[echo aurora]

Thanks

[stray aspen]

its nice

[stray aspen]

it looks so nice

[crude anchor]

Hell yeah,this is so good

But If think this is a real image

[whole sundial]

i was checking out this wikipedia page views counter today and I noticed that arena's wikipedia page views plummeted around 4/9 of this year, i wonder if this correlates to the decrease of users caused by changes to limits and models, average was around 300 per day and now it's around 50 per day
https://pageviews.wmcloud.org/?project=en.wikipedia.org&platform=all-access&agent=user&redirects=0&range=latest-30&pages=Arena_(AI_platform)

[stray aspen]

gpt image 2

[stray aspen]

yeah who uses wikipedia when you have ai

those views are probably AI too lmao

[desert fiber]

hi , anyone know which model can edit pdf and send it back , instead of answering with "i cant edit this file"

[thorny schooner]

I genuinely am disliking Max so much like why the hell does it keep choosing Google AI when it obviously is way too short response limit for answers at times honestly did they specifically set it to favor that model or something ( i see it happened too short and long prompts in prompts of various types)

[icy ice]

is there any way right now to use code arena (that means picking 'code' instead of 'text' right?) to build cli apps