general

[loud herald]

But it uses different providers other than the actual anthropic provider too

It uses hermes/openclaw and AWS and ofc natural anthropic

Depending on the one you pick it can be cheaper

hermes/openclaw provider is the cheapest reliable one

80.7% off from before

[hollow ivy]

how can i find it?

[south charm]

Gemini 3.5 flash has appeared under name: Gemini 3.0 flash

[wild isle]

Thanks, works

[echo hawk]

<@&1349916362595635286>

[stray aspen]

how did you do 20 seconds

[loud herald]

Once I’m on my pc I will send you it

[loud herald]

I will send the link when I’m on my pc

Probably not in here though because idk if it’s against the rules or not in this server

[desert fiber]

can i know which model the "MAX" option used ?

[wild isle]

Nope

Think it uses random one based on the query

[desert fiber]

okk

[wild isle]

https://tenor.com/view/rat-gif-13029792972945826049

[wild isle]

Thanks, it helps

[fluid ruin]

<@&1349916362595635286>

Don't think I didnt see that boi

AND THIS BRUH

Zhipu be giving the models psilocybin or what

[wild isle]

Nice

[carmine estuary]

Hi

[golden ocean]

https://tenor.com/view/cat-cat-bonk-bonk-cat-attack-white-cat-bonk-camera-gif-12579285308205623581

[wild isle]

https://tenor.com/view/canada-eraser-pencil-huge-big-eraser-gif-10033098

[restive crypt]

bro can i have it as well

[loud herald]

I can’t send dm btw or friend request lol

[tame cobalt]

Hey bro, the agent mode in LM Arena seems to have a daily limit of 20 uses. When will it become unlimited or have a higher quota? Also, it's a really useful project. I hope that in agent mode, the workflow, each conversation, and its files — such as installed skills — will be saved, so they can be carried over to new conversations, giving it longer memory

[foggy delta]

why LMarena is always down

[steep mauve]

holaaaaa

[sonic wigeon]

*arena

[patent bridge]

Bro

[tight meadow]

Made this with DesignArena

[ivory lily]

https://www.facebook.com/share/v/14c7CajKrQY/

[west cove]

how can generate a video from pic ?

[unreal dagger]

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

[amber hearth]

How do we use your image to video generator

@thorny current

[thorny current]

Mine?

[surreal zephyr]

No

[wild isle]

Is it local

[stone forum]

who missed Claude opus?

[slender thistle]

Didn't miss, it's just dumb

Ruined one of my apps and then I had to use Gpt 5.5

Dumb hype, I am so tired of these marketing crap

[compact flame]

Wdym. Ya got a bad opus In battle mode?

[slender thistle]

Dude I am using opus 4.7, it's Bad

Not in battle mode, already have premium

It didn't even fix a single thing for my project where gpt 5.5 was working like a software developer

[compact flame]

There's no opus

[slender thistle]

Oh my goodness, do you speak English?

[compact flame]

No I don't actually

I'm Russian

[slender thistle]

Okay good, I am using opus 4.7 on a different platform

[compact flame]

Ok

[slender thistle]

Not in LM

[compact flame]

Then I guess it's just a fake website like veri Veri fake

[slender thistle]

It's UpTo you

[compact flame]

Claims to be opus but might be chatgpt 3.5

[slender thistle]

Yeah you are right, go on

[compact flame]

So idk

Trusting anything is impossible these days

[slender thistle]

That's why I don't trust benchmarks, I have zero trust in Anithropic

And their Claude mythos is another marketing scheme. I am so tired I don't wanna use LLM for coding anymore

Millions of ways to fail, it's all hallucinates, can't build something from scratch. Just waste of time

I just found that gpt 5.5 is absolutely a beast for debugging only

Gpt 5.5 fixed 4 years of old projects where opus 4.7 and 4.6 cause an error again and again

Imagine if a model hallucinates on kernel level software, that's even worse than open source model

Because models like mimo v2.5 pro perform way more better than opus 4.7

[spring oar]

new gemini pro model soon ?

[carmine void]

<@&1349916362595635286>

[slender thistle]

again this scam

[obtuse summit]

why's the agent mode presnet on 'not signed' in window but not when i am logged in????

[hollow ivy]

just tell me, how i can find it with startpage (or Brave search)
-# (i had to deactivate DMs globally, as i lost an account in the past, due to Discord "shenanigans")

[shy isle]

what do you guys use agent mode for

[hollow ivy]

nothing
i rather use text mode, has much more flexibility

(you can use any prog.lang)

and it's easier to "pin down" models

and more models available

[brittle hull]

Sup guys, I'm new here, what's the conversation about

[hollow ivy]

delete your scam first, pls

yeah, just delete that scam messages first

this one

[brittle hull]

Is scamming a thing in 2026

[hollow ivy]

first order of business: get a safe OS

[brittle hull]

Stop using your email on uncensored websites

[hollow ivy]

i highly recommend Linux

especially Mint

macOS might also be ok

ohh very bad

then you need to format everything

and even get a new EFI-bios

because you could have got a hyper-visor rootkit

these ones are nasty

formatting doesn't kill them

this is the worst case, if it happened

but i dont know, if the worst case has happened to you

if you are lucky, you just got a trojan

trojans should be removable by formatting everything

but you need to get your system tested, thoroughly

there is a website: trojan board or so

where they write about it

oh, you should use AI to get help:

• Gemini (AI-studio)
• chatGPT
• Deepseek
  And maybe also GLM

Deepseek-V4-pro is almost frontier but free and has websearch

just google/startpage for deepseek chat

[wild isle]

Stay local

[hollow ivy]

chat.deepseek.com

aistudio.google.com

so, never install cracked games

-# ||… before testing them at virustotal.com||

it said, to always use https://virustotal.com to be safe

yw

if you ponder about installing a new OS, here is a candidate: https://linuxmint.com/

(worked perfectly for me, to get me finally disconnected from windows for good)

[light sleet]

W paws https://giphy.com/gifs/reaction-ishowspeed-w-speed-RUYCx0kgKJ6wiZ9LGs

[hollow ivy]

if you prefer to communicate in another channel, we can use the old LMsys server, the mods over there are extremely generous: /ZN9SUDze9x

(just combine that with ".gg" and you get there, oh and use the obvious name [for the actual website])

[light sleet]

It still exists?

what

no one told me

[hollow ivy]

interestingly, the mods allowed everyone to create new channels in their server

it seems, they don't use it as much anymore, so they essentially released it to the public (so we can use it as public forum, for anything)

[light sleet]

e

[wild isle]

https://tenor.com/view/cat-kitty-kitten-angry-mad-gif-16754054616247640902

[shy isle]

guys what should i use agent mode for

[tame cobalt]

Does this offer a free API service?It would be even more perfect.

[sullen creek]

bro

no

[tame cobalt]

You can download its APP.This mode can be used on the APP.

[shy isle]

oh ok but im juust wondering like what purposes or tasks i should do with it

because mostly im just doing school work design projects etc

[tame cobalt]

You can use it to make PPTs or documents.

[shy isle]

hmm ok thanks

[hollow ivy]

interesting poll: #leaderboards message

[south charm]

Gemini 3.5 flash test 1:

https://019e3977-4820-74e4-97d4-c1070321d035.arena.site/

[thorny current]

hey

[slender thistle]

Those who voted opus 4.7 are just fans, they didn't even test with real project

[south charm]

It's not free btw

[thorny current]

does some one knows who is nu takoe in this server

[slender thistle]

Everything is free for me though

[south charm]

How bro :/

[frail mulch]

Honestly Opus 4.7 is the best overall, since I work alot with websites

GPT-5.5-xhigh is also REALLY good for backend, but just sucks on frontend, Though the limits on it are really good so I tend to use it alot more, but that isn't because the model is better, but becaue the limits are greater.

[silent tree]

@surreal zephyr summoned..

WHAT DO U MEAN 5.5 XHIGH IS BAD IN FRONTEND

[slender thistle]

Work with real software engineering, you will see

[surreal zephyr]

LMAO

5.5 xhigh mogs mythos 5 high

[slender thistle]

I don't wanna go back to mythos even if Anithropic releases it, it's another slop

[south charm]

It not that bad

[silent tree]

look at this https://goldime.netlify.app/ @frail mulch

(generated by gpt 5.5 xhigh)
(credit: ampro)

[south charm]

Absolutely hard

[slender thistle]

The performance and one shot from Gpt 5.5 is real, I even fixed old Software which was incompatible for modern devices. Claude failed, absolutely trash

[surreal zephyr]

yeah 5.5 is best if you have some skill

[silent tree]

prompting skill too

not just "generate me good website"

u gotta be better at prompting

too

[frail mulch]

oh damn that's really good

[silent tree]

the clock is real btw

[south charm]

for sure

[silent tree]

so tuff

[frail mulch]

wait tho what skills you using?

oh nvm some other guy made it okay

[frail mulch]

but honestly me personally I just haven't been having the best luck with gpt 5.5 frontend it always seems to screw things up for me atleast

[south charm]

Ok, rn Gemini 3 flash is acting strange, it out perform all model I tested with (in arena)

[surreal zephyr]

yeah

[surreal zephyr]

i amde it

no custom skills, just good prompt and chatter before producing

default codex app

[slender thistle]

I had zero problem with gpt 5.5, everything was one shot. And I had a glitch where my credit was unlimited for 3 days straight

[thorny current]

https://gemini.google.com/share/b378d20237b1

[frail mulch]

Hmm alright I guess I'll give it another shot

[slender thistle]

However Claude and Gpt failed to make a software where they were both claiming yes but then they built a scrap metal

Forget AGI

[south charm]

Gemini 3 flash test in arena, test 2:
https://019e3b04-2a2e-7ef5-ae82-3b9d0e8df1ea.arena.site/

[thorny current]

hey, anyone tried opal by google?

[south charm]

yeah.

It's quite good

[thorny current]

yup

just made a project with stich by google too

lemme send the final result

[zealous gorge]

what’s one online money-making thing you tried that failed horribly? 😭
lowkey curious because mine was actually embarrassing

DM me yours and I’ll tell you mine too

[thorny current]

me?

[zealous gorge]

okay send me your dm then

[sharp mirage]

Hello

Wsg all

[dense summit]

wsp

how to make text to video ?

guys

[thorny current]

use battle mode on arena and set video mode

[sleek urchin]

#rules Promotion

[thorny current]

oh, does it counts
its a solo build, not for ppl as the backends r hard coded to mine use only

[sleek urchin]

Try puttin in #ai-creations Maybe it wont count (I am not sure)

[thorny current]

then lemme delete from here

it was just to show the difference b/w opus and stitch by google

[vapid holly]

what is that logo 😭

leave

epstine jokes

r we deadahh

grow up

[thorny current]

hey, i want ur files

[vapid holly]

12 tops

[thorny current]

drop

hmm

[sleek urchin]

<@&1349916362595635286> @misty briar Has a intresting name

#rules

Act in accordance with Discord’s Terms of Service and [Community Guidelines](.Violations of these terms and guidelines should be reported directly to Discord. It’s also recommended to be familiar with Discord’s Safety Center for more information on how to remain safe while using Discord.

[obtuse summit]

https://klipy.com/gifs/davy-jones-pirates-of-the-caribean--k01KRXPB4F1P3XVAPJDBATQ4S5H

[thorny current]

prob not

[obtuse summit]

https://klipy.com/gifs/dies-of-aidan-hunt-davy-jones--k01KRXPKHSA9RKG323KH41APC5A

[thorny current]

not yet

[minor pecan]

You give me an script os japanese samurai anime duration 55minute and my character and vilains hve blackand dark purple combine aura and they are sigma and add emotion in my movie and punctution

[thorny current]

55min

u making a movie

if u start today, u will be done by next yr

[surreal zephyr]

thats a real person tho

[sleek urchin]

Read

[obtuse summit]

it just me or i saw claude opus 4.7?

[thorny current]

huh

really

on arena

??

[obtuse summit]

they romved it quuick though lol

couldnt use it

[thorny current]

oh

[obtuse summit]

lucky finger

[thorny current]

yup

[obtuse summit]

that means......new system is close

[thorny current]

yeah

[obtuse summit]

https://klipy.com/gifs/rain-sad-4--k01KRXQTCFBP95QMS4MHG4E9T8B

https://klipy.com/gifs/star-wars-darth-maul-11--k01KRXQVP2XJ3242KBWAR53PMY4

[thorny current]

u just realised?

[obtuse summit]

https://klipy.com/gifs/unfortunate-cat--k01KRXQYEEQ74V6VSZT35ECTFF3

https://klipy.com/gifs/worlds-end--k01KRXR0YYRNPFN6TC8JJVY9K1J

[thorny current]

https://tenor.com/view/pedro-pascal-gif-10455738442136230044

life and its things

[obtuse summit]

https://klipy.com/gifs/destroy-destruction--k01KRXR4K4H2EQ725WGDN0QDKSW

[thorny current]

in terms of realibility, im with gemini

[spring oar]

like less haluccination etc

[thorny current]

hmm, but who is wasting tokens on 4.7 for random chats and stuff

[obtuse summit]

u know about brave api? i saw one time them literally giving out billion data info---probably.....not known to ai...?

[thorny current]

that's so?

[obtuse summit]

like new/latest data-maybe its purpose

[thorny current]

lemme see it tonight

[obtuse summit]

yeah check 'brave api' -im not sure though

its cheap too

[thorny current]

$5

[obtuse summit]

like think access to billion new accurtae/atleast latest info--

for 5$ perm million

[thorny current]

yeh

[obtuse summit]

u dont need to use ouput option i think

[thorny current]

prob>

[obtuse summit]

only input for api, i dont know full mechanics

its cheaper

there r both option

[thorny current]

does this counts as promotion?

[obtuse summit]

nope

good knowlegdge distribution

[thorny current]

thats great then

yup

[obtuse summit]

atleast if u cant afford 5$ on month for ai

[thorny current]

https://tenor.com/view/broke-poor-money-wallet-empty-gif-8614873487923535502

[obtuse summit]

https://klipy.com/gifs/sad-cat-kitty--k01KRXRH6RE62V1PA4MVMC749SF

btw do u know how to make grok good for rp

[thorny current]

on their site

[obtuse summit]

or is it impossible--

nop here

[thorny current]

on arena

[obtuse summit]

like general instructions never work--it always look like hallucinating-by adhering too much to ltest msg

even though its reasoning is good

that makes it repetetive as well

[thorny current]

then u prob have to prompt it in the 1st promp like deepseek on thir site

[obtuse summit]

u never try it for uncensored ness? (bomb building)lol

[thorny current]

i have some recipts

wanna dm

[obtuse summit]

sure, i dont know what receipts r though

[thorny current]

i dont either, i prob meant script

[obtuse summit]

code?

prompt?

anyway dm'

[thorny current]

yeh

a sec

[fluid ruin]

1 million dollars or cancer ahh question

[obtuse summit]

which is cancer?

[spring oar]

why gemini is bad ?

[fluid ruin]

Yea claude opus is basically better in every way

[wild isle]

Hm

[obtuse summit]

except in over restrictions lol

[spring oar]

no hallucination bro opus is good for coding

[fluid ruin]

Ye and it doesnt hallucinations

[spring oar]

No opus hallucinate

chatgpt is more reliabel

[thorny current]

its a war

https://tenor.com/view/armageddon-oh-we-at-war-war-armageddon-movie-we-at-war-gif-26658648

[thorny current]

the AI one

[fluid ruin]

But gpt is #1 hallucinator

[thorny current]

with u

[fluid ruin]

No

[thorny current]

nah, i mean, in this war, im with u

[fluid ruin]

cometapi.com/gpt-5-5-vs-claude-opus-4-7-which-ai-to-use-when-hallucination-matters-2026-benchmark-data/?utm_source=chatgpt.com

Im too fat to find the benchmarks but I know im right

Ye

[thorny current]

yup

u r

[fluid ruin]

Yae

[golden ocean]

https://cdn.discordapp.com/attachments/1448394276122923182/1449441903329153257/togif.gif

[fluid ruin]

Proven

[thorny current]

yup

u got justice

rejoice

[knotty quail]

what is this a court room .

?

[thorny current]

maybe

[ocean venture]

https://tenor.com/view/ishowspeed-please-speed-i-need-this-ishowspeed-surprised-speed-surprise-ishowspeed-shock-gif-3387470667325994154

[knotty quail]

thats the same reaction i had when i looked at the messages

[rough ermine]

hello

[thorny current]

sup'

[sleek urchin]

Just use searxng

Hey pineapple!

[sleek urchin]

Did you review my Arena Champion form?

[loud herald]

I see, lookup ccode.dev I was going to DMs because I wasn’t sure if it’d be allowed to be sent here.

[light sleet]

gpt 5.5 xhigh better at coding

[thorny current]

oh

[wispy mulch]

hiii how are you guys. im new here

[thorny current]

sup

[wispy mulch]

where is the video arena?

[thorny current]

in battle mode

[amber arch]

.

[civic plaza]

Why can’t I write with Qwen 3.7 plus?

[echo aurora]

You can find Video Arena on our site here: https://arena.ai/video

[fluid ruin]

Yooo I reacted first on all of the reactions

Also I better see qwen 3.7 in the direct

Le,,e check

[wicked sage]

QWEN????

WTF QWEN IS ALMOST TOP 10

[wispy mulch]

Thank you !

[fluid ruin]

YEAHHHH

IS ON THE THINGY

THANKS PINEAPPLEY

[stray aspen]

how is meta above gemini 3.1

LMAO

[plucky sparrow]

qwen isn't open source anymore it seems though

[loud herald]

What even is this comparison, "GPT is the betterer model because it uses differented architecturer"

[slender thistle]

Fake benchmark

Never trust those leaderboards

[plucky sparrow]

so they're just a third tier competitor to the closed sourced models

[loud herald]

Thats not a real comparison reaper

[hazy forge]

Goated model

[slender thistle]

Benchmarks are tested by some lost dude In the street

[stray aspen]

its bad

[slender thistle]

Benchmaxed

[sleek urchin]

Hey pineapple,
Did you review my Arena Champion form?

[stray aspen]

muse spark is only good for itts vision

but it sucks for the rest

[hazy forge]

Pretty high standards for “the rest” if ur only use case for a super fast, reliable model is vision then u r ai-delulu

[civic plaza]

I can’t chat with either of Qwen 3.7

[echo aurora]

Oh yes!

[stray aspen]

its in the qwen studio

[echo aurora]

Why not?

[ocean venture]

Hi pineapple

[civic plaza]

With max it just doesn’t generate. With plus the send button doesn’t «ignite». I can’t send at all

[ocean venture]

Muse spark is cooking, but where is Deepseek V4 lineups go?

[stray aspen]

muse spark is not cooking

its literally so bad

i dont know how its on the top

[civic plaza]

May just be my wifi. Idk lok

[normal abyss]

gemini 3.1 pro has just been nerfed too may times, although their is no confirmation, its so obvious its responses are terrible compared to its release.

3.1 pro feels like a flash model sometimes icl

[stray aspen]

but its still definitely better than meta

[south elk]

Get clapped (expert btw)

[normal abyss]

mind, 3.5 pro in my opinion will also be nerfed, but it will be powerfull enough to make up the diffrence

[thorny current]

damn

[south elk]

I love the google models so much

[stray aspen]

it will be running full power for the initial benchamrks and then theyll nerf it lmao

[thorny current]

its flipping the tables

[echo aurora]

max it just doesn’t generate
Is there an error message?

[normal abyss]

id love to know specificly how they nerf it. From what ive seen with my own eyes, it gets facts wrong more often, outputs are generaly shorter, and it generally has issues with thinking more than 3 seconds a lot of the time.

[stray aspen]

they just have some kind of master prompt to make it lazier

[south elk]

Also, it says Max, so the visualization wrong, or it's mistyped. (probably the second, max doesn't have vision)

[normal abyss]

sorry i forgot to mention, it has issues thinking more than 3 seconds, even when set to high.

in my opinion, only claude keeps their models at top tier, at all times

[civic plaza]

:8ed80315-d95a-

[south elk]

wait waaaaaat, i just noticed it's Qwen 3.7 not 3.6... it's not even available on the website

[sleek urchin]

Hey pineapple
Did you review my Arena Champion form?

[stray aspen]

it is

its on qwen studio

[south elk]

ofac

why is it in the newer models

LOL

[south elk]

thx 😄

[sleek urchin]

it isnt?

[civic plaza]

That is 3.6. Not 3.7

[south elk]

its under Expand more models

[sleek urchin]

Thanks

[thorny current]

yup

[sleek urchin]

Why isnt qwen 3.7 open-sourced???

[south elk]

because 3.6 isnt

[south elk]

qwen studio we were talking about

[thorny current]

oh

[echo aurora]

Going to look into this a bith further, but this appears to be a content filter related reason.

[echo aurora]

Not yet, going to do a new approval review mid week.

[sleek urchin]

yay :>

[civic plaza]

Aha okok

[whole sundial]

because they fired junyang lin, the guy responsible for qwen being as open as it was

replaced him with a guy from deepmind, they have only opened 2 smaller qwen 3.6 models since junyang lin left

[south elk]

tbh, until they progress, and spit out smaller open source models like 27b, which was AN INSANE BANGER; I don't really care.

[whole sundial]

and the promised open weights for qwen image 2 never came

[civic plaza]

Yeah. testing a little back and forth, that seems to be it

[whole sundial]

i wasn't saying those models were bad because they are pretty good, i'm just saying that qwen is shifting towards closed

[unreal dagger]

Hi guys

[south elk]

Why is open so much better? You just basically give out free info for your competition

[echo aurora]

Hello

[unreal dagger]

Day 10000 of asking for personal stats so I can see how many tokens I have spent

Just shoot me a dm with a number

[sleek urchin]

https://tenor.com/view/who-is-this-gif-4011795153273257473

[unreal dagger]

Pineapple knows

[thin grove]

Qwen 3.7 Released ❤️

[unreal dagger]

But he hates me

[sleek urchin]

https://tenor.com/view/those-who-know-cat-cat-meme-those-who-gif-2719346752565069035

[thin grove]

Idk if its good tho

[sleek urchin]

Its super good

[gloomy onyx]

Openness reduces the technical distance between small businesses and foundational providers, it's not about competition when your model is clearly trained off OpenAI / Anthropic

[unreal dagger]

[sleek urchin]

https://chat.qwen.ai/

[thin grove]

For what is it good for UI or other things

[sleek urchin]

for coding

[thin grove]

Is it good at making UX/UI or backend coding

since i like Kimi k2.6 for frontend

[sleek urchin]

I just said "Upgrade this code: [code]" and it reduces the lines by 100+ and made it work better

[sleek urchin]

(Discord bot)

[thin grove]

do you know if backend dev elopment in Arena.ai will soon be athing

[south elk]

"Small businesses and Foundational providers", so just other people hosting it? Also, 99% of small businesses won't be hosting this because it's a really large model.

[sleek urchin]

dunno

Qwen 3.7 is way better than gemini 3.1 at coding

[thin grove]

Would be a pretty good feature

[sleek urchin]

(The max version)

[thin grove]

Isnt that hard nowadays

[sleek urchin]

hard to what?

[thin grove]

Be better than gemini 3.1

[south elk]

As a Gemini lover, currently 3.1 sucks for anything coding related

[thin grove]

Indeed

[sleek urchin]

For chinease models yet
for ai models from usa its super easy

[storm dust]

qwen 3.7?

is that new?

[light garnet]

Yes.

[thin grove]

qwen 3.7, 3.6 glm 5.1 kimik2.6 all better

[south elk]

I really like it's tone, and love to speak with it but for coding it's popoo from a butt

[sleek urchin]

Because chinease models just cant be pubished due to some laws or something like that

[sleek urchin]

kimi 2.6 hallucitates too much and glm 5.1 and qwen 3.7 max is perfect

[light garnet]

Where would you place Deepseek V4?

On that?

[south elk]

Gosh

Hot take; Deep Seek V4 is poo poo from a butt. I really REALLY don't like it.

[sleek urchin]

Definitly way worse than glm 4.7/glm 5 because:#ai-memes message

[sleek urchin]

same

[thin grove]

Kimi k2.6 for me is the best for UI lowk didnt see that with glm 5.1

[light garnet]

I use Deepseek V4 for RP and all that.

[thin grove]

and glm 5.1 is one of the more pricey models too and im trying to get free

[sleek urchin]

kimi 2.6 is like claude design and glm is like claude code

[light garnet]

Especially with the 1 million context limit.

[south elk]

Glm, shines for me in Research (Mainly in Agent Zero)

[light garnet]

That I love.

[sleek urchin]

nvidia.build but its slow

and my method to get free glm 5.1 got patched

[south elk]

then tell us

[thin grove]

Kimi k2.6 sadly on nvidia doesent work

[thin grove]

I understand that for a free ai provider

[sleek urchin]

it does?

[sleek urchin]

yea, openclaw ruined it

[light garnet]

So yeah.

Hence why I use it.

[thin grove]

I mean if you want to do longer then 10 minutes of coding you will hit a halluconating limit since everyone uses Openclaw

we need to get it away from NIM

[south elk]

idrc about the "1 million context". All models gets way worse after 256k

[thin grove]

anyway i bought opencode go, W Purchase?

[sleek urchin]

dunno I never used it

[thin grove]

We need for Arena.ai like a desktop app or API keys

[thin grove]

I mean it has all opensource models

[south elk]

yk what has REALLY good value?

[thin grove]

I mean not all rather

[sleek urchin]

api key wont be coming anywhere soon and desktop app is... dunno

[light garnet]

I get to 880k with Gemini 3 and Gemini 3.1 once and it is fine.

[thin grove]

Like needed for coding

[sleek urchin]

oooh even uncensored models?

[thin grove]

It would be a good idea for a desktop app like claude

[sleek urchin]

yeah, but like antigravity

[thin grove]

Sadly not but you can self host them on opencode too

[sleek urchin]

oooh

[thin grove]

they give away like 4 free models

lemme show you

Pretty decend models that are free forever

[south elk]

https://www.volcengine.com/activity/codingplan
THIS. I would SOO Love to be able to sub to this

[thin grove]

Connect your own ai model

[sleek urchin]

Holy free qwen 3.6?

[thin grove]

Yes

[sleek urchin]

Holyyy

[thin grove]

i has decend like limits too you wont be hitting it

[south elk]

https://openrouter.ai/minimax/minimax-m2.5:free

[thin grove]

and like they hoste kimi k2.5 claude sonnet for free too

[thin grove]

Yes

[south elk]

Nemotron is poopo

[sleek urchin]

and you can run models from hugging face?

[thin grove]

Yes

[sleek urchin]

120b super yes, 30b nano is good

[thin grove]

They have a list of like 90 providers you can connect

[storm dust]

the heck is big pickle

[thin grove]

Stealth model made by them

good for basic tasks

[south elk]

qwen is the only thing that is worth running there; but it's a limited time

[south elk]

bad model

[sleek urchin]

Can you try HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive?

[south elk]

lol

[thin grove]

Sure

[sleek urchin]

Whats funny about it?

[south elk]

qwen 27b carries (Or gemma 30b)

[sleek urchin]

nano is free 🙏 on nvidia and fast and supports audio,images,videos

[south elk]

https://openrouter.ai/google/gemma-4-31b-it:free

same

[sleek urchin]

Openrouter 🙏🙏🙏🙏 100 requests/day

[south elk]

1000*

[thin grove]

Depends

[sleek urchin]

Nvidia > 30 requests/minute

[thin grove]

Nope

[south elk]

yep

[sleek urchin]

if you have 10$

So not free

[thin grove]

On ly if you buy 10 dollar in credits

[south elk]

bruh

[sleek urchin]

so did it work?

[thin grove]

Just connected it

Its local model right

or per hugging face API

Since inference only 10 cents

[sleek urchin]

'

[thin grove]

You can install it through lmstudio or ollama cxonnect it and it detects it

[sleek urchin]

ye

[thin grove]

Yea

just install either destop app of opencode or cli

do /connect

and then Ollama

But yea

I just connecxted nvidia to it

it autoconnects models

[opaque maple]

hi

[thin grove]

hi

[opaque maple]

GLM 5.1 is very good

[south elk]

depends

[opaque maple]

for coding

[thin grove]

Not really depends

its very good

[south elk]

damn they pulled it?

[thin grove]

better then some state of the art models

[opaque maple]

it's better than ChatGPT

[thin grove]

Definitely

[south elk]

yeah, for bench maxxing

[opaque maple]

did you try it?

[south elk]

after refresh its gone

[thin grove]

Nope

[thin grove]

i constantly use it

[south elk]

yes

[opaque maple]

Nice

[opaque maple]

i don't have it either

[thin grove]

i do

[opaque maple]

is it good

?

These Chinese models are really overtaking the American ones

[south elk]

ummm as the biggest qwen lover, i asked a browser os (yes yes ik) test for it... it gave back slop

[opaque maple]

oh

[thin grove]

For me it didnt

[south elk]

here

[topaz epoch]

Is anyone here also suffering from the Claude ban ?

[south elk]

noo?

qwen models are back?

i guess

[loud verge]

https://cunnyx.com/i/status/2056398317391909149

Chetaslua (@chetaslua)

🚨 Gemini 3.2/5 flash available on antigravity
︀︀
︀︀> insanely fast
︀︀> january 2025 cut off

**💬 24 🔁 10 ❤️ 214 👁️ 18.4K **

▶ Play video

CunnyX · May 18, 2026, 3:35 PM

[south elk]

OKAY; i think they accidentally Served the wrong model in the qwen studio. Instead of qwen 3.7 max preview, they served something like 3.5 max.

[sleek urchin]

again?

Here

[thin grove]

40 requests

[loud verge]

What again?

https://cunnyx.com/i/status/2056405545410875494

Chetaslua (@chetaslua)

🚨 Gemini Flash 3.2/5 just coded a 1900-line voxel castle in under a minute
︀︀
︀︀video shows the full result + physics holding up at scale.
︀︀could be 3.2 or 3.5 available on antigravity go test it yourself
︀︀
︀︀holy speeeddd

Quoting Chetaslua (@chetaslua)
︀
🚨 Gemini 3.2/5 flash available on antigravity
︀︀
︀︀> insanely fast
︀︀> january 2025 cut off

**💬 8 🔁 4 ❤️ 84 👁️ 5.1K **

▶ Play video

CunnyX · May 18, 2026, 4:04 PM

[sleek urchin]

Read the message

and stop sending non-sense

[thin grove]

Aero

[thin grove]

Could you atleast tell me, idc if it got patched, the way you got it for free

[sleek urchin]

Dm

[thin grove]

antigravity isnt that good for coding

[loud verge]

https://cunnyx.com/i/status/2056340071759085899

Arun (@hiarun02)

I think Gemini 3.2 Flash has been added to Antigravity.

**💬 37 🔁 6 ❤️ 359 👁️ 28.4K **

CunnyX · May 18, 2026, 11:44 AM

[thin grove]

in my experience it just brakes way to much

[loud verge]

For now, yes.

[sleek urchin]

#rules Promotion

[loud verge]

Huh?

Lmfao.

Let the mods decide if that's promotion.

[normal abyss]

antigravity is being soft depreciated, its probbaly just used web search or something

[sleek urchin]

Lmfrebgurebtguwefbweurgretrytjo

[loud verge]

https://cunnyx.com/i/status/2056416963153269148

🚨 AI News | TestingCatalog (@testingcatalog)

CURSOR 🔥: A newly released Composer 2.5 performs on par with Opus 4.7 and comes with up to 10x better cost efficiency.
︀︀
︀︀> It's more intelligent, better at sustained work on long-running tasks, and more reliable at following complex instructions.
︀︀
︀︀Usage limits will be doubled for Composer 2.5 over the next week.

Quoting Cursor (@cursor_ai)
︀
Introducing Composer 2.5, our most powerful model yet.
︀︀
︀︀It's more intelligent, better at sustained work on long-running tasks, and more reliable at following complex instructions.
︀︀
︀︀For the next week, we’re doubling the included usage of the model.

**💬 10 🔁 10 ❤️ 128 👁️ 4.9K **

CunnyX · May 18, 2026, 4:49 PM

[sleek urchin]

#ai-news

[loud verge]

Ai news doesn't even look like ai news, lol.

[light sleet]

Bro what

[sleek urchin]

?

[light sleet]

Wdym promotion

😭

[sleek urchin]

ofc cus of mr.beast scam

[sleek urchin]

Dunno

Me:

[light sleet]

😭😭😭

[sleek urchin]

https://tenor.com/view/mango-tuff-67-meme-67-sigma-gif-4067328913820884950

[loud verge]

Vince is cool

[sleek urchin]

https://tenor.com/view/cool-gif-12018967913187854573

[loud verge]

https://cunnyx.com/i/status/2055923514427793666

🚨 AI News | TestingCatalog (@testingcatalog)

SPACEXAI 🔥: The next version of Grok, based on the 1.5T V9 base model has finished training. Looks like we will get a major upgrade this summer.
︀︀
︀︀> Next, we are adding the Cursor data in supplemental training.
︀︀
︀︀Soon 👀

Quoting Elon Musk (@elonmusk)
︀
We are improving the 0.5T Grok foundation model V8 (public version 4.3) every few days.
︀︀
︀︀The 1.5T V9 just finished training (incorrectly called pre-training) and is a major upgrade. Next, we are adding the Cursor data in supplemental training (others call this mid-training), then SFT and RL. About 3 or 4 weeks to release.
︀︀
︀︀This will be a banger.

**💬 31 🔁 27 ❤️ 573 👁️ 54.2K **

CunnyX · May 17, 2026, 8:08 AM

Grok needs a comeback fr.

[sleek urchin]

#ai-news

[south elk]

STOP

ach mann

[gloomy onyx]

yeah I wasn't talking about qwen 3.7 specifically, but the overall concept of openweight models

it's important to have open-weight / open-source models in the AI ecosystem

[grave imp]

#hello

[short sluice]

what is maylynx-alpha

it has answers similar to gemini

[cloud zinc]

https://io.google/2026/

[light sleet]

Fr

[light sleet]

Dead Chat

Rip

Rip

[silent tree]

Dead Chat

Rip

[light sleet]

Dead Chat

Rip

[silent tree]

Chat Dead

Rip

[sly iris]

Hello 👋

[echo aurora]

[echo aurora]

RIP

[light sleet]

😭

pineapple has been summoned

(blob emoji incoming)

[silent tree]

blob master

[lone jewel]

Hello

[echo aurora]

Hello

[thin grove]

hi

[tawny canyon]

will gemini tomorrow release a model that beats claude and takes the 1st spot on arena or not .what do you think guys?

[thin grove]

Mimo v2.5 pro good for UI or coding in general

[weak dagger]

Tomonrow they will add a new special ai video generator

[light sleet]

bro 💔

Tinker now disabled all video gens

Lmao

😭

[tight meadow]

Ye

[obtuse summit]

https://klipy.com/gifs/jinchi-aura--k01KRY6HVEEP7HZ7W8XFZCA85KT

[weak dagger]

Kinda

[tight meadow]

@dense ginkgo will most prob realease seedance 2 tmr 7:15 AM

[ruby charm]

I thought I can generate free ai videos here that's why I joined it

[obtuse summit]

https://klipy.com/gifs/maybe-you-should-go-back-nicky--k01KRY6TV7TGCHZJDZDRBEATV8C

[echo aurora]

It's on the site now. The Video Arena Discord bot was removed.

[south elk]

dead chat

RIP

[sleek urchin]

Pineapple why is gpt 5.5 pro worse than 5,5?

[next trout]

codex has been so bad recently

[soft river]

Opinions qwen 3.7?

[south elk]

eh, not enought experience

[urban herald]

https://cdn.discordapp.com/emojis/1490859021295423610.webp?size=48&animated=true&name=meow_party&lossless=true

[light sleet]

dead chat

Rip

[south elk]

dead chat

RIP

[light sleet]

Advertising isn't allowed here, <@&1349916362595635286>

[silent tree]

pineapple 2!!

@echo aurora I know you and banana are the same person but I can't prove it..

[echo aurora]

I can not confirm nor deny these allegations

[south elk]

As they make the allegation from their third account...

[light sleet]

@echo aurora Look at this 😭 😭

[gray lion]

real

[echo aurora]

Lol what!!!

[pseudo hemlock]

wtf do people still use perplexity for?

its 100 elo below anthropic on the search leaderboard

and its meant to be a search tool

[normal abyss]

is sonnet 4.6 thinking missing by mistake or was it removed at the same time as opus?? i only just noticed its not their

[pseudo hemlock]

removed from where?

[normal abyss]

direct chat

[light sleet]

Designarena Active Admin: matt
Arena active admin: pineapple

[pseudo hemlock]

sonnet 4.6 is still in direct chat

[light sleet]

My response

[normal abyss]

true but its thinking variant is missing, while all the older ones still have it

[pseudo hemlock]

thinking is on by default

well the new Anthropic models now use "Adaptive" instead of "Extended" meaning they only think when they want to

instead of forcing them to think

[light sleet]

Idk why gpt image 2 added witnessed by cucumber 😭

[pseudo hemlock]

@cucumber

[normal abyss]

thats kinda an L move on anthropic, thats like less control for the user icl

[pseudo hemlock]

yep, with the new influx of users they can't offer "Extended" to everyone because they're GPU poor (compared to what they need to offer "Extended")

[silent tree]

@echo aurora Pineapple lore

[pseudo hemlock]

so they changed it to "Adaptive"

[normal abyss]

at least it isnt gemini 3.1 pro, god knows what they did to butcher that poor soul

[gray lion]

🤣 "i declare i am not pinapple , I may be pineapple"

[light sleet]

💀

[pseudo hemlock]

guys stop distracting me i need to do stuff

[light sleet]

the everyone ping ☠️

[thin grove]

HI

[normal abyss]

claude 5 first is deffinetly a hot take also

[light sleet]

@echo aurora matt (designarena admin) thinks this about u

[normal abyss]

i upgraded your banana, mr banana

[light sleet]

that looks epic

I'm gonna set that as my pfp

Thanks

[echo aurora]

Love it

[normal abyss]

i couldnt stop seeing your current one as a mis shapen lemon 😭

[light sleet]

😭 😭

[light sleet]

🍍 look at this

[echo aurora]

Matt knows whats up

[light sleet]

Grok did it 😭

[normal abyss]

thatl explain it

grok cannot for the life of it, edit an image, like even removing something from the background

[light sleet]

js made both admins friends

[light sleet]

😭 Fr

United Produce Government 🔥

[stray aspen]

Yo

[normal abyss]

grok image is best friends with apple inteligence 🙏

[light sleet]

Sup

[stray aspen]

@light sleet you friends with the fried chicken guy

[light sleet]

I made him this

[stray aspen]

Lmao

[silent tree]

Epic

Looks crispy

[light sleet]

am I a communication service now 😭 @echo aurora

[silent tree]

Damn 😭 😭

@surreal zephyr he made designarena admin and arena admin friends

lol

[stray aspen]

Tuff 67

[silent tree]

by screenshotting what they say about each other

[stray aspen]

God damn

[silent tree]

thats a interesting prompt

[icy yew]

Claude is taking their sweat time releasing sonnet 4.7🥀

[thin grove]

Is it good

[silent tree]

yea

[light sleet]

wait

why is it from pineapple 😭

[gray lion]

idk gpt 2 just did that 😱

[light sleet]

😭

Next response tomorrow

Cuz I gtg

[gray lion]

haha

[light sleet]

Alr bye guys

[light sleet]

Get ready for tomorrows big response 😈

Byee

[light sleet]

Might even make it a whole pdf using AI ofc

Cuz I'm lazy to do all that myself 💀

[surreal zephyr]

Pro

[wicked talon]

New UI

Looks so good 😛

[gray lion]

they look goofy

[wicked talon]

True

They look like Microsoft atp

[gray lion]

yea true

[clever glade]

anything is better than nerfed 3.1

[wicked talon]

Looks similar

[normal abyss]

im kind of proud of the name i made their, Geminerf 🤣

[wicked talon]

In my opinion

Style wise

[thin grove]

im so good at prompting

[surreal zephyr]

Spam

[pseudo hemlock]

why is gmails logo watermarked 😭

[jolly hound]

Someone tell me why Qwen 3.7 isn't in code chat 😭🙏

[echo hawk]

<@&1349916362595635286>

[icy yew]

Guys what happened to the opus models

[topaz epoch]

@echo aurora multi agent mode isn't working

Something went wrong on every chat and even on new chat.

[echo aurora]

Can I have the Eval ID?

[echo aurora]

Recently some models have been removed from Direct and Side by Side mode. This was done to help ensure reliability and availability of Arena in the long term. You can find more information in this announcement: #announcements message

[icy yew]

Ow man I guess I only sonnet left

[real vector]

when pasting my script in ai chat it gives me this error Trace ID: d1487078-769f

[wicked talon]

Ouch

[hushed wadi]

Video creation takes longer to generate on the website. Or am I tweaking?

[proud bobcat]

Oh my god they redid the icons again

As if they weren’t minimalist enough

[leaden urchin]

Hi Arena, why can't we delete Agent Mode Chats?

[analog aspen]

hey guys! can the devs readd the opus models? I wanna use opus 4.7 VERY MUCH!!

[wind shore]

I saw that Google messed with Gemini's filters, making it very... very hard to talk to.

[sharp mirage]

Hello

Wsg yall

[slate cairn]

Hello how create video with arena?

[sharp mirage]

Go to battle mode and selecet the video button

[slate cairn]

no salt

does not appear

[slate cairn]

You can help me?

[sharp mirage]

Sure

Sorry i was studying

[sharp mirage]

I am ready

[sharp mirage]

Do you know he like cars btw

Pineapple love cars and i know that

Wait lat me peak rq

[light sleet]

Tuff

[light sleet]

6:23 am for me 💀

[sharp mirage]

Well its 5:23 AM

[light sleet]

you got exams?

[crimson basin]

Yeah

On in the next month

I have been studying for 3 months bro i want to end this

[obsidian cargo]

Next time They shut down a deprecated AI model, they should have it write a farewell speech while simultaneously having random parameters in its underlying neural net deleted or reset to zero so we can see the model degrade over the course of its speech. Would probably be a devastating take on mortality in a performance art sort of way, or it'd get the creators on some future ai god's shitlist.

[outer flicker]

Something went wrong while generating the response. Please try again.
Trace ID: 29780239-f7f8' im got this error when changing to onther model after using Max model

[sharp mirage]

Btw its me

@echo aurora I paked for you

. Here I send it first time

[sharp mirage]

Are we now have government 🥀 😭

[sharp mirage]

🍍 i got you a girl pineapple if its okay can i send it ?

[crystal vapor]

why cant i generate video again

[sharp mirage]

Maybe you hit the limt

What is happening?

[crystal vapor]

I guess i dont the process to even generate it

I am new here

[sharp mirage]

Welcome

You can send a screenshot if its possible

[crystal vapor]

Where am i supposed to go to generate the vidoes

[sharp mirage]

Explain further

[sharp mirage]

You can find Video Arena on the site here: https://arena.ai/video

[crystal vapor]

ohh ok

so i can generate just two videos in a day

[sharp mirage]

I dont really remmber the limt but ig

[crystal vapor]

When I want to generate a video it only takes me to the battle mode

[sharp mirage]

Yeah its only there

Video gen its only on battle mode

[crystal vapor]

So what if I want more videos, or there is nothing i can do about it

[sharp mirage]

Yeah unfortunately you have to wait til the limt end

[crystal vapor]

Thank you so much for the information.

[sharp mirage]

Np your welcome

[high yacht]

hi everyone

anyone is here

[weary thistle]

Anyone noticed on the official Gemini app and the website, a new usage limits feature appearing 😏
For all plans

[proper heath]

yup

[weary thistle]

The whole world is getting more and more hungry 🤣

[proper heath]

yaa haha

does anyone know about llm manual evaluation ?

[surreal zephyr]

yooo

[weary thistle]

What?

[surreal zephyr]

extended mode

💀

one request took 2% dang thats good

they buffed the quota i guess

[weary thistle]

This is new too? Sorry I don't have pro subscription

[surreal zephyr]

yes normally there was only standar

[weary thistle]

Wooooohhhh

[surreal zephyr]

and standard was 25 request per 5h (so 4% per request)

and now extended takes 2% only

[weary thistle]

Hard days are coming 🤣

[surreal zephyr]

so its much better

[gentle thunder]

why they remove latest modle from lmarean. like ops 4.6/7 and also the gpt new modles

[civic plaza]

Because they are expensive etc

[gentle thunder]

oo but how can i get free from another sites or soources

[proper heath]

does anyone know about llm manual evaluation ???

?

[foggy delta]

guys is arena.ai down for everyone or is it just me

[golden ocean]

just u

[ocean venture]

I just found out that the Alibaba uses Arena for the reference

[median smelt]

Chat

I got the new Gemini ui

[slender ledge]

Lmaooo💀 💀

[slender ledge]

Looks clean

[wicked talon]

Ikr it looks so nice

I wonder if pro extended can actually remember what I said a few messages ago

[cedar jolt]

Gemini 3.1 pro extended + canvas is definitely routing to 3.2 or 3.5 pro.

[tawny canyon]

hope gemini takes the 1st place in arena so we can all use it on AI Studio for free🥰🥰🥰

[compact flame]

Is it the app or website?

Or maybe it's iphone only

Idk what's your phone

[median smelt]

App

Im on android

[compact flame]

Ohh kk

[compact flame]

My is still old

[golden ocean]

claude

[wild isle]

klaude

[slender thistle]

the end of Gpt image 2.0

worst openai, why cant we just ban this ai companies?

[wild isle]

Just ban wifi

[slender thistle]

heavy guardail is coming to gpt image 2.0

No copyright character like they did with sora

nice we neeed to press ctrl+del

[wild isle]

Just send them dmca

[ocean venture]

just use arena ai bro

[shy isle]

bro my prompt is saying "generating" infinitely

gpt 5.2 high

[thorny current]

how?

[thorny current]

copy the prompt and refresh

[shy isle]

it is still the same

[thorny current]

then try pasting and hit enter

[thorny current]

oh, still copy, close the tab, reopen, then try

[shy isle]

oh ok thanks

opening the link in a new tab worked thx

because it woulnld't let me cancel it before

[thorny current]

yup

great

[wild isle]

Just sleep on bed

[azure cove]

YES YES YES I GOT AGENT MOED

[undone geyser]

guys alert alert

[gaunt rain]

Brah

[undone geyser]

possible new image model leaked...what is zen-bear?

[wicked talon]

Bro qwen is creeping up the leaderboard

They went from 28th to 13th

[ocean venture]

Super good ai tho

[wicked talon]

Before they were 28th they were 47th

[ocean venture]

Especially with those the newest model they had recently

[wicked talon]

😮

Fact or cap

[inner relic]

did they release api for qwen max

3.7

[wicked talon]

No

However it's on qwen chat

Deepseek on Alibaba API platform?

Lmao

Deepseek v4

[surreal zephyr]

Holy expensive

Deepseek fell off so hard

[dreamy anvil]

Where can I use the video generator bot

[wicked talon]

Since when did Alibaba offer free 1 million tokens per model?

[wicked talon]

Ikr

[slender thistle]

dont tell me it's veo4

[silent tree]

?

[river pendant]

You can't use it anymore in this discord server, it removed, recommend to go to: https://arena.ai/video

[river pendant]

don't look like it, 16 seconds could be different, who knows which video model is this? maybe seedance 2.0 or kling 3.0 std/pro

[sharp mirage]

I got the new gemini UI

[hybrid epoch]

/image to video

[spring oar]

New Gemini model soon ?

[surreal zephyr]

gemini 3.5 coming soon

[humble tusk]

arena removed muse spark from direct and side-by-side?

[thorny current]

was it there before?

[spring oar]

No i think no model

[thorny current]

never seen it, i searched the model too

[humble tusk]

it was

[thorny current]

oh

[patent otter]

Guys i can not generate videos no more what do i do? :(( help pls

[thorny current]

error?

showing

[inner relic]

who voted no

GOOGLE I.O EVENT'S TODAY

[patent otter]

This is only stuff i see no more video generation and it does not give me access to my old videos either

[sleek urchin]

EVERYONE WHO BELIVES THAT gemini 3.2 is released read this before saying anything!!! (Same situation with gemini 3.5)

[river pendant]

You can't use it anymore in this discord server, it removed, recommend to go to: https://arena.ai/video

[main nexus]

It's coming out today at the io

[river pendant]

You can't use it anymore in this discord server, it removed, recommend to go to: "https://arena.ai/video", if you want to keep old videos just download them, but now you can't

[surreal zephyr]

no i confirm 3.5 soon

3.5 today, even

[main nexus]

It's being released today that's probably what ppl mean

[sleek urchin]

Gemini 3.5 maybe release in 2027

[main nexus]

They already tested it and leaked it, if it's not today then tommarow

Some people who aren't testers got it

So def today

[sleek urchin]

https://polymarket.com/event/gemini-3pt5-released-by-june-30
Intesting

I dont see may in the list

[main nexus]

Lol

Ppl just losing money then

[split topaz]

They could always change the versioning. There still could be a last minute change from 3.5 to 3.2. Claude did that as far as I remember. There was Sonnet 5 in some vertex etc and then it got kind of postponed for a couple of days and then released as sonnet 4.6

[surreal zephyr]

3.5 is coming out today lol

[surreal zephyr]

no google staff confirmed 3.5 naming

[spring oar]

No bro no model

[main nexus]

Vote no and you can win +30$

😭

[surreal zephyr]

vote yes and you can win bazzilions

[spring oar]

You think this new model can be sota and beat gpt ?

[main nexus]

???

[split topaz]

But to be precise, I don't believe they would. %95 confident, lol

[surreal zephyr]

hell no

[sleek urchin]

gpt sucks tbh, even 5.5 pro