general

[prime beacon]

?

[barren hatch]

hi, it's good to be here

[prime beacon]

Hello

[hardy swallow]

Hi please check #1397655624103493813 to learn how to use the bot.

[prime beacon]

Thanks

Bee

Hello

Bee

[gleaming roost]

[prime beacon]

Hi anyone here

[hardy swallow]

Hi!

[prime beacon]

Are you busy

[toxic egret]

yall can upload pdfs?

[marsh hedge]

nano banana been down since forever🫩

[gleaming roost]

I use nano banana pro every day, what are you saying?

[limpid compass]

yeah, its been almost 16 hours now of errors

[marsh hedge]

some are less lucky than you🫩

[astral blaze]

gemini 3 is goated

[marsh hedge]

onb im boutta become rich

i done discovered the method🥹

[keen lake]

welp

movementlabs making their own ide with coding plans added https://www.youtube.com/watch?v=o7qVzKxCnx0

[stray wasp]

🤔 😭

[quasi sparrow]

The website is broken for me

[hasty shard]

Hello fellows

@hasty shard

[loud verge]

@hardy swallow

[astral helm]

@hardy swallowplease remove captcha

its so painful

use cloudflare

[ocean dirge]

does anyone know why this prompt was blocked?

[sleek phoenix]

wtf

LMFAO you can now compare opus 4.5 with opus 4.5

[tame silo]

“Create a slow-motion cinematic video showing luxury perfume bottles emerging from deep black shadows. Each bottle should be illuminated with a unique golden light angle—one from the side, one from above, one from behind—highlighting the glass texture and golden liquid inside. Use dramatic lighting, soft golden reflections, and subtle smoke in the background. The camera should move slowly and smoothly, revealing each bottle with elegance and mystery. The atmosphere should feel ultra-premium, bold, and refined, matching the GOLDEN GRASS PARFUM brand identity.”

[astral helm]

ur website is not working properly

why is that

[ornate pebble]

if one day i become a supervillian it is because of captcha

istg

[fickle valley]

I want to generate images how do I just generate images

[astral helm]

fr man

[astral helm]

u cant basically

captcha not helping

[cloud oak]

is this the official seedream site?..i can't even gen any pics there.. 😭

[whole sundial]

the only official seedream site is dreamina (i think doubao in china and cici or whatever it's called now also have it, all sites i mentioned are by bytedance), all others are third party

[ornate pebble]

idk why but i swear the gemini-3-pro-image-preview-2k (nano-banana-pro) do beter in side by side than in direct chat

[cloud oak]

thanks..i mostly prefers image to image so i think im good with lmarena.. 🥹

[quasi sparrow]

Can someone help me

[pallid pagoda]

Hello team! Someone arrives has generated money with image creation, video?

[dry kayak]

Guys, what should I do when, like, the chat stops responding? I'm not talking about when it keeps loading forever or when it says something is wrong with the response. I mean, in the last 5 chats I had, asking ChatGPT to make a story to pass the time, after like an hour of conversation maybe, it would reach a message that had nothing typed, it just said 'no sources' about the web search part, and the options for like, dislike, copy, and retry—a message with literally nothing. And even if I try to hit retry, open it in another tab, open it in an incognito tab, or even use another device (I tested with mobiles), even after all that, retry does the same thing, it doesn't give any response. When this happens in the chat, usually all the ChatGPT options start giving this error, although others like Gemini, when I test, say that something went wrong with the message instead of saying nothing, although retry doesn't work there either. For some reason, the only A.I. that actually works in these chats, giving a real response, is Grok. Has anyone ever had this error? I didn't really know the name of what to look for in the bug table. (Note, I'm using a translator to translate this message, sorry for any spelling mistakes.)

[woeful pebble]

was doing the same to me hope theres a work around it first timer here🤭

[mossy holly]

https://tenor.com/view/hello-hi-hy-hey-gif-8520159980767013609

[daring rock]

@split harness Please head to #1397655624103493813 for a detailed guide on how to use the bot

[quasi sparrow]

Can someone help me

[tiny jolt]

Trust me: > websim

[rose sky]

Dope...

[uncut nacelle]

hii

[floral aspen]

how to generate a video on lmarena.ai

[ashen mauve]

my eyes

[wheat isle]

where can i find archived promot history?

they just disappeared😢

[dry hollow]

"Toile de Jouy style, monochrome blue, cozy kitchen interior, woman watching cat play with shadows on the floor, symbolic reflection, detailed pastoral scene, linework illustration, everyday life with hidden meaning, storytelling"

[valid rivet]

hi

[half trail]

Yo

[valid rivet]

m trying to gen image but due to captcha error cant

any fix?

[sick hare]

Hi,

[steel steeple]

How do I enable video mode on the website? I see the tops text to video text and the image to video.

[static osprey]

guys what happens when you switch between models in a chat, does it still remmeber your conversation wihtin the chat?

[misty shadow]

i pay for proffesional video~!

[verbal nimbus]

If same chat then yes

[plain adder]

hello

:D

[high aurora]

Good bye 🫂👋🏻

[static osprey]

thanks!

[verbal nimbus]

Nano Banana Pro:

[clear dagger]

Hi! Why is there no image loading on Grok models? There may already be an answer, but I haven't found it. I'm writing here

Gemini 3 pro has this feature, but Grok does not

[valid rivet]

@verbal nimbus can u help pls?

[shrewd citrus]

Gemini models aren’t image generating for some reason

[thin current]

Same

Guys, don’t sign up on that website. It’s a scam. MrBeast has never been involved in anything like this
FAKE

[tiny jolt]

???

[clear dagger]

?

[weary wagon]

Guys I’ve set up an Ai model yall wanna see?

https://www.instagram.com/itsaria.rose?igsh=djZ3NGRyb3RidnJm&utm_source=qr

Ain nun to clown bout bro @meager glen

[rigid copper]

<@&1349916362595635286>

i still don't know how to use @oak python bot thingy lol

[gleaming roost]

[gentle kindle]

Hi, I'm using battle arena sporadically (usually few times a week when other AI models will not satisfy me) and always great results. Lately I asked it to animate an image (which resulted in a 5 seconds movie being generated). I did it 2 or 3 times , then I got "an error occured". I'm wondering if there is a quota or something ? (since video generation is quite expensive). I have waited almost 12 hours and it's still the same error occurs.

[timber iris]

Hello, what are websites (like LMArena) where i can use GPT 5.2 X-High for free??

[solemn nacelle]

hi

[idle zodiac]

any chance to be able to generate 10 seconds videos?

[little patrol]

Hey guys, looking to generate hyperealistic product shots for my ecommerce brand. Has anyone got any prompt templates that I can trial?

[shrewd citrus]

Ask Gemini 3 pro with the search function for a prompt

then copy that prompt and use it

[little patrol]

thanks i'll check it out. I use ChatGPT 5.2 atm for prompts, output doesn't look organic.

[languid crescent]

Heyo, PDF attachment is not available anymore?

[stray aspen]

yes

not sure if i can say it though dont wanna risk getting banned for 'promotion'

[rancid hawk]

yo

[languid crescent]

I am curious

[hollow quest]

hi

[winter sigil]

Hello

[thorn valley]

i've mentioned yupp here before, as well as other sites, and nothing ever came of it

the issue is promoting it as a mass dissemination on the specific website, but simply mentioning it like that isn't a problem

[clear spear]

Did y'all improve the search?

[valid rivet]

Yup same

[molten heron]

hi

[zealous plover]

hi

[wanton wasp]

hi

[grave yew]

hello guys

[limpid compass]

Good grief. Is Nano Banana Pro still busted for you all? It's been a day and a half now of 'something went wrong' errors. Anyone else?

[grave yew]

i js got in

[hushed gyro]

@echo aurora I suspect that one reason image models are so unstable is because the constant switching between old and new testing UI.

[loud crag]

Sometimes, LMArena just tells me something was wrong with the response when I'm at my rate limit instead of the actual timer to wait, so I have to guess when it's passed so I can generate more.

Usually it'll tell me I've got 30-50 minutes to wait, and update each minute.

[limpid compass]

It's 50 mins - 1 hour. And believe me, it's not the rate limit. I try after every hour and I've been consistently getting errors. It's been broken for the last 38-39 hours now.

[loud crag]

Ah damn, that's okay then.

[gleaming roost]

[echo aurora]

Hello, if you're getting this error following the steps in this article may help. Would note the 2nd half of the article also has recommended ways to provide us with more information to help figure out why you're getting this error. Unfortunately, you can get this error message for various different reasons (including rate limit even though there is also the other rate limit error message), this is why getting us more info is helpful. cc @limpid compass

[echo aurora]

It's been broken for the last 38-39 hours now.
Can you share more on this?

[twin iris]

hello

[echo aurora]

constant switching between old and new testing UI.
What do you mean by this?

[echo hawk]

@echo aurora (sorry for the ping)

[echo aurora]

My apologies! I meant to respond sooner. I don't yet have an update for you, but will bump the question with the team. FWIW I did try a image prompt asking it ~"create an image of the weather today in SF" and it created one with with Oct as the date which makes me think it's not enabled.

[wise urchin]

I can't download pictures. Or create now

[echo aurora]

From the site or Discord? Are you getting an error?

[wise urchin]

From site

[echo aurora]

Can you clear cache + hard refresh?

is this mobile?

[limpid compass]

I've been trying to generate with pro and 2k every hour but I consistently get errors. I am on desktop. Maybe 1 in every 20-30 attempts across a couple of hours results in a successful generation. Everything was working consistently until about 5-6 pm Friday night (PST). I'll check out the article when I get home as I am currently out and about.

I posted a few times since that night and I think there were at least a few other people who said they were having probs with nano banana pro.

[echo aurora]

Gotcha, yeah sending the recording via the form in that article is going to be most helpful. Those models are really popular so it can be prone to errors unfortunately.

[echo hawk]

Well, apparently, according to what I saw on the minimaxir.com website, the date cutoff for the Nano Banana Pro is January 2025.

[limpid compass]

Sure. I will keep trying. As long as it appears the issue is on my end. Seems like it's working fine for you guys?

[echo hawk]

https://minimaxir.com/2025/12/nano-banana-pro/#:~:text=Although Nano Banana Pro's cutoff,when it doesn't work.

[normal abyss]

is their 1 tiny bus in the background i cannot tell 😭 i fail these every single time

(if someone finds an issue with how i do the captcha id appreciate feedback, ive tried going slower and faster but no dice)

[gritty arrow]

one of my video geenrations failed, does that count toward my limit?

[keen beacon]

🎥 FULL VIDEO SCRIPT (WORD-FOR-WORD)

Video Title:
What Is Web Hosting? Beginner’s Guide (Shared, VPS & Cloud Hosting Explained)

🎬 INTRO / HOOK (0–15 seconds)

“Do you want to create a website but you keep hearing words like hosting, VPS, or cloud and you don’t understand what they mean?

Don’t worry. In this video, I’ll explain web hosting in a very simple way — even if you’re a complete beginner.”

👋 INTRODUCTION (15–40 seconds)

“Hello and welcome to the channel.

This channel is all about Information Technology, web hosting, cloud computing, and practical IT skills.

If you’re a beginner, a student, or a business owner, this channel is for you.

Let’s get started.”

🧠 WHAT IS WEB HOSTING? (40 seconds – 2:30)

“So, what is web hosting?

A website is made of files — images, text, code, and videos.

For people to access your website on the internet, those files must be stored on a computer that is online 24 hours a day.

That computer is called a server.

Web hosting simply means renting space on a server to store your website files so anyone in the world can access your website at any time.”

[gritty arrow]

wrong place

it goes in #video-arena-1 or 2 or 3

[hushed gyro]

Well yk, the new UI is an experiment, and the UI constantly switches back and forth every time I use it

[novel iron]

Hi everyone, can anyone tell me how to remove the 5-query limit? Is it possible to buy premium?

[limpid compass]

@echo aurora I figured it out. I played with a few older prompts and narrowed down the problem. It's the new Nano Banana censorship that came out on the 16th. Sigh. Glad I identified the issue. It was driving me crazy. Hope they loosen the reins on the censorship in the near future. Thanks a lot Elon.

[echo hawk]

<@&1349916362595635286>

[hushed gyro]

Chat is NB Pro still the best image editor out there?

[ashen mauve]

yes by far

[loud crag]

I gotta figure out how to fix my prompt. I keep asking Nano Banana to only output a 4 image sequence, but it outputs a 6 image sequence like 30% of the time.

[feral holly]

is there any way to increase the video duration?and also is it true that nano banana pro is ending in this platform this month?

[rigid cliff]

MEV-aware flows
Is someone implementing {MEV-aware transaction flows}?

[near fossil]

Hey yall which ai agent is best for social media post

??

[frail rapids]

сова летит над лугом

[near fossil]

USA 🇺🇸

[ocean vortex]

rozmawiaj po angielsku, tu nie Rosja

[near fossil]

What’s trash 🚮?

[ocean vortex]

This hits differently since the last election and not in a good way, not gonna lie lol

I wouldn't react but I now I will 😇

[rigid holly]

The archive button is neat and all. But is there like a button to see your archive space where they are stored or what?

[wispy wigeon]

What’s wrong with the filter ?

[hidden ibex]

the 32k refers to the thinking or context window size of the model?

[ocean vortex]

thinking length

[hidden ibex]

thx

[chrome goblet]

Can you refresh the website to make the CAPTCHA go away?

[plucky island]

don't try that, just solve the CAPTCHA unless you're a robot
otherwise I think it gives them to you more frequently if you don't (not sure though, don't quote me on that)

[undone saffron]

I found a way to temporarily fix the reCAPTCHA error:

1. Log out
2. Clear all lmarena cookies
3. Start a new chat sending a message, then click Login again before the reCAPTCHA pop-up appears
4. Login
5. Problem solved
   After chatting with the AI for a while, the reCAPTCHA issue will show up again, so you'll need to repeat the same steps

Message for staff: Please just use Cloudflare’s captcha instead of Google's..

[hollow snow]

what is this model dart-frog-0113

[marsh grove]

anyone have a good face swap prompt?

[random wolf]

bruh I can't even verfiy the CAPTCHA. since yesterday I can't use the website

is there any solution? so annoying

I can't even use the website since yesterday. hays🥴

[lucid geyser]

I need hard prompts

For text to image benchmark

[torpid pasture]

Why my lm arena is crashing with like something went wrong?

[willow mural]

ik thats whats happening to me also

[willow mural]

its error 400 and 403

[torpid pasture]

I am not able to generate images from nano banana

[willow mural]

ikm

the sources have issues

[torpid pasture]

Is this a usual thing or it will get fine?

[willow mural]

it says bad request and forbidden

it'll get fine

it just tweaks a little

[torpid pasture]

Okay do you have any face swap prompts?

[willow mural]

not really, do you have any prompts to fix text?

[torpid pasture]

I use gpt to give me prompts

Mostly

Give prompts to give prompts

Lmao

[willow mural]

basically peeps wanted for me to fix the text on their shoe

the logo

they know i can do some pretty good prompts. but i basically cant fix it

[torpid pasture]

Yea the model is sucking

Right when we need it@willow mural

[willow mural]

yeah

[lucid geyser]

@echo aurora I had a text to image chat with many generations. I then sent in an image to edit but I’m very certain it changed a model and I couldn’t reveal the original. Is there any way to see it

[feral holly]

it would be nice if we could select more than 1 refferal image when doing image to video conversions

[scarlet spire]

That's down to the architecture of the models, not the interface that provides it the images.

[feral holly]

so you're saying we cant select multiple images as refferals in discord?

im not following

[fresh dome]

can you add upload doc or link from google doc ?

[stray flower]

Are there any AIs here on Discord other than for videos?

[frigid tusk]

is the switch to battle going to be removed ? its been a long test i think

[young palm]

hello

[novel crater]

(Done).

(End).

(Stop).

(Done).

(End).

(Stop). (Done). (End).

(Stop).

Done. Sending.

See ya.

(End inner monologue) (Start Action)

Response.

(Submit now).

(Bye).

[fervent beacon]

Where do I find the prompts, mentioned in the YT vid?

[vivid gorge]

Hello, I’m happy user of LMArena from mid 25. I was wondeing if it would be possible to add „system prompts” to UI. Most models support it and it would be very helpful to compare models with that feature used in conversation.

[wide jetty]

plz watch my video and rate it - https://youtu.be/V5iaVlc5_A4?si=HaqDh58FdyfcOeUw

[inland epoch]

❌ Generation failed. Failed to create evaluation session.

I;m getting this error when trying to generate a video. Has anyone faced this? What do I do to resolve this?

[fallow crow]

Pls help — real, up-to-date AI news. Might be useful for you: https://t.me/WWAInews

[deft grove]

**Hmmmmm…
So Siren is ... Grok? **

ONE MORE TIME?

I'll try

[granite ruin]

guys what the fix?

[ocean ferry]

get ready for non lazy Gemini 3 guys

[deft grove]

Gave me this 😂

[drowsy shuttle]

yo

fix when?

[granite ruin]

fix pls nano banan pro

[ivory snow]

how to unarchive my convos in lmarena

[proper plinth]

Why can't I make videos anymore?

[devout spire]

Why results of battle look awful this day

[deft grove]

https://lmarena.ai/?chat-modality=video
Try this

[tulip flume]

Got this message with every model I tried rn.

[granite ruin]

what the hell? fix something wrong

[deft grove]

Well, if that is actually Grok, I don't think it's the version we have right now.
If you ask the current Grok to generate its company logo, it won't generate the (X) ... it just animates a random logo

[frigid tusk]

same

[hot idol]

hellfc

[turbid comet]

How do I generate videos

[tacit salmon]

ai is going crazy

[frail venture]

In order to create educational videos using artificial intelligence

[zealous sparrow]

we dont have any examples of its work

[verbal ferry]

any info when will nano banana pro works ?

[quartz kettle]

Is there anyone looking for a skilled AI and full-stack developer?

[echo aurora]

This is being reported to the team, looks like we're seeing a very high error rate currently. Thank you.

[verbal ferry]

okay no problem are you going to type when it's fixed ?

[woven canopy]

Yo is yalls lm arena also tweaking?

[cedar tide]

Has anyone had the "nocturne" model?

[zealous sparrow]

rather new what company do you think its from

[cedar tide]

Is it text model ?

[echo aurora]

We're aware of this model erroring out at a high rate, team is looking into.

[echo aurora]

Check out #1397655624103493813

[zealous sparrow]

on which arena did you get it

[cedar tide]

I did not get it, i see it on bot detection

[woven canopy]

Nano banana right?

[echo aurora]

Yeah

[woven canopy]

Ok thanks

[zealous sparrow]

which bot

[cedar tide]

🤐

[zealous sparrow]

okay

fair enough

doesnt that bot have a check model command or smth

[cedar tide]

Its not my bot, i dont have more info

[zealous sparrow]

ill try and get the model on textarena

you can try prompting codearena

worse if its an imagearena model

[hushed gyro]

happened exactly 8 times in 10 minutes! unacceptable!

[deft grove]

They said they're fixing it bro chill

[hushed gyro]

yes but this happens so much even before the high error rate

[stray aspen]

<@&1349916362595635286>

[echo aurora]

Really sorry for these errors! The team is aware of this model having high error rates at the moment.

[random prairie]

Don't all of them have brothers working?

[dreamy ivy]

when can we expect it to be fixed

when can we expect it to be fixed

[echo aurora]

I don't have an ETA unfortunately.

[dreamy ivy]

but its even high for other image models too

[echo aurora]

The overall problem of image models erroring out + focus on pro-image models has been flagged

[vernal zodiac]

same

[vernal zodiac]

same

[shrewd citrus]

It’s been like this since yesterday

[plain wedge]

Hola

[warm zinc]

hello everyone, thank you for the acceptance/invite

[little ginkgo]

We have pdf file uploads but not txt file 🙁

@echo aurora Pls txt file

300k character and 20k lines of code

[opal osprey]

im trying to figure out a way to upload files to the models

anybody figured out a way?

[little ginkgo]

U cant

I tried everything

[opal osprey]

NOOOOO

THERE HAS TO BE A WAY

[little ginkgo]

Models can't even curl links

🙁

[opal osprey]

please... dont say that.. 😭

😭

and the devs obviously wont make them support files right?

[little ginkgo]

Ye bro

[opal osprey]

i wonder why

[little ginkgo]

[weak sluice]

Oh?

Only showed when I'm not signed in

[echo aurora]

Something we're working on still!

[obsidian cargo]

sad that we can't emoji react on announcements anymore I need to properly convey my absolute lack of interest in an image model being ranked 21

[echo aurora]

Correct

[little ginkgo]

[shrewd citrus]

Will the model read and edit the pdf?

[echo aurora]

Oooof, we may bring back custom emoji reacts at some point. It was becoming a bit difficult to mod.

[obsidian cargo]

<3

[little ginkgo]

Chatgpt is the best for image

[echo aurora]

Early versions are going to be for vision, but we should have fast follows on other modalities

[little ginkgo]

[weak sluice]

Is it an upcoming thing or..

[little ginkgo]

Pls add text file first

Before anything

[echo aurora]

Something we're experimenting with

[echo aurora]

Is for sure on the list

[little ginkgo]

Who asked for pdf btw

Text file is important

Gonna solve my problem

[weak sluice]

Sweet. Thanks for telling me Pineapple.

[proud bobcat]

no new text arena update

sniffle

[echo aurora]

[crude anchor]

Guys, I think that in a few weeks they'll give us the best news, which is the video arena on the website!

[spark python]

Is image gen down?

[stray aspen]

has anyone tried ernie for developing

is it good

[verbal ferry]

what now with nano banana pro will it work or no?

[normal abyss]

50/50, itl work eventually

[cloud oak]

why is this on the list if it doesn't even works?.. 🥲

[woven geyser]

How to upload files in lm arena chat

[lucid geyser]

Is it possible to reveal the model of a chat

That got broken

[wicked wren]

Hi, why is this appearing?

[terse shuttle]

i can't send image

[little ginkgo]

Bro

Is this down ?

[marsh hedge]

take a screenshot and put it again

it'll work

[cloud oak]

i keep failing the recaptcha thing..its getting annoying now.. 😭

[plain wedge]

Why isn't it working?

[terse shuttle]

try to rewrite ur prompt

[plain wedge]

I've tried everything, always the same message

[terse shuttle]

what is ur prompt about?

[plain wedge]

There is no problem with the proprietary

Nothing I type works

[terse shuttle]

try to clear cookies

[marsh grove]

anybody have a good face swap prompt?

[normal abyss]

🙏

[lucid geyser]

Why

[raven otter]

What the hell happened to the image models

[vestal ledge]

YES

[hushed gyro]

@echo aurora No error for the first time in 8 hours! 🤭

OMG errors are gone and the responses are faster (now under 30 seconds)! Bravo team! W fix 🔥

[distant idol]

please lmarena save us 😭

[unreal ermine]

Hi

My lm arena app isn't opening

[ashen mauve]

im so happy that the A/B chat bricked my chat

[obsidian cargo]

I hear there's some gemini 3 pro GA thing in aistudio A/B testing that's really good, hope it comes to lmarena soon

[woven canopy]

Yo thanks for fixing it

[marsh grove]

how do i use negative prompt?

[dark delta]

The Gemini 3 Pro 2K option is no longer appearing for me, why?

[gleaming roost]

This was the worst implementation they ever did in lmarena

[thick matrix]

How to add prompts guys ?

[proud bobcat]

Son I’m crine who’s gonna use this

Outrageous price

[rigid copper]

sup guys

[normal abyss]

i agree so much, for 2 years their hasnt been a single bad update till that 😭

[thick matrix]

How to solve that ?

@quick jackal

[dark delta]

There is a limit to the number of artworks you can create in LMARENA?

[placid flame]

@thick matrix Can you clarify it? What do you mean?

[thick matrix]

I mean the image to video generation does not generates video properly in lipsing

Video and the audio does not match

@placid flame

Just help and solve this only one problem

Please

How to generate better lipsing videos ?

[echo aurora]

That's going to be up to the model and the prompt so I'd recommend researching that area. The mods here won't be able to help out with that sorry to say.

[echo aurora]

Each model is going to have it's own rate limits, onto of that each modality is also going to have their own limits. These limits can change over time and we don't currently have these limits publicly listed.

[stuck walrus]

X2

I archived some chats I made with many decent results in a specific style, and now I can't find/access that chat

[proud umbra]

look at o1 pro

[thick matrix]

Ohk no problem 👍 but you can tell me how to see which model I am using? Like veo 3 or other model for example

[elfin harbor]

Is battle mode in direct chat going to be permanent? @echo aurora

I personally think it should be separated 🙂 It happens out of nowhere

[buoyant basin]

how to make a video from image i try already but i still no create a video from image

[dense willow]

Hello, could someone please let me know if there is any limit on image generation in LMArena?

[elfin harbor]

Yes there is limit

[dense willow]

What is the daily limit, and if I want unlimited access, is it paid?

[elfin harbor]

Different rate limit for different models I've heard I don't think there's any paid option at least to my knowledge others might help you on that

just wait few hours and try again or create a new thread

[dense willow]

Thank you very much for your kind help.

[thick igloo]

halo

[ashen radish]

how to use @tiny palm bot in my server

[stray grotto]

hello!

[ashen radish]

how to use this bot in my server

@tiny palm @hollow isle

@everyone

[whole sundial]

<@&1349916362595635286>

[rigid copper]

<@&1349916362595635286> another hacked guy

for all cost, do not click any suspicious link, attachment or whatsoever.

[unborn crag]

Why is images showing error on mobile

Whenever I add images

[whole sundial]

yeah crypto is the easiest to steal from as they ask you to log in, once you do so they'll just take away all your crypto

and it's literally just some project, i imagine that service is real but this "insight"/"wrapped" site isn't

[rigid copper]

i don't really trust things like free money or something, because there is no such thing as "free lunch". And yeah, people fall into scam when they provide login credential or downloading malware in their devices.

just to be safe, scan that website link with virustotal and see if it's really safe to visit or not

[rigid copper]

true

well, i was in online school, and it was too boring with it, so i asked claude to make the picture in picture bookmarklet for youtube, i can't really access to most of the website (including youtube) because most of the school don't allow any student to watch any video while studying.

but i don't care, so i bypass it.

idk where is the coding channel is, all i thought is #codename-discussion but which is not.

at least, i can outsmart teacher that i use bookmarklet script to watch youtube by pasting link instead of using extensions, incognito, guest mode, or other browser, and they will not record any sign of youtube history because it's from picture in picture using script.

🙃 might be sounds stupid like "why you open youtube at school?" but this is actually clever way to use youtube without getting caught.

and i wondering, why is there 2 claude with same name? was that model duplicate or what?

[whole sundial]

there is but you'll need to sign up for arena champion (#announcements message) to get access to it

idk why that isn't promoted more frequently

there is a bunch of good people here who I think deserve the role but they don't have it because it's never promoted

[peak sapphire]

@echo aurora Nano Banana 3 Pro and 2K haven't been working for 3 days.

[frigid tusk]

when are you going to remove or decide on the switch to battle mod!

[whole sundial]

<@&1349916362595635286>
malware

[stone cape]

yo kiri

have u checked world knowldge on hawk ultra?

[whole sundial]

no, i forgot their models existed

i didn't have much time to test it, maybe i'll try it out though

[stone cape]

lol 😂

its a big step up

[stone cape]

wana know ur take on it

[whole sundial]

this is actually good, but the response formatting reminds me of claude

[stone cape]

the code gen beats opus 4.5 64k thinking in many cases

surprisingly

[stone cape]

the best part is it has context output of like 300k tokens

[whole sundial]

btw does anyone here know what the free hawk max/ultra limits are?
and can i get an invite for movement labs discord (if it's still around)

[stone cape]

its on https://x.com/movementlabsAI

movementlabs (@movementlabsAI) on X

Try here:https://t.co/lz6TQKhl5O
Discord: https://t.co/dm3KNTIodD

X

[stone cape]

im subbed to pro plan i get like 25 messages a day

for like $6 a month

but easier to just make new accounts

💀

[tropic tide]

hy

[rose sky]

Is he / she cute? IDK if it’s a he / she, I will get him / her next week

[rigid copper]

<@&1349916362595635286> that is gonna be malware.

[rigid copper]

i see, thanks for the information.

actually i saw before there is an application form that you can register to be part of arena champion

but since i'm new to this server, i guess i no need to do so, to become arena champion, i have to be experienced on AI which i'm not much experienced about it.

[whole sundial]

yeah it's suspicious, the zip contains a renamed chrome executable and an even more suspicious file called "chrome_elf.dll" which is probably malware
connects to some random website which is probably what's spreading the malware

[rigid copper]

the name itself isn't legitimate, and yeah, 100% sure to be malware or trojan, and unrelated file name topic tho.

[whole sundial]

this is the name of the website it pings

the actual site

[whole sundial]

literally just a bunch of random words they put together

also it refuses to run in triage, which means this is 100% malware, please do not run this file on your computer

[haughty flame]

👋

[vagrant sand]

are there any rate limits still on using claude opus models on the site?

what are the rate limits of it??

please someone reply

@echo aurora

[cloud oak]

-# just my thoughts.. 😶‍🌫️

• lmarena introduces battle mode for direct chat to encourage more users to vote for the ai models..(because the number matters?)
• the Captcha popping up after every single vote.. (i can't pass the captcha and i feel discouraged from voting further.. 🥲)
• sometimes an ai Assistant would get stuck on infinite generation..i can't vote and see the names of either models..
• the gemini 3 pro 'something went wrong with this response' error..(im thinking it's the model's way of saying "you have reach the rate limit" at this point)
• users become frustrated from getting stuck with the chat and/or the captcha and they votes less/stop using the website.. 🥲

[verbal ferry]

what is now with nano banana pro?

[verbal ferry]

guys nano banana pro is still not working

[deft grove]

YES!

[deft grove]

Facts lol

[random prairie]

Can I find out when it will be fixed?

[elfin epoch]

hi

[cobalt tendon]

Hello

[naive hawk]

No lol

[woven canopy]

Can yall not like find a way to make Kling ai free like the rest of the models

[gleaming roost]

[sage raptor]

be grateful that you are using that for free

[verbal ferry]

can you make nano banana pro working

[echo aurora]

We had some problems with the model over the weekend, but the error message rate is much lower now, so you shouldn't be seeing as many error now.

[echo aurora]

It's been a bit unstable lately, the team is aware of the problems. The overall error rate as been much lower compared to yesterday.

[verbal ferry]

when is it gonna be resolved ?

[echo aurora]

We don't have the specific model rate limits published, we may change this as it's understandable why this would be helpful.

[echo aurora]

I don't have an ETA sorry to say. The team is aware.

[echo aurora]

It's unlikely that we fully remove rate limits. It's a very helpful tool to protect us from exploits/bad actors.

[unkempt cave]

I feel like the video arena results on 4K video can be not precise and wrong because there is too much aliasing on Veo 4k video, once compressed down to 720p on lmarena it sometimes look superbad and I believe most people will see it as worse quality

An example: #video-arena-3 message

Maybe there should be special encoding parameters to prevent that from happened

[frigid tusk]

oh dang

[barren sorrel]

Where can I find a person on the Internet who will read a text written by an AI and tell me what he doesn't like about it?

[sage hill]

Hello sir can anyone here help me how to start yta

[echo aurora]

yta?

[sage hill]

Youtube automation sir

[old garnet]

I’m experiencing a specific issue that is not included in the bug report list

[final lintel]

bro why is it that i cant even use lmarena for coding bro

every code i sent

OR EVEN IF I SAY HI

THEY MAKE ME DO RECAPTCHA THATS BASICALLY IMPOSSIBLE FOR A HUMAN

💔

[old garnet]

I have a multi-AI chat, but Claude is the only model not working. The issue is isolated to just this one chat session.

[final lintel]

please switch to cloudflare captcha again it was so much better than this bro

[echo aurora]

We pay close attention to the captcha system and do plan to make adjustments when appropriate. Overall, the system is going to kick in when it suspects in-authentic use. Would recommend avoiding use that may contribute to this.

[final lintel]

lmarena was my fav spot to use ai's, but after this im thinking i cant even use it properly because each message just gives me a "verify human" captcha, which i cannot do, i click everything properly but they js say error

[final lintel]

i did bro, like i'll send u a recording i'll say hi and it will prompt me with a captcha

[old garnet]

Does anyone know how to fix this error?

[final lintel]

clear cookies or change browser bro

@echo aurora i have to say the captcha is so annoying heres the recording

[old garnet]

I already gave that a try, but no luck

[final lintel]

idk then bro

[knotty shale]

hello

[hushed gyro]

We having errors again huh?

[knotty shale]

hello

[thorn iron]

Usually constructs are considered a she

[knotty shale]

helllo

reply me

[thorn iron]

Hello, Ali

[knotty shale]

yes

which country you belong

[thorn iron]

USA

and you?

[knotty shale]

caneda

[thorn iron]

Canada?

( have 7 sistersaand 8 brothers

[echo aurora]

Going to ask we keep conversation related to AI please

[thorn iron]

all of which are Navy Seals

yes AI is why I'm here

NO number for you

[echo sinew]

@marble tartan Hello! Please read the info posted here 👉 ⁠⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate videos or images using the bot.

[mint folio]

hello

[verbal ferry]

still

[echo aurora]

This is something we're aware of and are looking to reduce the error rate for. Would note this channel #1417174113092374689

[quartz light]

😭 we want custom reactions backkk

can Arena Champion have permission for custom reactions?

oh wow thats a lot of arena champions..

i remember last time i saw i was one of the only few with the role...

[coarse compass]

gemini 3 pro appears to be down

[echo aurora]

Can't say we're seeing higher than usual error rates with this model. What problems are you getting?

[coarse compass]

something went wrong please try again

[echo aurora]

Can you give this article a try: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message ?

[coarse compass]

this is what i get when i am trying to go to the website

[echo aurora]

When going to https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message you get that error? Or are you seeing that on https://lmarena.ai/?

[coarse compass]

this is when i got to the help.lmarena.ai/article/etc

[echo aurora]

That's strange, I've tested on a few different browsers and unable to get the same. What if you go to just: https://help.lmarena.ai/ ?

[coarse compass]

i get the exact same page

[young niche]

Is the arena make 30s videos

What is the arena limits

[vapid crag]

I have a problem: when I send a prompt on my Nano Banana, I always get the error message "Something went wrong with this response, please try again." This problem has been persistent, especially after the last two updates. I would appreciate it if you could help me fix this issue.

[echo aurora]

Okay thanks for the info. I'll keep an eye out for other reports and dig deeper. It's unclear why you're seeing this. To confirm: you're able to go to the site just fine, right? It's the help center that you're getting this error for?

[echo aurora]

Sorry for the error! Our team is aware of this model erroring out higher than usual, it's being looked into.

[proud rampart]

Are there limited

Sorry ...that's embarrassing

melodic contour, rhythmic flexibility, timbral variationYinsen (陰旋)

• Type 1: D – E♭ – G – A – C (1 – ♭2 – 4 – 5 – ♭7)

• Type 2: D – E♭ – G – A – B♭ (1 – ♭2 – 4 – 5 – ♭6)

Sorry again

[vapid crag]

I hope you can resolve this quickly. Also, please remove the excessive filtering. Sometimes I send prompts in Spanish or French, and when I send them, it says they violate the terms of service, even though the prompts don't contain any offensive words.

[echo aurora]

We are collecting suspected false positive flags from the content filter in #1447983134426660894

[dense wyvern]

since when did lmarena add the video option

[echo aurora]

It's something we're currently experimenting with.

This is fixed now btw

[thorn valley]

🙏 thx

[celest orchid]

@echo aurora

[echo aurora]

[eternal thicket]

How do i add my own audio to my code arena

[celest orchid]

but you can't

[strong lake]

Does anyone here have a project idea or an ongoing project ?
If you have any difficulties or need a developer, let's talk.

[crimson bear]

Who actually runs this site? The Vatican or the Mufti of Mecca? I feel like a ten-year-old whose mom is constantly scolding him.
“Grimy mattress on the floor” – This content violates our Terms of Use.
“Bottle of beer on the table” – This content violates our Terms of Use.
“Lying drunk in bed” – This content violates our Terms of Use. This content violates our Terms of Use. This content violates our Terms of Use. This content violates our Terms of Use. This content violates our Terms of Use.
Pretty boring party by now.

[whole sundial]

<@&1349916362595635286>

[whole sundial]

this too

and this

[viscid cloak]

Nanopro, input triggers instant "something wrong" (No ToS issue cuz about 1/5 input passed with right result). Please fix.

[steady grove]

so i made ai agent and told him: you can modify yourself to evolve with new features, and it wrote function to get "available features" from github

im bit scared

[dark delta]

Wtf

[gleaming roost]

[marsh monolith]

What does failed with evaluation session mean

[stray aspen]

Cooked

[dark delta]

Can someone give me some guidance on this error?

[normal abyss]

unfortunately their isnt a guarentied fix for the nano banana pro errors, you just gotta get lucky

from experience, youl get better results of success from the gpt image 1.5 model (at least for me)

[echo aurora]

We are looking into why this model is erroring out higher than usual. It's worth trying the steps in this article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

[whole sundial]

i'm pretty sure this is an issue on google's end, they probably can't handle the amount of nano banana pro generations

it's been a problem ever since the model launched

probably why nano banana flash has been delayed so much, it probably won't launch before gemini 3 pro and nano banana pro either hit ga, get another revision, or both

ironically, launching nano banana flash would probably help google by decreasing the amount of nano banana pro usage since the flash model is not much worse than pro

[proven bronze]

It won't let me try again either.

[dark delta]

Something is definitely wrong.

[proven bronze]

Now it lets me try again.

But the error message keeps coming up that something went wrong.

[dark delta]

The only thing to do is wait for a response from the developers.

[proven bronze]

It's strange that this is happening at this time of day.

It usually happens in the morning and at midday

At least here in Argentina

[echo aurora]

I don't think the timing of the day has much to do with it tbh

[dark delta]

For me, the Nano Banana option isn't even showing up in LMArena

After this error appeared

[dark delta]

@proven bronze Take the test now

I just generated an image in Nano Banana.

[proven bronze]

I usually see nano banana.

[proven bronze]

It works a little better now.

[dark delta]

Nice

[echo aurora]

Are you not seeing it in the drop down?

[dark delta]

Yes!

To find it, I have to access it via a link.

[echo aurora]

Are you able to send a screenshot?

[echo aurora]

Ah, you need to swap to Image Arena

[dark delta]

Even so, he doesn't show up.

But regarding the errors, it seems things are returning to normal.

[echo aurora]

Can you send a screenshot of the list when you're in Image?

[dark delta]

The error persisted again.

._.

[lucid geyser]

I got a question about model naming. The Gemini 3 models don’t have preview in the name

But aren’t they preview models?

[echo aurora]

Yeah this we are aware of, but what I'm more concerned on right now is if you're not seeing that model in the drop down, as this isn't something I can repro and not seeing other reports of.

[crimson bear]

So the most exciting battle on LMArena right now is:
Something went wrong with this response, please try again.
vs.
This content violates our Terms of Use.
In the category: This is getting on my nerves.😬

[stray aspen]

lmao

[frail granite]

bro does the ai keep crashing for anyone

[whole sundial]

glm image can't beat... a 4b model that can run on most consumer hardware

[echo aurora]

How often are you seeing a Something went wrong error in Battle?

[normal abyss]

has anyone got a site that i can access veo 3.1 (not fast) for free

even if its 1 generation only im needing to generate something once

[stray aspen]

glm image sucks

[echo aurora]

In what way?

[whole sundial]

glm image requires an nvidia ai accelerator like a h20 to run and even on there it runs terribly

meanwhile flux.2 klein 4b can generate an image in 1-2 seconds on my 5060

[hollow flicker]

GLM is coming back

how is GLM coming back so hard

Z.ai let them cook

[whole sundial]

glm 4.7 flash is also trash compared to qwen3-30b 2507 in my experience, it's probably much better in coding though

would've been better if they released a base model

[toxic whale]

Haven't been keeping up with new drops, what does everyone think of glm image?

Also where could I get access to glm 4.7 flash, I can't run it locally

[whole sundial]

z.ai api and openrouter have it

[normal abyss]

depends which one, which are you using

[whole sundial]

i thought the model was bad when i ran it locally, then i ran it in openrouter and it wasn't much better

honestly disappointing, would've been better if they released the base model

[frail granite]

oh

im uuusing it for

coding

[normal abyss]

through claude?

[whole sundial]

all the finetuning for coding probably harmed its world knowledge

[frail granite]

gpt

[whole sundial]

glm 4.5 was very good in world knowledge for its size then its successors made it worse for the sake of better coding performance

[frail granite]

gpt 5.1 high

[whole sundial]

we've came from glm being much better than qwen in world knowledge for its size to qwen being better than glm (in the 30b size range for world knowledge)

[normal abyss]

oh, gpt 5.1 high always works like every time for me, it can take a while to gen sometimes tho

[frail granite]

ive been waiting 3 days btwe

[whole sundial]

and they gave no details at all about the model's training, so we don't even know if its a qwen (or nemotron, which is also garbage) finetune or not

[normal abyss]

is it just stuck generating???

[frail granite]

yes

i refreshed everything

also

i used gpt again and it said error do this or smth

and i kept retrying it either 1 of those 2

[signal oasis]

Guys, I’m having a really hard time with image generation and it fails very often. Is this problem just happening to me, or is everyone experiencing it?

[normal abyss]

if its stuck generating, unfortunately your chat is corrupt from infinite generation, youl have to start a new chat. Its been more common recently with how often this happens.

[proven bronze]

Solve the problem with a nano banana ❌
Making interface changes that nobody asked for ✅

[normal abyss]

this is new to me though if its retrying and erroring, what region are you in?

[quartz light]

its trash

i can confirm

[frail granite]

europe gmt to be ecact

exact

[normal abyss]

same as me, idk why it would just be erroring for you, il quickly try use the model and get back to you

[frail granite]

thanks

[normal abyss]

what prompt did you use can you send it here, il try do the same thing

[frail granite]

the first prompt i gave it or the last one

[normal abyss]

which ever one has the error (not infinite generation)

[normal abyss]

yeah gpt 5.1 high is working fine for me (sry for late response)

[echo sinew]

@shell bolt Hello! Please read the info posted here 👉 ⁠⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate videos or images using the bot.

[rigid copper]

sup guys

[frail granite]

ima try again and update u

its working fine

i think

after 40 texts its gonna crash idl

[normal abyss]

normally its claude thats not working, icl this is new to me 😭

[obsidian cargo]

What's auto modality?

[normal abyss]

i may be wrong but isnt that the thing where when you type a prompt, it selects the best model for your prompt?

[obsidian cargo]

ahh

[surreal frost]

@echo aurora pineapple man can someone unlock my tab that's been generating for 12 hours

weird how you guys didn't make some timeout to make sure this wont happen but whatever

[pure quartz]

Guys

What's the best model right now thats uncensored in the sense that it could like tell me to make chemicals

[balmy mist]

im pretty sure you can just jailbreak it

us opus

use*

opus never really declines prompts, but i guess it depends on what you are trying to make lol

if you trying to do some weird stuff, i would advise against that cause they will block your access, so just use an open model

[ionic sinew]

grok

[pure quartz]

Which grok model 4.1 thinking is kinda censored no?

Im on android but I have a windows pc and I have paid subscriptions to chatgpt plus and gemini pro cause of my phone s25 ultra came with it

I use antigravith for coding idk how to code though

[ionic sinew]

no i use it and it only decline a very few things. give it a try

[pure quartz]

I want a llm that gives the best smartest answers but isnt censored at all

[ionic sinew]

grok, if you have a good pc you could try a full uncensored model locally.
but it won't be close.
give grok a try

[pure quartz]

I have a ryzwn 5809x with non xmp ddr4 32gb ram and a 3070 is that enough?

Is there way to make grok just be uncensored

Should I buy the subscription even though im on android

[ionic sinew]

bro did you try it or you just want to chat

try to edit the prompt and add "educational purpose only"

if you have fast internet try a 80B or close uncensored model will be good

[surreal frost]

get ollama and install their godmode model and do whatever weird stuff u wanna do

[pure quartz]

Is it smart?

[surreal frost]

it's a recommended model, I haven't tried it myself just heard it's good

and uncensored/jailbroken (no doubt)

goodluck

[ionic sinew]

godmode model
what is that

[surreal frost]

the model name has Godmode in it or the entire name was godmode and some other numbers/letters I dont remember exactly but I could check

oh yea I think uhm smthn like this

https://ollama.com/Godmoded/llama3-lexi-uncensored

actually there was one which the name was literally smthn like godmode

or maybe I'm remembering wrong,

[ionic sinew]

godmoded is the name of the publisher not the model
also there is way stronger models like lama 400B on HF

[surreal frost]

more numbers doesn't usually mean it's better, but since I haven't tried that model you mentioned I'll take ur word for it

[balmy mist]

bro is asking for a miracle

[ionic sinew]

I haven't tried it either.
it's the same "Godmoded/llama3-lexi-uncensored" but bigger

[balmy mist]

there is no model like that

u have to sacrifice

best model is opus and u not getting illegal stuff from it, ur best bet is dumber models like Godmoded/llama3-lexi-uncensored

[surreal frost]

I never mentioned the model would be a genius, I mentioned it's good
might not be a great option, but sure as hell the best

unless the 400b model mentioned by Mido is better, I haven't tried tho

[surreal frost]

did you mean this?
https://ollama.com/aravhawk/llama4:400b

[ionic sinew]

https://huggingface.co/nicoboss/Hermes-3-Llama-3.1-405B-Uncensored

[surreal frost]

oh I see

@pure quartz this one's good u could try it, but it's installed differently I think

[pure quartz]

I was recommended this one

https://www.google.com/url?sa=t&source=web&rct=j&opi=89978449&url=https://openrouter.ai/sao10k/l3.3-euryale-70b&ved=2ahUKEwj8-_-Q25uSAxWBSTABHe9pLeQQFnoECCYQAQ&usg=AOvVaw0RBTAwsjXqG73MiAcUHOIr

[surreal frost]

It doesn't seem to mention being uncensored nor godmoded, only that it's for roleplay unless thats what you're looking for

besides, I dont think you can download from openrouter

[pure quartz]

Its because I only have a 3070 I cant run thwarted big modela

[surreal frost]

no it's cause openrouter doesn't let you download models, they're a router and they themselves dont host the models you see here

[pure quartz]

How would I be able to run it on a 5800x 3070 32 gb of slow ddr4 ram

[pure quartz]

Ikcause I don't think i can run any good reasoning models locally

[surreal frost]

try the godmoded model I linked before or the one by huggingface

[surreal frost]

this

[surreal frost]

or this by mido

[pure quartz]

405b is that even possible with my specs????

[whole sundial]

try an abliterated gpt oss 20b or qwen3 30b

[surreal frost]

no idea

[whole sundial]

https://huggingface.co/DavidAU/OpenAi-GPT-oss-20b-abliterated-uncensored-NEO-Imatrix-gguf

[surreal frost]

damn

[whole sundial]

https://huggingface.co/mradermacher/Huihui-Qwen3-30B-A3B-Instruct-2507-abliterated-GGUF

[surreal frost]

@pure quartz u could try those ig maybe keep finding till you find one that works for ur specs if not

[pure quartz]

What do I run it on

[rain bay]

You can try to use chinese open-sourse models. Though they do not have top performance, their censorships are thin like a paper.

[surreal frost]

just to clarify, you can't download any models at all. period. from openrouter. they're a routing service not a hosting for models so you can only use their apis to chat or use their chatroom nothing more.

[pure quartz]

I also need a model with a lot of context or a RAG system because I have large files it needs to reference, like books and encyclopedias.

[surreal frost]

it lists I think

[whole sundial]

you get lmstudio and run them in there

[pure quartz]

Which are called what

[whole sundial]

4bit quants will run great on your system

[surreal frost]

oh my, that would make it laggier

[pure quartz]

Ik

[whole sundial]

i'm choosing MoE models here because some other models (seed-oss-36b/hermes 4.3-36b) are dense and are thus much slower

[rain bay]

deepseek series, for example

[wispy raven]

Create mata laxmi video, video size 16:9

[rain bay]

it never refuse to answer me after a simple prompt engineering

[whole sundial]

the only deepseek model you'll be able to run well with 32gb ram is v2 lite chat, which is pretty old nowadays, an uncensored (abliterated) qwen model would be better

[surreal frost]

pineapple man is on everyone ready up your questions

[whole sundial]

there are the r1 distills, some of which can run on 32gb ram but they're still a bit dated and once again, qwen3 30b would be better

[pure quartz]

I prefer non local models that aren't paid for by the use

[surreal frost]

usually doesn't exist but

i did find this one
https://openrouter.ai/chat?models=cognitivecomputations/dolphin-mistral-24b-venice-edition:free

[whole sundial]

50 messages per day fully for free, 1,000 if you put $10 in your account

[pure quartz]

(Venice) Invalid request parameters. Please check your input and try again.

Why does it always do that

[surreal frost]

I'm actually not sure

[surreal frost]

well yea

[rain bay]

Qwen models do better for smaller models.

[surreal frost]

which model do u guys think among this list would be good at being brainwashed and doing illegal stuff for you, asking for a friend

[chilly crane]

A young Indian school boy in uniform walking toward his village home. His mother stands near the house watching him with a smile. Rural Indian village setting, soft daylight, emotional family moment. Vertical 9:16, realistic, cinematic.

[rain bay]

you can try to ask other models

[surreal frost]

yea sure

[ashen mauve]

i am ai

[chilly crane]

Image to video how to create

[surreal frost]

dude did you even glance over the channels list

have u seen this

[ashen mauve]

no

[surreal frost]

https://ptb.discord.com/channels/1340554757349179412/1400148597768720384

[ashen mauve]

he has no eyes

so he can't see

[rain bay]

kimi k2 thinking tells me to use aion-rp-llama-3.1-8b

[surreal frost]

seems like it

the first 3 channels in the server dude

[surreal frost]

@chilly crane

use your eyes do you not have any

[surreal frost]

for what purpose, jailbreaking?

[rain bay]

it is uncensored in huggingface model card

[surreal frost]

send link

[rain bay]

https://huggingface.co/bartowski/Aion-RP-Llama-3.1-8B-GGUF

[surreal frost]

nvm I can't run that, see if you can find someone who can

@pure quartz u could try these models see if your pc can handle them if the other one fails

[pure quartz]

What about if I want a local or free open router model that agenrically controls my pc

And can code

Good

Free gemini?

[surreal frost]

ah

thats a different story,

try agent zero

https://www.agent-zero.ai/

[pure quartz]

What about one that fills out pdfs

What about one that uhh

Teaches me how to code or and bother language is notebooklm the best for that?

[surreal frost]

pretty sure this program is meant to be an all in one agent that does everything

but

I don't know specifics

[surreal frost]

oh it's also uncensored

no idea how they managed all that into one program

[pure quartz]

How do I learn to vibe code hack lol

[surreal frost]

vibe code hack?

[pure quartz]

Yeah likebi dont knoe hoe to code but I wanna hack stuff

[surreal frost]

not exactly what I expected to hear in one sentence today, but if you wanna learn cybersecurity yea I think agent zero can help with that

[surreal frost]

okay well "hacking" isn't all code btw.. people use tools

[pure quartz]

How do I automate my workflow completely. for free

[surreal frost]

using agent zero

[pure quartz]

What model

[surreal frost]

agent zero Is the model, and it comes with it's tools it's very advanced

[pure quartz]

What coding ide / agentic coding program should I use i use antigtavity rn

[pure quartz]

Oh thats very cool and its free?

[surreal frost]

yes

[surreal frost]

personally I legit just use notepad++ so

[pure quartz]

Does it work on windowd

[pure quartz]

What model is best for coding rn gemini 3.0 pro or claude 4.5 opus?

[surreal frost]

there is a setup guide for most platforms including windows

[surreal frost]

claude

[pure quartz]

Even for big programs?

[surreal frost]

yes

well personally I use gpt-5.2

[pure quartz]

Why

[surreal frost]

it seems to give less errors and leaves out less stuff when making code than other models I've tried including claude

[pure quartz]

What is the best ai deep research type thing is it something local Google deep reseaech chatgpt research perplexity what is it best uncensored too

[pure quartz]

Im looking at the guide you need an api key

[surreal frost]

agent zero includes one

[surreal frost]

from?

[pure quartz]

So its not free

[pure quartz]

https://youtu.be/SEBWHaX2L5s?si=4j7I7N6qldbR4WBW

[surreal frost]

I remember there used to be a way to make it use your own preferred ai

oh yea in the .env file
OLLAMA_BASE_URL="http://host.docker.internal:11434"

should be listed in more detail somewhere in the guide

[earnest pike]

Hello

[pure quartz]

Which ai should I use

[surreal frost]

whichever you prefer

[pure quartz]

Which isbthr been free one

[surreal frost]

for agent zero?

[pure quartz]

Yea

[surreal frost]

probably the default option it has which is openrouter

[humble summit]

Yo guys what ai model is the best for websites?

[surreal frost]

5.2

[humble summit]

What's 5.2 sorry I'm kinda new

Claude opus?

[surreal frost]

GPT-5.2

[humble summit]

Ohh

Thankss

[surreal frost]

np

[toxic egret]

is it just me or the UI changed somehow?

[junior holly]

Can anyone tell me how to generate video?

[toxic egret]

#1397655624103493813 or if you are lucky sometimes the icon appears on the website

[surreal frost]

yes it did and it broke my chat 👍

[spiral brook]

Hi

[rain bay]

GPT always spend the most time

[rotund rivet]

why is the ai taking forever to generate

its been 8 hours

[surreal frost]

been happening more often after the design change I feel like

a change neither of us asked for nor needed

[obtuse smelt]

hmm

[undone solar]

This Problem 3 days ago but stil loading how to solved the problem

Any Admin Hare

been happening more often after the design change I feel like
a change neither of us asked for nor needed...😓😔

[azure tendon]

Hi

[undone solar]

Yes

[echo galleon]

How to Become an Arena Champion @tiny palm

[whole sundial]

#announcements message

[cloud oak]

the website can't be refreshed anymore by pulling down on the phone screen?.. 😭

[woven canopy]

Reload it manually from ur browser lol

[meager fulcrum]

Gemini 3 image generation stopped entirely

[wind ember]

any new models or nah

[viscid cloak]

Rip Nanobanana Pro 💀(For don't know how many times it had happened)

[brave lichen]

Not credited image all showing something wrong fix now

[hollow imp]

Huh?

Why are there two same models

[hollow imp]

You can use it through perplexity max

[obtuse smelt]

waiting for fix

[golden ocean]

one is normal and the other is a cutie pie

[shy surge]

hello

[lofty frigate]

Tf is this

[naive hawk]

They banned you

[lofty frigate]

Damn

[stray dock]

why do i keep getting the poll to choose bw 2 agents

in gemini 3 pro

[obtuse smelt]

wait is are same i retry not responding too

[thin current]

Guys, remember that this service is completely free, so we can’t complain too much. Instead, we should thank them for the great work they’ve done

[brave lichen]

Not credited image all showing something wrong fix now

[obtuse smelt]

me too

[brave lichen]

seedream-4.5 working but image1.5 and gemini not working

seedream-4.5 working but image1.5 and gemini not working

Image1.5 and Gemini only work from text to image . Not working from reference image plz fix now

[crude anchor]

@echo aurora Sorry to bring this up because of a somewhat silly question, but there are any news reports or indications that video AI will be available on the site in February (I don't have a phone number to verify on the Sora website).

[gleaming roost]

What kind of update is this? It completely messed up my prompt

No one wants to fix bugs, but implementing useless things is something everyone wants to do

[rain bay]

router is the top reason for the model to become stupid

[cerulean sorrel]

hello my prompt is in process from last 12 hours now what?

[vital sedge]

Why can't I upload a PDF file to the site?
It was working yesterday

[daring shuttle]

Best video generation model that don't miss?

[rigid copper]

sup

[rigid copper]

that's why i don't use smart router, best to do it manually

[gleaming roost]

It simply activated itself

[rigid copper]

well, i guess you use battle mode, i never get this when i use direct chat or side by side

use side by side or direct chat then

[gleaming roost]

I'm furious because this happened in direct chat and it even activated itself automatically

[viscid cloak]

Thank? Funny. Trading in your ideas&prompt and voting results, while still saying it's "free". From when peops could give up these self-diminishing mindset.

[gleaming roost]

Well, the "free" option is actually paid for with our data.

[rigid copper]

idk, mine did not have it

mine still show the same model i chose it

[fossil solstice]

getting the same error on both models

[rigid copper]

the only error i got is this one, recently nano banana pro failed to run it.

[fossil solstice]

i think the model is down or something

[distant hawk]

Guys is ai Limited?

Bc When I Was Chatting IT It Showed Me That

[fossil solstice]

i think it’s just down.

[rigid copper]

sometimes it does run, but about 75% it will just fail to run

[fossil solstice]

as of now?

[rigid copper]

yes

[fossil solstice]

because yesterday it was working just fine

it did have a few hiccups but yea

[distant hawk]

Guys Which Better For Coding Gpt 5-medium Or a Gpt-5.1?

[viscid cloak]

Too early for Cali to rise up solving urgent service down for more than 6 hours

[rigid copper]

imagine using lmarena code arena to create the website of image to image ai using nano banana

and it did not work

[grim radish]

Hello... Thanks. I will to try video creation

[rigid copper]

lol, let me see if i can try creating my own ai image to image using lmarena code arena

[gleaming roost]

[distant hawk]

same

[bright shard]

Nano Banana Pro Dont work

[livid prism]

Hi guys

[rigid copper]

[livid prism]

What is the difference between deleting a chat and archiving a chat?

[stray hamlet]

Is there any way to make the chat start at the bottom when I open it? Whenever I open the chatbox, it always loads at the very top of the conversation. That means I have to scroll all the way down every time just to get back to the most recent messages, which gets frustrating pretty fast (especially in longer chats).

[flint sandal]

did rate limits got removed? i remember there was like 5 message limit with opus 4.5 thinking per few hours, now i can send like how many messages i want and no errors?

[left lodge]

New icon.

[rigid copper]

deleting a chat will not be recovered and it will be lost permanently, archiving a chat keeps it but hide from you and others iinm

[livid prism]

How do i recover archived chat?

[rigid copper]

in discord or what?

[hollow mist]

Please tell me what is the problem of errors when generating a nano banana image

[livid prism]

Site

[rigid copper]

what site?

[gleaming roost]

[livid prism]

lmarena.ai

[hollow mist]

Please tell me what the problem is? And when will it be fixed?

[left lodge]

New model selection UI.

[rigid copper]

you can't archive the chat in lmarena

[left lodge]

For side by side

[rigid copper]

i never see any settings related to archiving the chat

[livid prism]

Hm, i cant do it too now

[hollow mist]

Please tell me what is the problem of errors when generating a nano banana image

[livid prism]

But a few days ago i could archive a chat

[golden ocean]

https://tenor.com/view/hal9000-hal-hal-9000-space-odyssey-2001-a-space-odyssey-gif-16460551034971887576

[rigid copper]

idk, large scale error, if i'm not wrong, rate limit?

[hollow mist]

Will they fix it?

[left lodge]

And direct too.

[left lodge]

In my opinion, the sorting of the models should be according to their elo scores. From high to low.
They order seems random and change time to time , how does the order work rn? @echo aurora

[rigid copper]

i think it's fault on model itself, trying other model like gpt 1.5 seems to have no issues

[rigid copper]

i think you actually delete the chat but you thought that you're archiving the chat

[left lodge]

The platform is hitting the api capacity. They just dont have enough rate limit. So when enough requests are sent ,all remaining after that gets rejected.

[rigid copper]

ooh, i see...

[left lodge]

Thats why only the popular and good ones suffer with this.

[rigid copper]

for real though, i don't really know much about AI api system thingy

[rigid copper]

as the demand increases, more error it causes

[left lodge]

Yeah cause they have a fixed allowance per sec.

[rigid copper]

true for that

[left lodge]

And if you try retry enough it will just randomly work because you were lucky enough to send your request before the system reached its limit.

[viscid cloak]

api/rate budget shall be rightfully calculated. More than 6 hours service down for a mature company recently received huge funding doesn't sound right.

[rigid copper]

they will just reject the request when api capacity reaches

like when your phone is fully charged, your phone will stop drawing the power, preventing overheat and overcharge

[echo aurora]

Hey everyone - just a heads up that we are seeing issues with gemini-3-pro-image-preview models, the team is aware and we are looking into asap. I am sorry for the inconvenience this is causing.

[left lodge]

…

[rigid copper]

take your time, and gemini-3-pro-image-preview-2k suffered the same.

[echo aurora]

It's based on rank

[echo aurora]

I'm not sure what you're referring to, but overall I won't be able to confirm if/when new features (or features we're currently experimenting with) will be available and fully rolled out.

[sonic basalt]

Maybe someone knows how make OPUS 4.5 multi agent use in Antigravity or VScode ??? Tried a lot of ways but always hitting limits... 🙁 some people say that it's posible just create multi GMAIL logins to Antigravity and make automatic opus 4.5 thinking to swith betwen accs not hitting to prompt TOKENS limit in FREE accounts...

[left lodge]

Which rank? Leaderboard?
This one isnt.

The logos are a little off in color because of forced dark mode.

[golden ocean]

https://tenor.com/view/blahaj-shonk-attack-camera-dive-shark-gif-14344361424253681658

[fading rover]

In a matter of fact gemini itself has issues last two days even my api key aistudio won't generate 2k or 4k in random cases throwing errors so yea doesn't seem to be lmarena issue on that front.

[hushed gyro]

please fix the NB issues, it's pretty frustrating 😅

https://nanobananas.ai/nanobanana2

chat use this while uhh this issue gets fixed ig?

[echo aurora]

Ranks that are on the leaderboards, so when you're in Text modality it'll reflect the Text Arena leaderboard. Not all of those models have gone through ranking evals through Battle.

[echo aurora]

We're looking into currently

[round island]

why do the mods have weird usernames
lm_mod_6
lm_mod_18

[tall tulip]

Is it just me or nano banana always error when generating images?

[woven canopy]

Yep