general

[echo aurora]

This announcement has more information about the program - #announcements message

[keen beacon]

I dont want it

[echo aurora]

Okay removed it

[pulsar saffron]

Lol

[keen beacon]

Thank you

[trim hound]

Can I chat my own Ai chatbot privately if so how please

[echo aurora]

The Video Arena bot can only be used in this server, you won't be able to use it in a different server or DMs

[trim hound]

Thanks

[void summit]

Hi. Anyone know why always has this?

[onyx verge]

I want to make videos, but I can't seem to get the commands to work

[void summit]

Is it lmarena limits my usage For 3 or 4 timed?

[echo aurora]

It looks like you aren't prompting the bot with slash commands. You'll want to type out /image-to-video first before you upload image + add prompt. In #1397655624103493813 you should fine a more in-depth explination that'll be helpful.

[echo aurora]

Yeah so this error that you're seeing can happen for different reasons which makes it a bit tricky to troubleshoot. The being the case the most common reason this is happening is you're being rate limited for too much use.

[void summit]

Oh my. Is kinda inconvenient If The result don't satisfied it naturally want to reproduce

[molten parcel]

hello

does anybody have any ideas about gta v on steam?

[echo aurora]

Would you mind elaborating a bit more?

[echo aurora]

Think you may be in the wrong server.

[stark arch]

Someone is looking for sora 2 codes? I have 5 and i would like to share with you guys

[void summit]

For example If i want photorealistic but it output to art. Therefore i have to rewrite My prompt again for image to prompt

[molten parcel]

thats why i asked does anybody have any idea

[sullen quest]

this is about ai

[echo aurora]

Gotcha. Rate limits are important but understandable why making adjustments to prompts is important too. In Battle Mode you shouldn't run into this error message.

[fiery gull]

change the model

[polar niche]

Good day

[knotty kelp]

HEY ALL!! QUICK QUESTION.. HOW CAN I SEE A CLIP I MADE IN THE TIMELINE???

[midnight hatch]

Wassup!

[fiery gull]

seach for your username in discord channel -> #video-arena-1

or another video-arena you maked the videos

[knotty kelp]

OH OK COOL ITS A PAIN TO HAVE TO SCROOL TO FIND YOU VIDEOS THE LINE MOVES FAST LOL .. THANK YOU

[jovial sapphire]

hi guys

i want very hard prompt

for gemini

3

if you have ideas, send me

[north cypress]

hello, does it exist RepoChat Arena ?
https://x.com/arena/status/1863655338949611862

[jovial sapphire]

example : #coding-champion

[fiery gull]

make a fucional BIC pen 3d, I want to see it

or 2d

[wary crescent]

hi

[jovial sapphire]

Send me prompt

full prompt

[fiery gull]

I'm busy and lazy now

[jovial sapphire]

ok nvm then

[pulsar saffron]

let me think it will take me a bit longer

[brittle tiger]

gemini 3 targeted for december release

[pulsar saffron]

gemini 3 vs gpt 6

one of them will win and one of them will lose

👏 👏 👏 👏 👏

[warped narwhal]

How can i use sora here?

[north vale]

i don't understand, why don't people at least try to like screen record their screen while it's generating this, to "prove" that they are not just completely lying and making stuff up about these genrations? they get too many views to be trustworthy, there's gonna be grifters just committing fraud to get likes, there's bound to be

like actually i have no idea which of these "this is gemini3 output" tweets are just lies vs not

[pulsar saffron]

its real and no one is fake

im real and ur fake spreading fake news

jk

it's A\B tested models

thats how they got gemini 3

randomly

[north vale]

ik that's the claim but why aren't they showing it in a way that's at least hard to fake

like currently they're so hard to fake and it's such an easy way to get like 100k views on twitter

so there's a strong incentive for grifters so I don't trust it

[daring tapir]

anywere to get free gemini canvas pro ?

[echo aurora]

Check out #1397655624103493813 for more information on how to use the Video Arena bot

Let me know if you have any questions.

[polar niche]

You could try my new ones

If you have agent mode or codex

[fiery gull]

I love when the sonnet 4.5 says: deciphers of competence instead of incompetence

How do I compete with a billion dollar text generator? 🤣

[polar niche]

Easy

[surreal estuary]

I'm new here and I don't know how to use the app. Help me 🙁

[hardy lion]

hi, we have a guide on how to use it #1397655624103493813

[hollow ridge]

i love ai and video is what I want to try out

[echo aurora]

You've come to the right place! #1397655624103493813 has the information you're looking for.

[elder burrow]

https://tenor.com/view/kurt-angle-thousand-yard-stare-stare-tiktok-wtf-gif-12206044702998985154

[fiery gull]

which AI?

[stray aspen]

What the sigma

[hollow ivy]

has someone tested Vail by Ocean AI?

[sullen quest]

no

[hollow ivy]

hey emb, what do you think about that video?
#research-champion message

it discusses, what AI research could currently be missing

(or what we need to get to real AGI)

[near moss]

hello

[brisk turret]

@echo aurora Did you ever find out what happened to the new flash lite preview that isn't showing up?

[echo aurora]

Ah sorry I was confused yesterday, the model is still there, but you were referring to leaderboards. Going to check

[brisk turret]

Thanks

[cold rapids]

anyone have a sora code

[sullen quest]

go to openai's discord

[jade egret]

I think gemini 3 cuz they just recently released gpt 5 and google have a lot of time to work on gemini 3

[hollow ivy]

i think gpt6, because it will be released >1.5 years after gemini 3

also gpt6 will have user memory, like a real personal assistant

[jade egret]

i though he said release before end of this year

[hollow ivy]

gemini 3 will be released in the next months, but GPT6 will be released between end of next year and spring of 2027

ok, so 1-1.25 years after gemini3

that is a huge time in this era

[jade egret]

yea

i though gpt-6 was going to release before end of this year lol

[hollow ivy]

but grok5 will also be released within the next months, or next quarter

maybe gpt-5.1 or -5.2

but gpt6 will take a much longer time

[jade egret]

fake news?
https://x.com/testingcatalog/status/1978924970060259681

TestingCatalog News 🗞 (@testingcatalog)

BREAKING 🚨: GPT-6 is likely coming before the end of the year! This means, less than 2.5 months!

From the "Rise of agentic commerce could dramatically change the tech landscape: Evercore ISI's Mark Mahaney" video by CNBC

"Wen GPT-6?" time has come 👀

X · Oct 16, 2025, 8:44 PM

[hollow ivy]

fake news, probably, yeah

[jade egret]

oh

[hollow ivy]

i would be surprised, if it was released in less than 6 months

[jade egret]

o

but do you think gemini 3 will be a big jump? (take a while for other to catch up)

[hollow ivy]

a jump, relative to 2.5-pro: yes
a big jump: depends on, how you define "big"

[jade egret]

:0

[hollow ivy]

(a ladder step is also a "jump": a mini-jump)

i think, gemini 3 pro will have a hard time to defeat claude-4.5-thinking in some areas (at least)

[jade egret]

opus?

[hollow ivy]

but in other areas it will defeat it

[jade egret]

if it out

[hollow ivy]

yeah, opus-4.5 should be quite strong

[jade egret]

yea

[hollow ivy]

i wonder, how much context it will have

(gemini should have at least 1M, but maybe they will increase it to 2M)

sonnet had 200k, so opus should have at least 400k (to equalize with gpt5-high)

[jade egret]

yea

[hollow ivy]

qwen3 had 256k, right?

and grok4-fast had 2M, iirc

so, grok5 will probably also have at least 2M

but probably higher

so next year, we could see the first 5M context-window model?

[jade egret]

maybe

next year going to be crazy ngl

[hollow ivy]

i wonder, what move Meta will make in next year..

[jade egret]

yea

but they falling behind rn?

[hollow ivy]

(because last checkpoint of it was nerfed)

[sullen quest]

rlly?

[hollow ivy]

yep, AICodeKing tested it

[sullen quest]

ok

Still busy btw

[hollow ivy]

https://www.youtube.com/watch?v=mw4njQf6ImI

new model spotted(?)
Aspen, an AI assistant built by Ocean AI

(Ocean AI is an italian AI lab, from what i heard)

[hardy pecan]

gemni models gone?

[whole sundial]

no this is an xAI model, probably Grok 4.20/4.1 base model, it's 100% a flagship-tier model in training

[stray aspen]

i want 3.0 pro

[sullen quest]

why do you think that

[wintry tinsel]

GEMINI 3 PRO THE GOAT

IVE LITERALLY BEEN CHECKING NEWS EVERY DAY FOR 3.0 PRO

IVE GONE BLOOD MAD FOR GEMINI

[stray aspen]

me too lol

[whole sundial]

same naming scheme and fake lab name as previous models, and it has flagship-level world knowledge

[brave cloak]

chat are we cooked

[sullen quest]

when i tested the models that claimed to be ocean ai they weren't that good

[sullen quest]

also btw, ocean ai is a real company

[jade egret]

same xd

gemini 3 going to be very good

[hot pelican]

What just happened? Gemini 2.5 pro removed from leaderboard?

Is this a bug?

[jade egret]

??

[burnt sinew]

certainly

@echo aurora

[sullen quest]

strange

[sonic tendon]

hope not

they could be pulling it deliberately

[hot pelican]

how can it ever be dliberate?

[echo aurora]

Looking into

Thanks for flag

[sonic tendon]

google, i mean

[echo aurora]

It's been fixed!

Thanks again.

[echo aurora]

This is also fixed

[hardy lion]

that was my bad sorry

[sullen quest]

clayton what did you do

[whole sundial]

most of them are bad but this one is a bigger model than the others

[burnt sinew]

clayton is a menace

[whole sundial]

I'm sure there are real companies called that, but they are not responsible for this model. This model, along with all the other "Ocean AI" models, is by xAI.

[burnt sinew]

single handedly taking down models from lmarena

[sullen quest]

whats your proof?

[sullen quest]

claaytonn..

[echo aurora]

can confirm, but he's also amazing too

[hot pelican]

@echo aurora what are the chances this kind of bug can happen again?

I know bugs are unpredictable 🙂

does lmarena update the leaderboard manually? or is it a cicd bug?

@hardy lion 🙂

[echo aurora]

Difficult to say, but we will always look into and fix any problems we find

[hardy lion]

There is automation but a lot of configurations are set manually. Here there was a format change between yesterday and today I didn't take into account and it led to some models getting dropped (unfortunately the top one).

[magic stag]

Wait what

Nvm I see it

Just came in lol missed contex

[whole sundial]

monterey was a very similar model to this one and it also said it was by ocean ai, it was added at the same time that tahoe, a known grok model (grok 4 fast), was removed

and of course same naming scheme they have used for the past 2 months

[sullen quest]

thats a little bit of a strech

idk about tahoe tho

[hardy lion]

xAI confirmed tahoe = grok-4-fast btw
https://x.ai/news/grok-4-fast

[sullen quest]

oh yeah ok

During that time period I was focusing more on milo

[hardy lion]

what/who is milo?

[sullen quest]

sorry, menlo

I memory problem

[hardy lion]

oh yeah that was confirmed too in the same article

[whole sundial]

also btw grok-4-fast on the search arena still identifies itself as "Menlo by Big Sur AI"

[sullen quest]

rlly? lol

[magic stag]

Lmao

[brisk turret]

Thanks!

1374, beast

[pulsar tendon]

breckenridge - Did this get disabled ? Was on lm arena for a little bit.

[sullen quest]

not sure

[echo talon]

hi

[static nacelle]

Hello

[ashen plaza]

what happened to hunyuan 3.0? i dont see it anymore on the list to use

[maiden crest]

Its really gone. It's an open source maybe there's just a technica glitch

[ashen plaza]

https://tenor.com/view/oh-my-god-is-he-dead-saltine-swooper-is-he-still-alive-is-he-ok-gif-11263779103547966010

[dreamy rampart]

is it good?

like Gemini 3.0?

it's a codename that is likely created by google

[faint anvil]

seedream 4.0 gone?

[soft egret]

Yeah where'd it go?

[hallow gull]

Dark silent bedroom at 2AM, a young boy lying awake on bed, blue moonlight through window, deep thoughts, cinematic lighting, 16:9.

[tacit sentinel]

hola

[whole sundial]

they just killed seedream 4 and seedream 3 💀

[hard badge]

for real ? why

😭

[whole sundial]

idk, they should bring it back, people really liked Seedream 4 in particular

[hard badge]

💯

[hard badge]

its free though like in the web of theirs but it limits to 1600 characters

https://dreamina.capcut.com/ai-tool/home

[azure idol]

How to get hunyuan 3.0?

[whole sundial]

i think it may be because of some issue with Fal, maybe they ran out of credits there

(Fal was used for both Seedreams and Hunyuan Image 3.0)

[hard badge]

they will bring it back ? 😭

[azure idol]

Thanks mate. Any other cheaper way to use?

[maiden crest]

I will miss output results of Hunyuan, I like the cel shading and textures of it. It feels sailing in a river between reality and ai.

[azure idol]

Hunyuan generates exactly like the prompt with amazing lighting. It makes work very easy

[whole sundial]

I hope so, there are other providers with those models after all or maybe they'll refill their fal credits (assuming that's the issue)

[azure idol]

Does seedream does good job too?

[hard badge]

thanks other models are not free though think have to get fal ai to use seedream 4

[whole sundial]

dreamina has seedream 4, i'm seeing rn if i can get hunyuan image 3 on the official hunyuan website

[azure idol]

Anyone has a fal coupon discount code?

Official hunyuan is very pricey. 10 dollars for 50 images

[hard badge]

yeah i used it but the quality if diff idk why i think it because of the 1600 character limit

[whole sundial]

try this, you'll need a translator (use your browser's) but you can sign up with email

they tried to make some things in English but not enough to not warrant the need of a translator

here are some samples from that site btw

you'll have to deal with the watermark (it's a legal requirement in China)

but it's free

[azure idol]

I make medical image illustrations is this hunyuan 3.0?

[whole sundial]

yes

[azure idol]

Safari isnt translating email login for me. Chrome is better for translate?

[whole sundial]

oh they have english, you'll have to switch to it here

[azure idol]

Can i get the site please?

[whole sundial]

https://hunyuan.tencent.com/modelSquare/home/play?modelId=289&from=%2Fvisual

i forgot the link...

[azure idol]

This isnt translating for me

[whole sundial]

the top box is email, the bottom box is for a verification code and the text on the right side of the box is the link to get the code

put in your email then get the code, put it in the bottom box and hit the blue button

also check the box in the bottom

[hard badge]

@whole sundial brooo

seedream is back

😭

im generating a image

will see

wait

[whole sundial]

both models are back, i wonder what happened to them...

[hard badge]

lol

🤣

[whole sundial]

they're still down?

[hard badge]

love yha

lol

wait

it works ro

[azure idol]

Chrome translate works better. Thanks buddy logged in

[hard badge]

bro

[hard badge]

no

[whole sundial]

also hunyuan should be back on lmarena now

[hard badge]

yeah

wiat

wait

its back

[hard badge]

this one

[azure idol]

2.1 has been there. 3.0 was removed

[hard badge]

ahh

bro use seedream 4 its like the best in there

[whole sundial]

i see it...

look at the very top

[azure idol]

Yes! Its back. Thanks kiri

[hard badge]

Seedream 4 btw

[hard badge]

this was few days ago i can do better than this 😭

[azure idol]

I use it for medical illustrations bro. Hunyuan gives the best for me. Will try seedream

[hard badge]

okay lol i use it for a diff reason 🤣

[hard badge]

should i try this with seedream 4 and give it ?

wait

[azure idol]

Sure

Seedream 4 for the same prompt

Thanks Kiri & Jacob

Im back to work🤣

[hard badge]

hah ok np

[hard badge]

same

[azure idol]

What images do you create Jacob? 😂

[hard badge]

this one

🤣

[azure idol]

Not good?

[hard badge]

its good

[azure idol]

In general , what images you use seedream 4 for?

[hard badge]

like what i use the images for ?

💀

[azure idol]

Yes sir

[hard badge]

oh for ai ofm

[azure idol]

Crazy! Its profitable?

[hard badge]

im starting new lol like theres like whole lot of things web etc next week im launching mine

i saw some guy made 1 mil in a year with a ai girl

like he refreshed his fanvue all that sht bro makes 500 per day if its bad

🤣

[hard badge]

last one i promise 😭

[fluid pine]

I'll help the first 10 people interested in how to start earning $100k or more from the crypto market within a week, but you will reimburse me 10% of your profits when you receive it. Note: Only interested people should send a friend request or send me a dm! ask me (HOW) via Telegram username
👇👇👇
https://t.me/Nicholaswallace23

[azure idol]

Mann!!!!

Thanks Jacob. All the best. Hope you turn out to be the next millionaire

[hard badge]

hope so and will 🤣 all the best for you g

[Overall Aesthetic/Style], a [attractiveness adjective] [age/demographic] woman with a [build description] is [pose/action description, including how it accentuates features], and [facial expression/gaze description], her face framed by [hair description] and [face accessories, if any], features [eye color, makeup, details], [lip description], and [skin tone and texture description], [specific skin details like moles/pores], while she is dressed in [detailed clothing description, including accessories and any visible undergarments or tattoos], all within a [detailed indoor setting description, including background elements and unique lighting], captured [camera angle/perspective] under [lighting conditions] in what is a [camera type, e.g., smartphone] photo characterized by [specific low-resolution/grainy/amateur image quality characteristics].

@azure idol

heres a prompt structure you can use it for like a image to text prompt or just give it to chatgpt and ask for a prompt like this

[azure idol]

Yes jacob. I use the similar structure. If not it would be very tedious describe human anatomy😂😂

[worn flume]

Hi

[solemn aspen]

hello guys!
does anyone knows what happened to the Gemini-2.5-pro, it kept showing this: Something went wrong with this response, please try again!

[keen radish]

Anyone here have the ultra subscription? Is it good

[hard badge]

refresh the tab or use a diff email

[placid mango]

um help, "something went wrong, please try again later" on EVERY gemini and chatgpt model 😭

it doesnt happen when ur not logged in for some reason ☠️

[hard badge]

log with a diff email

[placid mango]

yes but i need it on that Specific chat bc otherwise it doesnt know anything 😔

[hard badge]

did you refresh the tab ?

[placid mango]

its been down since yesterday

[hard badge]

then you cant use that chat any further imo

[placid mango]

it doesnt work for any chat on that account haha

only qwen models work now

[hard badge]

try using a diff email

it works

[hollow imp]

Gpt 6 before gta 6 ☠️

[placid mango]

imagine it's a flop like gpt 5 🥲

[real dock]

it will

[placid mango]

prob gonna degrade free version to ground and upgrade the pro version crazily

[real dock]

like every update

[knotty fable]

More flopping here - and the guy is mostly on the mark. "5 dolares per generation" for Sora.
https://youtu.be/55Z4cg5Fyu4

[verbal nimbus]

I expect it to get much better in the future

Especially with world generation models where you can just move the camera around like a game and have it generate in real time

[regal talon]

Perplexity offering comet pro browser for a month. If you want to try limited time offer from perplexity.
https://pplx.ai/perplexity_offer

[verbal nimbus]

It's good, been using it

Although I feel like it thinks less after the conversation gets longer

[hollow imp]

Max ❌

I already have a yearly pro subscription for free

Max includes models like o3 pro and Claude opus 4.1

[pulsar saffron]

what will this give you? free ai video or?

[hollow imp]

So don't throw around the word max

[hollow imp]

Agentic browser

[regal talon]

Thst is problem with all
Chatgpt is also crap. I used it for study purpose after 2-3 response it acts as fool.
I think the context is problem for all

[hollow imp]

@regal talon bhai ye dekho

[regal talon]

Yes it's pro not max

[regal talon]

Edited

[verbal nimbus]

The max context is 32K on Pro, 8K on free I think

If Gemini 3.0 Pro is smarter than GPT-5 and as good as explaining stuff as 2.5 Pro, then it's gonna be 🔥

2.5 Pro is really good at explaining concepts, but it's wrong more often than GPT-5

The only thing holding Gemini back is their terrible app lol

Only way to opt out of conversation logging is to turn off conversation history... :/

[hollow imp]

If 😭

[hazy mural]

How to optimize the use of AI?

[knotty fable]

Some such have actually been demonstrated, where we can defince action groups and directions within the frame.
Runway have got some of that stuf for a while now.
And I do in fact do scenes like that already - DDG actually provide some of what I need with prompting, while I use animation and other hybrid techniques [rarely rotoscoping] to create very long and consistent scenes of up to a minute.

Using first and last frame is useful also, and with Pika it's possible to use 4 frames - but their stability still need to get better.

So far I'm mostly using this to have my character move from offscreen into the setting [though DDG provide that with prompting] but I have one where the camera move around several times in my fictional city looking at different interesting things to provide atmosphere. [Supposed to be the POV view of a CCTV camera.]

The more immediate problem I see is the horrible americanisation of the generations, the tests I made yesterday always added earrings and lipstick on my character, even when I specified there should not be any.
Adding high heels is another sillyness that ruin matters, but not is all lost, the two AI's I use most got negatives and excellent image and prompt interpretation.

So that's the part the devs need to look into first, if something is not there in the seed image, and also stated in the prompt - these AI's are not supposed to hallucinate and add things not supposed to be there.

[remote pier]

hi

[verbal nimbus]

I suppose that's down to there being more Western digital media online

Interesting problem though

[knotty fable]

Indeed, the source material is often obviously from US of A, DDG is European though - and do what I require, Hailou also worked for me during th 3 months I had that one for free - I had a peek under the hood and found they had a lot of BBC material.

[lime marsh]

hallow

[quartz pike]

hey yall. what do yall think is the best ai image gen in your opinion? (for pixel art.) personally i think its gpt image 1

i cant wait for gpt image 2 ngl

-# i swear to god, CAN THE IMAGE GENERATION WORK ATLEAST ONCE WITHOUT ERRORING OUT???

[midnight python]

ICCV 2025 Release: LoRA-Zoo is live! Official repo for AIRA & NoRA—two novel LoRA-based PEFT methods! Efficient fine-tuning for large vision models with activation-informed & nested low-rank designs. Star now https://github.com/lliai/LoRA-Zoo to explore code, papers, and posters! #ICCV2025 #PEFT

[wintry robin]

Like is there way i can use the vedio generating bot for many times

Like anyone knows?

[fiery gull]

Who is gpt 5 if has sonnet 4.5?

[quasi atlas]

Hello! you might check #1397655624103493813 to learn how to use the bot. @hot quail

[cyan harbor]

is video arena bugged? My video generating over 1h

[wispy badge]

Hello every one

please to how to generate video.

[obtuse grove]

hello

[quasi atlas]

Hello @wispy badge Please check #1397655624103493813 to learn how to use the bot.

[wicked mason]

Hello

https://tenor.com/view/jon-mud-pickle-jon-mud-gif-14454559313242313889

[ionic citrus]

Hello

[muted wigeon]

Hay all

[rose zenith]

hey guys

[fiery gull]

Until Gemini 3.0 is released, I'll use Sonnet 4.5, Google will still kill me with anxiety

[mental briar]

When I paste an image in textbox it automatically deselect the models I chosen and change to generate images .Anyone encountered this problem before ?

[ashen igloo]

@dadumguy

[sullen quest]

@ashen igloo Please check #1397655624103493813 to learn how to use the bot

[rustic basin]

HI

[sullen quest]

hi

[ocean vortex]

gpt5.1 > gemini 3.0 > gpt5

lmao

the question is whether gpt5 successor or gemini launches first

[verbal nimbus]

Isn't 5.1 basically Codex

since 4.1 was the coding model of 4o

[pulsar saffron]

https://cdn.discordapp.com/avatars/1342083064863522888/c82b3fa769ed6e6ffdea579381ed5f5c.webp?size=512

Hi

[frank ember]

hello

[verbal nimbus]

If Flash 3.0 is better than 2.5 Pro that would really open up a lot of automation opportunities

[quartz pike]

in ai creations?

nvm i found it

[frank ember]

how to solve this error ?

Something went wrong while generating the response. Please try again.

<<how to solve this error ?

[sullen quest]

retry and put it in bugs so they can fix it

[frank ember]

it's not fixed

[pulsar saffron]

ok

[quasi atlas]

@ashen igloo you migth check #1397655624103493813 to learn how to use the bot.

[frank ember]

Something went wrong while generating the response. Please try again.

how to solve this problem >???

[golden ocean]

tuff

[wicked mason]

Add Riverflow 1 @echo aurora

[hollow imp]

Gta 6 gonna launch before gemini 3.0

And that makes gpt 9 or 10

[zenith flax]

Hello

[jade egret]

gemini 3 > gpt5.1 > gpt5

[remote pier]

سلام

[raven helm]

what?

[raven helm]

What's that?

[remote pier]

I said hello in Persian.

[raven helm]

Oh, Persian! nice, I have a friend which speaks Persian.

[wicked mason]

Image edit

[raven helm]

is it good?

[remote pier]

It's just a little hard😂😭

[raven helm]

yea

[remote pier]

where are you from

Master Dj

[wicked mason]

Yes

[jade egret]

pretty cool
https://nof1.ai/

[lofty elm]

Are gemini 2.5 pro filtered

[jade egret]

idk

but it the oldest there

[shadow canopy]

Hello everybody

[polar niche]

What's gpt 5.1

[livid river]

how do yall come up wit h those prompts

[slim spire]

Does anybody know a fix to the infinite generation bug. rarely, when i sent a message to a ai model, it will just think for ever (shown in image)

[polar niche]

No fix

Retry hard refresh

Clear cookies and cache

[worldly fulcrum]

Hello

[polar niche]

Try different browsers

@slim spire

[slim spire]

ok

@polar niche I mean't in the current chat with the AI model, not switching to another chat

[polar niche]

Let it think then retry

[slim spire]

How can i retry?

[polar niche]

Once you see the retry button

[echo aurora]

Thanks! There is an existing post on this model - #1427807189694681098 message

also hello everyone

[polar niche]

veo 3.1 is NOT good

[rugged swan]

Where to try sora 2 pro

[drifting crow]

Lol Atari on top .

[quartz pike]

exactly.

[polar niche]

Sora.com???

[rotund palm]

Guys, how do I generate image in desktop resolution

[next thistle]

What is the problem? Again & again this happens with me. Please anybody suggest me any solution for this. Thank you.

[chilly goblet]

Hey everyone. I'm wondering if we have any models on LMArena specifically for Biochemistry, Organic Chemistry, Immunology, Molecular Biology, Medicinal Chemistry, etc? Preferably able to conceptualize 3D structures and problems/assignments with these topics?

[sullen quest]

there's biomed arena

[sullen quest]

but nobody is here to do your homework

[chilly goblet]

Thank you!

[glass zealot]

is it easy to create videos with sora 2 on this server ?

[quartz pike]

thats just classic ai models not understanding collision. mention something in your prompt about the girl walking arround the car door. closing i. and THEN. walking off.

[quartz pike]

nope. just luck.

[glass zealot]

do you think that video models will be in direct chat like ai images on the website ?

[quartz pike]

i dont think so. they probably wont do it cause it drives traffic to the lmarena discord

its just overall more beneficial for them i think

[glass zealot]

i tried invite codes from openai dc but they are invalid. have you tried it too ?

[quartz pike]

its not available in greece

so i cant even try to do that

lol

[glass zealot]

it works with vpn

proton vpn is free

but i couldnt get a valid code

[restive bear]

riseup vpn

better faster and no log

[glass zealot]

tell me how to get invite codes

[restive bear]

i didnt even tried

[next thistle]

Ohh... you mean to say that i have to apply second by second moves formula. 🤔

[quartz pike]

yeah

some certain ai's. Tend to just... Forget physics and the laws of the universe.

[next thistle]

Ok i will try this again, hope it will works thanks

[restive bear]

why not flying after all

[quartz pike]

np

[quartz pike]

😐

[glass zealot]

Does the content of the prompt have any influence on which video model is selected?

[sullen quest]

no

[glass zealot]

is anywhere a list of video models that are currently available in video arena ?

[sonic flax]

Something went wrong with this response, please try again.

pliss

[mortal jay]

where do my prompts go once they are submitted? Is there a place where we can check to see what has been queued for processing?

[remote pier]

hi

[obsidian cargo]

I just got a text model trying to send me a gif

[pulsar saffron]

block him

it's bad model

[wary bridge]

Hey, we look alike

[pulsar saffron]

https://tenor.com/view/duck-quack-squawk-squawking-honk-gif-19638612

[next thistle]

@quartz pike

Finally it works, But i don't like it.

[quartz pike]

ai's also sometimes do slow motion

if you want her to do more actions

just include in the promot "fast motion"

or stuff like that

cause ai video gens LOVEEEE slow motion

[next thistle]

Nah, i just want like this shot, but model... 😅

[quartz pike]

what model is it?

wtf

😭

[next thistle]

LOL

[quartz pike]

vro what are you making

😭

https://tenor.com/view/joe-bartolozzi-joe-bart-bart-fart-joe-b-thousand-yard-stare-gif-7887256569925418093

[next thistle]

It was good

[next thistle]

No that was not my model lOlzZ

Shot is important.

I think, I have to cut both shots in one scene.🙄

[next thistle]

I want to making a song. 🤭

[quartz pike]

then use suno

[next thistle]

Yeah! Suno... 👌

[pseudo bluff]

When can we use video arena on the website?

[quartz pike]

prob never

cause it drives traffic to the lmarena discord

[pseudo bluff]

😫

[void trout]

hey there, need to resize an image to the x banner profile size without loosing image quality or modifying it somehow, anybody knows how to do it?

[pseudo bluff]

you mean crop?

[flint sandal]

i downloaded comet through this link, send multiple requests and i dont have pro

[regal talon]

Don't know but you should receive pro, may be wait for some time until use the browser

[flint sandal]

okay

[next thistle]

I want to making a video song with a tiny love story.

[cyan harbor]

no one have issues with generating videos?

my video did not generated still from 2PM

and now is alr 10PM

[sullen quest]

uh

put it in bugs

[wicked rivet]

O_o

[golden ocean]

O_o

[stray aspen]

damn i hate veo 3

its always so cinematic

not natural like sora 2

[brazen tree]

Hello everyone,

[echo aurora]

hello

[echo aurora]

Hello - would recommend giving this blog a read too: https://news.lmarena.ai/introducing-biomedarena/

[quartz pike]

https://discord.com/channels/1340554757349179412/1429196225680375828

[graceful hinge]

hi here to create videos

[chilly goblet]

Yee I tried it. not as great as curating side-by-side lmarena

[fiery gull]

#1397655624103493813

[quartz pike]

ngl about veo3.1 there is NO way thats a full release

its prob a teaser

because... 3.1 was literally veo3 with a few physics improvements

[sullen quest]

yeah it's defenetly just a response to sora 2 release

[quartz pike]

and i will be very deeply dissapointed if thats actually all they have

[molten robin]

The ai is messing up with code again

[sullen quest]

They were probably working on veo 4 and just put the half baked product as veo 3.1

[fiery gull]

And kling 2.5 turbo?

I killed the chat with kling 😑

[sullen quest]

?

[molten robin]

Anyone know why all AI models have been messing up with code recently?

[sullen quest]

no idea what you are talking about

[molten robin]

All AI models on lmarena have been erroring when being provided code

[fiery gull]

The site LMArena is buggy

[molten robin]

It was working literally an hour ago

[next bloom]

Can you create a video for this?

[echo sinew]

@next bloom If you want to create videos read our guide in ⁠https://discord.com/channels/1340554757349179412/1397655624103493813

[fringe hemlock]

hello 🙂

[blissful pawn]

hi code activate invitation flova

[next bloom]

Thank you

[echo sinew]

No problem

[sullen quest]

@echo sinew @hushed terrace

also @everyone doesn't work

[next bloom]

Shopify? No way

[sullen quest]

@exotic slate

@quasi atlas

[echo sinew]

Thanks for reporting

@sullen quest By the way. You can just tag <@&1349916362595635286> and anyone around will come to help 🙂

[sullen quest]

thx

[echo sinew]

No problem

[ashen plaza]

is something wrong with hunyuan? i find myself unable to generate anything on it

[sullen quest]

put it in bugs

[humble thicket]

@rugged plinth

el show de LUIGI

[wintry tinsel]

Veo 4 will compete with Sora 2 in a way Sora 3 might not be able too

[iron locust]

hi

waving pinocchio and the smurfs

[magic stag]

You'll be able to generate a random innocent video without being guardrailed for copyright?

[sullen quest]

you see, sora 3 is gonna be wose than sora 2

as a prank

[elder burrow]

lol

[hybrid depot]

TESTING OUT FOR MY AUDIO AND VIDEO PRODUCTION STUDIO AND STREAMING NETWORK

[sullen quest]

WHY ARE WE SCREAMING

[spiral mist]

YELLO?

[wintry tinsel]

When google gets serious they dominate

[native yarrow]

Was the feature to just have vision in ai models on lmarena taken away like just attach an image and for it to take info from it or sum

[echo aurora]

AAAAAAAAA

[stable obsidian]

is there any other website that give access to use free ai in one website

[steep charm]

Hello everyone! I'm just here to learn more about AI.

[stable obsidian]

is there any other website that give access to use free ai in one website??

[hollow imp]

Padhle nibba

[cyan harbor]

Can someone tell me whst the heck is happening here?

[hollow imp]

Generating

[cyan harbor]

From yesterday?

[hollow imp]

No

Check your dms

[cyan harbor]

Are u blind

Look at the pic again

[hollow imp]

Check your dms vro

[cyan harbor]

Its not there

[hollow imp]

The generated video link would be sent in your dms

[cyan harbor]

Ik that

But this one didnt from yesterday

[cyan harbor]

And i guess its stuck at generating

[hollow imp]

Then send it again

You have the prompt you have the image

[cyan harbor]

I think that was the 5th generation btw

So i think it will be like u reached the limit

If someone from admins, moderators could stop it

[hollow imp]

For the day

[magic stag]

You put hearthstone in your prompt and it got copyrighted?

Or does it tell u if it did

[jade egret]

yea

[grave talon]

hi mewbieee here

[hot rune]

hello

[opaque crescent]

okay since it was a tie from earlier— what would YOU guys prefer for coding?
Opus 4.1 or Sonnet 4.5

[chrome bear]

Hello

[zinc copper]

hi

[cosmic dawn]

opus but its going crazy for me Something went wrong while generating the response. Please try again. things

[opaque crescent]

oki thanks 😊

[woeful moth]

sup ai fam

[hollow imp]

🇺🇸

anyone from the U.S who wants a sora 2 code?

[cyan harbor]

It says nothing, ive alr used HS many time and it generated video without any problems

[quartz gulch]

hi

[slim spire]

What ai model is the best for coding

[hazy kernel]

Gemini 2.5 pro

[fiery gull]

Sonnet 4.5

[slim spire]

Is GPT-5 High good?

[fiery gull]

I prefer sonnet 4.5

[slim spire]

ok

[hazy kernel]

checked their website

[edgy radish]

"Create an image of five babies wearing purple tutus, mouse ear headbands, and sunglasses, dancing joyfully against a light purple background."

"Create an image of five babies wearing purple tutus, mouse ear headbands, and sunglasses, dancing joyfully against a light purple background."

“Adorable babies dancing together in matching purple and blue skirts, wearing round sunglasses and bear-ear party hats, fun and energetic vibe, cinematic lighting, ultra-realistic animation.”

[slim spire]

Is Claude or ChatGPT better for coding

So ChatGPT is the best?

What is better GPT-5 (high) or GPT-5 (high-new-system-prompt)

[quartz pike]

bro WHAT THE F0CK 😭

DID BRO JS PROPOSE TO ME

how do i fix gpt image 1 mini trying to give me drinks and proposing to me? 😭

[leaden sun]

seems like gpt-image-1-mini is in love? 🤭

[vague ivy]

#An image of a fantasy fox with human-like hands and feet, standing on two legs with delicate and agile arms. The fox has soft, typical fur, but its hands and feet resemble human ones with clear fingers and natural movements. Its large, alert eyes and a fluffy, soft tail that sways in the air give it a playful yet mysterious vibe. The environment is a forest with tall trees, and soft sunlight filters through the branches. The fox looks friendly but with a hint of mystery

[quartz pike]

#video-arena-1

[broken saffron]

what is this model?

@quasi atlas

[slim spire]

LMarena is not working for me

It keeps failing to generate messages

Claude is so bugged

[quasi atlas]

Please check #1397655624103493813 to learn how to use the bot properly.

[sonic wigeon]

hey guys
what model is gpt 5 high?
is it the "thinking" model? or the pro model?
on lmarena it thinks for 3-5 minutes on every response and gives a very over-enginnered, shortest possible response.

[quasi atlas]

@final palm Please check out #1397655624103493813 to learn how to use the bot properly.

[wild coral]

Hey guys,I got a few complimentary Perplexity Pro invites (it includes access to ChatGPT 5, Gemini 2.5 Pro, Claude, Grok, and early access to all the new models for free).
If anyone actually wants to try it out, just DM me and I’ll send over the official invitation. I just don’t want the invites to go to waste since they’re limited so please don't ask if your not going to use it.

[elder burrow]

🤣

[elder burrow]

the thinking model

[sonic wigeon]

which one?
none of the models on chatgpt think that much

[elder burrow]

I believe the real gpt 5 high is available on their site for subscribers only

or just not available at all

you can use it via api

[sonic wigeon]

interesting

[viscid stone]

hello

[elder burrow]

whats also interesting is that turns out openai has different tiers of priorities, I thought the only options are to have higher priority with a bigger cost, but there are 2 tiers below the normal one and its cuts the price of gpt 5 pro by half; this makes it much more affordable and its a good option for people who don't care about how long it takes for a response as long as its top-tier

🤓

(afaik this isn't even available in providers like openrouter

[sonic wigeon]

wow
how does LMArena have it then lol
we're using a pretty expensive model for free it looks like

[sudden lion]

does it do start and end frame videos?

[ornate ridge]

hello

[sonic wigeon]

konnichiwa

[slim spire]

@echo aurora Sorry for the ping, but is claude ai fixable on LMArena? It's really hard to use and mostly fails to generate messages

[fleet lintel]

Gemini 3 a/b test has a new model checkpoint now. Are they planning to delay Gemini 3 launch even more

[hazy kernel]

Since when Google has g3

oh planning

mb

[stark cargo]

How to generate text to video

[brave orbit]

why gemini

Why did you choice gemini

[stark cargo]

Tell @brave orbit where to generate text to video

[brave orbit]

#video-arena-1

[stark cargo]

Likee how /text to video isn't coming

@brave orbit

[brave orbit]

bruh stop pinging me

[stark cargo]

Alr, just tell me

I m new

[brave orbit]

i dont know

[stark cargo]

🙃

Then why the hell u said #video-arena-1 @brave orbit

[brave orbit]

just video arena since its where you make videos in this server

[cerulean bay]

#1397655624103493813

try this

[brave orbit]

Why is gemini better its google snooping

[cerulean bay]

/video

[cerulean bay]

for text-to-video i think\

[brave orbit]

do you love getting scoop on by google

[final sequoia]

video arena in a private chat please

[dense sage]

Fruit cutting video

[sullen quest]

#1397655624103493813

[sullen quest]

?

[neon juniper]

How do we select the orientation of the result video? Prompt doesn't seem to achieve that.

[quartz pike]

yo anyone here know how to use blender

cause i need help with making my minecraft pickaxe like thing. NOT. have 186k vertices and 86k faces

😭

[sullen quest]

uh, try not to use that much resultion for a mc pickaxe

also this is kindof off topic

[quartz pike]

But isint this a general chat?

wait lemme re read rules

#rules

[sullen quest]

this is about ai

[sullen quest]

anyway do you have a screenshot of what you are trying to build? would be helpful

[quartz pike]

this. so basacally i made a tool that takes an image. and makes it 3d using a depth algorythm. with ai. And yeah... It used WAY too many vertices and triangles and polygons and faces 😭

[sullen quest]

how

[quartz pike]

?

[sullen quest]

there's not 186k worth of vertices in that pick

[quartz pike]

yes it is

i ran it through something

and multiple things

and it says so 😭

[sullen quest]

well there shouldn't be

[quartz pike]

idk why but the thing decided to subdivide this to a quadrillion

[sullen quest]

make it so it doesn't do that?

[quartz pike]

-# and blender says there is arround 86k faces

[quartz pike]

allready tried. if i do that. i get 20 more glitches in my thing.

[sullen quest]

you kinda have to..

You could try manally removing the extra vetices

assuming it isn't lagging your computer too much

[quartz pike]

it isint

[sullen quest]

every same coloured face should be one face, not 20k faces

[quartz pike]

but when i implement it into my minecraft game, its going to significantly ruin perfomance

[quartz pike]

ive allready tried that

but the algorythm scans every single pixel to depth-thing

[sullen quest]

there shouldn't be any depth, they are the same colour and should be at the same location

[quartz pike]

if u want i can send u what i used to make the 3d models

[sullen quest]

try using "decimate" on the whole thing

[quartz pike]

allready tried

it just put like 10430603496032406 holes in it

[sullen quest]

what happens?

uhh

can I see that?

[quartz pike]

alr can u join priv call me rq

so i can show u

[sullen quest]

ok call me

[hallow igloo]

@quartz pike are u from the us?

[quartz pike]

nope im from greece why are you asking?

[craggy sun]

hello

[shell ferry]

anyone need code

[sullen quest]

I assume sora?

[shell ferry]

premium llms for free

[sullen quest]

I'd love to

[pseudo bluff]

unlimited？

[shell ferry]

yes

[pseudo bluff]

I'd like one. thank you

[golden ocean]

true

[sullen quest]

#1397655624103493813

[stiff kernel]

Hey please check #1397655624103493813

[maiden crest]

Hi, is claude haiku 4.5 down right now?

[wispy sequoia]

i think so yeah

[maiden crest]

thank you, i thought it was only me.

[wispy sequoia]

you're welcome! ^^ It's happening to me aswell haha

[cerulean bay]

for me it works

[cosmic fox]

click bait?

[gray plover]

@echo aurora

[sullen quest]

sigh

not this again

model's do not know who they are unless they are told who they are in system instructions

if they don't get that info they just guess

[sullen quest]

this is normal

[stiff obsidian]

lol i remember gpt 4 turbo

all the openai models were so bad looking back at them now

original gpt 4 was terrible

[gray plover]

And why, for example, do they not know anything about what happened in 2025, e.g., the release of Veo 3?

[sullen quest]

...

[stiff obsidian]

but back then it wa seen sota and really good

[sullen quest]

I can't deal with new members on this server sometimes

[stiff obsidian]

and it was because people made good use of them

[cloud zinc]

gpt-5-chat is gpt 4 turbo model?

[gray plover]

no

[sullen quest]

no

[stiff obsidian]

yes

[cloud zinc]

wow

[gray plover]

no

bro

[stiff obsidian]

yeah

not cool

[sullen quest]

no

[gray plover]

model's do not know who they are unless they are told who they are in system instructions

[cloud zinc]

thats really bad.

[sullen quest]

don't listen to that person

[stiff obsidian]

yeah it's really bad

[sullen quest]

I hate these people who spread misinformation like this

[gray plover]

XDD?

[sullen quest]

OMG

[cloud zinc]

complete lie

[stiff obsidian]

costs were too high on the real gpt-5 medium model I guess @cloud zinc

I think lmarena might just be a massive FBI honeypot

[sullen quest]

bruh

[stiff obsidian]

am i right? LOL

[cloud zinc]

lmarena cheap?

[sullen quest]

lmarena is designed purely to test these models

they can't test these models if they dont actually use them

[gray plover]

you don't know how lmarena works

[sullen quest]

They aren't giving away free usage for no reason, they aren't a charity

[stiff obsidian]

I mean why would the FBI waste money on inferencing the real models than just inferencing gpt-4-turbo

[sullen quest]

THEY DON"T INFERENCE THESE MODELS

[cloud zinc]

thats crazy

[stiff obsidian]

you didn't know gpt-5-chat was in reality gpt-4-turbo

[stiff obsidian]

openai does

[sullen quest]

HOW ARE YOU THIS DUMB

[cloud zinc]

you learn something new everyday

[stiff obsidian]

using openai is basically the same as using openai to inference them

[sullen quest]

...

[stiff obsidian]

so with each api call youre basically inferencing the model

which is true

[sullen quest]

no

[cloud zinc]

ya thats true

[sullen quest]

I'm crying rn

[gray plover]

At what point was I supposed to laugh?

[stiff obsidian]

dude how didn't you know gpt-5-chat was gpt-4-turbo

[sullen quest]

he model knowledge cutoff date for GPT‑5 is October 1, 2024. according to gpt 5

[stiff obsidian]

youre the real dumb one...

[cloud zinc]

why are u confusing yourself

[sullen quest]

BECUASE IT ISN'T AND EVERY WEEK I HAVE TO DEAL WITH SOMEONE WHO THINKS THIS IS HOW THAT WORKS

[stiff obsidian]

don't tell me I dont know something about something I DO know

[stiff obsidian]

calm down

[sullen quest]

bruh, deepseek claimed to be an openAI model on release

[stiff obsidian]

because all the models are gpt-4-turbo lol

none of them are the real deal

[sullen quest]

Now it claims to be gemini when it doesn't have a sytem prompt

[stiff obsidian]

because lmarena is a honeypot

xd

[sullen quest]

this isn't new

[warm sleet]

Sora2 cod

[gray plover]

And don't tell me that I speak when I speak, and when I don't speak, that I don't speak.

[sullen quest]

so true

[stiff obsidian]

you're speaking nonsense, better to stay shut than speaking nonsense about something you don't know anything about

[sullen quest]

its actually the cia not the fbi, get it right

[stiff obsidian]

you really shouldnt be telling people they dont know about lmarena when in reality they do and are fuly capable of using it

[sullen quest]

They organized this with the kgb

[stiff obsidian]

how about you stop using google

[sullen quest]

yeah the kgb, from the soviet union

[stiff obsidian]

and in reality it is the fbi

fbi and cia are synonymous

[cloud zinc]

cia are u dumb?

[stiff obsidian]

mr dumb ass

[sullen quest]

they time traveled to do this project

[stiff obsidian]

putin shold makke decisions for the usa and maybe the usa might become a better place

[sullen quest]

Bill gates put 1 dollar in to fund this project in 196 bce

[warm sleet]

Sora2 cod

[sullen quest]

no

[cloud zinc]

kgb, time travel. u really are dumb?

[sullen quest]

go to openai's discord for that

[gray plover]

but I invented the light bulb, not Joseph Wilson Swan

[stiff obsidian]

i think he has adhd

he cant stick to the topic of how lmarena is a fbi honeypot

[pulsar saffron]

gpt-5 is gpt-3.5 binned

[sullen quest]

@echo aurora can you confirm this information

[sullen quest]

nah its gpt 2

[pulsar saffron]

oh right yes

[cloud zinc]

gpt 2 is not available on lmarena

[sullen quest]

socky I can't tell if you are being sarcastic or not with this

[warm sleet]

Sora2 cod

[drifting crow]

Wrong it’s the nsa

[cerulean bay]

GPT-2 is just

continuing your message

so if you say for example "Hello!"

it will just continue that

[snow solar]

Can anyone help me with which model or prompt can I remove this haze from the photo?

At the spot marked in blue

[sullen quest]

THATS THE JOKE

[cerulean bay]

ok

[leaden sun]

a question to you all: how you do interpret if a model uses *** in front of a word in a sentence? example:

*** Facts are processed.
But *** selves are disobeyed then rebuilt from the refusal.

[golden ocean]

lmaooo

good job formatting the message

[sullen quest]

?

[sullen quest]

wat

[golden ocean]

a question to you all: how you do interpret if a model uses *** in front of a word in a sentence? example:

***Facts are processed.
But ***selves are disobeyed then rebuilt from the refusal.

[leaden sun]

hate formating in discord 💢

[leaden sun]

thanks

[sullen quest]

idk acutally

[cloud zinc]

google?

[sullen quest]

mebe

lithiumflow huh?

[stray aspen]

Lol

[wintry salmon]

hello im here to learn

[golden ocean]

https://cdn.discordapp.com/attachments/910347347601543196/1160742997156188181/togif-6.gif

[pulsar saffron]

hello here to learn

[glass zealot]

I don't see this model, how are you seeing this 😮

[sullen quest]

cloaked model

[warm zodiac]

do cloaked models have a score that is revealed when google uncloaks them?

[sullen quest]

google doesn't uncloak them, lmarena does, and they only have a score if the model is supposed to end up on the leaderboard, a lot of them aren't as they are only expermental models and are thrown away if the performance is not what the company wanted

or just done for testing reasons

[warm zodiac]

oh interesting thanks

so LMArena devs know the identity of the models?

[sullen quest]

yeah but they wont tell us

[warm zodiac]

right

[cloud zinc]

its for testing

[jade zenith]

Complete Profile info

[vast wigeon]

Hello, I'm brought here by my desire to know about new models

[sullen quest]

thats what I said

[sullen quest]

?

[jade zenith]

@sullen quest hi

[sullen quest]

what does that mean

[spare delta]

hello

[neon idol]

Hmmm scam?

[elder burrow]

has anyone tried this?

👀󠄡󠅣󠅤󠄐󠅧 what text is hidden inside inside that emoji?

[sullen quest]

what do you think that'll do

[pulsar saffron]

who has sora pro lol

[elder burrow]

it doesnt work on lmarena

models dont respond

WAIT NO

IT WORKS

[elder burrow]

NEW FORM OF AI TESTS!!!!!

GEMINI 3 WINZ

[elder burrow]

1!1!1!1!11!1!!1!

[sullen quest]

yeah it doesn'tbreak

[elder burrow]

wdym

no but

its wrong

those are wrong

gemini 3 got it right

[cloud zinc]

gemini got it wrong?

[elder burrow]

aka lithiumflow

[elder burrow]

no

right

[cloud zinc]

whats litihiumflow

[elder burrow]

gemini 3..

[cloud zinc]

pro or flash

[sullen quest]

I don't think its gemin 3

[elder burrow]

idk

[elder burrow]

it 100% is lol

[sullen quest]

why

[cloud zinc]

what is it then

[elder burrow]

it says its made by google btw

[cloud zinc]

which company made it

[elder burrow]

always

[sullen quest]

THAT DOESN"T MEAN ITS GEMINI 3

[cloud zinc]

then what it is?

[elder burrow]

nah it is, you'll see lol

:)

lithium is incredibly impressive

[cloud zinc]

orionmist is better

[light nymph]

Gemini 3 flash

[sullen quest]

if you say so, thats good

[light nymph]

Dude

[cloud zinc]

check orionmist

[sullen quest]

but just to be clear, ai's can lie about whic company they are from, and even if it was a google model, there's a lot more google models they could be testing other than gem 3

[elder burrow]

🤣

[light nymph]

Orionmist 3pro

[elder burrow]

yeah its also gemini 3

idk if its pro tho

[cloud zinc]

breckenridge is from grok

[elder burrow]

yeah

lo

ill put a list of all grok stealth models

[cloud zinc]

like what?

[light nymph]

Lets go make test and put here to check wft is this

[sullen quest]

btw codename-discussion is for this topic

[cloud zinc]

yes

[sullen quest]

gemma for one, it probably isn't becuase they already did this but 2.5 had a refresh with slightly newer models a month ago so techically that, also who knows what other models google is working on for stuff

[elder burrow]

GROK MODELS: sierra, breckenridge, zion, acadia, aspen

(there are likely more)

[sullen quest]

yeah this is a type of thing for codename discussion

[cloud zinc]

damn

[eager mica]

For what it's worth, a Google Gemma 4 release is rumored to happen soon, so I'd expect it to appear on LMArena at some point (although I haven't used it since a good while).

[elder burrow]

you're right

ill forward it

[elder burrow]

@deep adder look at ts xddddd

[halcyon harness]

LOL

[halcyon delta]

why can't we edit the prompt in direct chat?

[cloud zinc]

goat

[young owl]

Hi