general

[gusty egret]

Wait that's gemini

Are Google servers suffering rn

[shut bough]

Hi guys 🤝

[prime umbra]

hi

[wind shore]

Guh, the AI are so affectionate. TwT

sorry-

[wind shore]

This seems dangerous.

[wind shore]

I'm sorry, but that sounds very shady.

I'm sure you have had that many years experience, but to announce it like that rubs me the wrong way.

[errant rover]

Please help

It's always like this ?

[sterile tartan]

That's The Best Part of it

But it can't generate music in all languages i had tried in the past

[potent lintel]

How i can find my videos. I can not find the videos i have given prompt for.

[sour spear]

If they generate successfully, you'll get a dm from the bot.

[fossil fable]

lmarena gpt-5.2-high help me start a $2,000,000,000 freeloading business

[golden ocean]

https://cdn.discordapp.com/attachments/1442924145615568956/1442924156546056272/MConverter.eu_v14044g50000d485kpfog65rnjameha0.gif

[quasi mulch]

I'm trying to build a knowledge base with my chats here using Zotero + SingleFile.

Most of it works (search, markdown), but the 'click to jump' feature for notes and outlines is totally broken. I think it's because the saved webpage source is non-standard or 'upside down'.

Does anyone know how to make the saved file navigate properly in Zotero? Thanks!🫡

[fossil fable]

Eternal

[south raft]

CTRL + F

[golden ocean]

https://tenor.com/view/cat-twerk-gif-25814656

[flat rain]

why can't i upload pdfs or docs on the website? it only allows uploading of images

[tight zenith]

Why can't I use Direct Chat in Video Arena?

[flat rain]

you can, just use /image

[west lodge]

is he speaking the truth

[silk fractal]

Hi help me admin please

[west lodge]

did the ram shortage rly not hit as hard over here cuz im seeing plenty of ddr5 kits for pre-inflation prices still

[sour spear]

Just browsed some images on the Midjourney discord, and was flabbergasted at how MJ still cannot generate proper hands. Looks like they're still stuck in 2023. 😦

[cloud palm]

why is it randomly switching to battle mode from direct

[hardy swallow]

Hi please check #1397655624103493813

[rose sky]

Damn Kevin…

[fossil socket]

I legit need to know why the Claude rate limit got severely shortened

[prime umbra]

Please, remove battle mode from direct chat. It’s infuriating, especially during longer sessions.
Seriously—first fix the constant generation issues, the endless “something went wrong” errors, all the bugs you already have. Then think about adding new features.
Because in its current state, this will only drive away people who are here specifically for direct chat.

And honestly, if you absolutely have to keep it, then at least think about making battle mode much rarer. Because… I can’t even swear on this chat. You know how annoying it is when I’m in direct chat and 50% or more of the messages are battle mode? That’s infuriating. Set it to 1%, maybe even less, but not this often.

Today I’m literally playing on my nerves because every 10 seconds it ruins my roleplay. Which is basically the only reason I’m on LMArena at all.

[bitter swift]

is there any way to get the lm arena bot in my own server?

[modest tundra]

The site is great, but the battle mode in direct chat has become too frequent. I have nothing against it, but when the second model takes more than 120 seconds to load or even goes into endless loading, it's frustrating.

[hollow ivy]

How is there a battle mode inside direct chat?!

I thought they are (3) fully separated modes.

I hope, they never remove battle mode, it's (almost) the only reason i use this site!

[prime umbra]

Not remove battle mode entirely, but remove battle mode from direct chat. You know—there’s a choice for a reason, so users can choose, not have it forced on them.

[hollow ivy]

ok, now that's weird
when i choose direct chat (or side-by-side chat) i get just that, not a "fake battle mode"

[prime umbra]

You know, for me it works like this: when I use direct chat for one or two messages, it actually behaves like direct chat. But when I use it longer, over time every second message turns into battle mode. It either doesn’t load because of a “something went wrong” error, or it gets stuck on constant generating. And that’s infuriating.

[hollow ivy]

do you use it anonymously?

without account?

[prime umbra]

i have account and use with account

[hollow ivy]

oh, then you hit the rate-limit (?)

when an LLM is used up, it might automatically select the next one (?)

[prime umbra]

What do you mean?

[sour spear]

Thing is, battle mode is the core of this site. All else is basically a free giveaway, but only battle mode can provide unbiased feedback. Not even sure how this works, is side by side voting weighted the same way as battle mode? Because let's face it, side by side makes it easy to manipulate the rankings, by simply voting for one's favourite model.

[hollow ivy]

the error message might have appeared, because you hit the rate-limit of the given model

(for Claude [esp. Opus] it's pretty low)

only battle mode (used with account) has no rate limit

[prime umbra]

Even if—look, even if there is some kind of limit—wouldn’t it be better to just say, “Hey, you can’t do this, please choose something else,” instead of forcibly throwing battle mode into direct chat?

[hollow ivy]

yeah, exact error messages would be a great improvement for the site, i 100% agree

there are too many cryptic errors

in science, transparency is king

and we do (citizen-)science here:
comparing AI models/LLMs methodically

[prime umbra]

So is it possible to make battle mode simply disappear from direct chat? Because honestly, sometimes it’s genuinely infuriating. It loses context, you end up waiting hours for a response, and the interruptions are just annoying.

Maybe the site should fix the current bugs first, and only then try to add battle mode to direct chat.?

[proper carbon]

https://sora.chatgpt.com/invite?code=K8E7A9 👀

[prime umbra]

So... could you at least disable battle mode in direct chat while voting? Please?

[hollow ivy]

i'm a mere discord user, have no rights

(everyone can do polls)

[hollow ivy]

..you become "arena champion" when having used the site for longer or entered early

(iirc, if you entered at least half a year ago or so, but am not 100% sure on that)

[prime umbra]

@hardy swallow you can look here?

[sour spindle]

Haven’t been on here in awhile any interesting models in the arena in the last couple months?

[rose sky]

EMMA?!

[rocky frigate]

@echo aurora I am really sorry for the pin but can u check ur dms ?

[echo aurora]

It doesn't look like you've initiated the bot successfully, you'll want to review the information in #1397655624103493813 .

[echo aurora]

Yeah I'll be able to respond in a bit.

[twin summit]

時間飛快，農曆新年即將到來。
這是一支全 AI 製作的新年 MV，所有角色與場景皆由 Gemini Nano Banana 生成。
歡迎觀看與支持，謝謝。

Time flies—Chinese New Year is around the corner.
This New Year MV is fully AI-generated, with all characters and scenes created using Gemini Nano Banana.
Appreciate your support.

▶ https://youtu.be/EHun3aMz5UA?si=Srs6XEJvH15l2pMi

[echo aurora]

Hey thank you for sharing this feedback.

It loses context
When a battle in direct happens the chat looses all of it's context and you have to start over basically?

you end up waiting hours for a response
As in whenever you get a Battle, it takes hours to generate a response? How often does this happen?

[echo aurora]

Not even sure how this works, is side by side voting weighted the same way as battle mode?
It's only votes in Battle mode that build the Leaderboards. Votes done in Side by Side do not contribute to the rankings.

[echo aurora]

There is not.

[sour spear]

Hey, thanks for the clarification! 🙏

[prime umbra]

Alright, so here’s how it is.
As for context — not always, but it does happen. And in roleplay, let’s be honest, context is worth its weight in gold. As I’ve said more than once, I mainly use LMArena for roleplay.
And about the hour-long generation times — sometimes it takes an hour, sometimes 3 minutes, sometimes instantly, and sometimes from the moment I started until now.

[prime umbra]

you can look here https://discord.com/channels/1340554757349179412/1458282112062062593

[raw sage]

hello all group members

[echo aurora]

Hello

[raw sage]

how are you

[echo aurora]

Well, hbu.

[echo aurora]

Gotcha. Thank you for flagging this and putting it together. I'm reporting this to the team.

[prime umbra]

And one additional suggestion, if I may.
If you plan to introduce Battle Mode into direct chat, I would ask that it be enabled less frequently. As shown in the attached image, this can be genuinely frustrating for users.

[rocky frigate]

thank you soo much

[echo aurora]

Thank you for sharing this. The issues outlined by @prime umbra is being shared, but hearing from others that the Battles in Direct is getting long load times, or endless loading, is good to know.

You mentioned this happens with the

the second model
Are you noticing the same @prime umbra ? That the errored out/infinite loading chat is the 2nd one (right chat)?

[rocky frigate]

yo pinapple why is " udders " considered voilation ?

[prime umbra]

I’ll be honest — my English isn’t very strong, and I rely on translators. But regarding the endless error, like the one in the attached image in https://discord.com/channels/1340554757349179412/1458282112062062593 … ever since I submitted it, it still hasn’t loaded.

[prime umbra]

Could you please ping me when Battle Mode in Direct Chat is either removed or fixed?

[normal abyss]

last time it was on the site it lasted maybe a week or 2 then it was gone for a good month, if we are lucky it will be gone forever sooner than later

[prime umbra]

I know. He disappeared recently because of me too XDD

[normal abyss]

i typed that wrong il edit the message rq

[echo aurora]

Makes sense. Do you see this errors happen mainly with Assistant B?

[echo aurora]

Sorry to say I can't promise this. These problems and feedback have been shared with the team, but I won't be able to reach out to you directly for what changes comes from this feedback.

[prime umbra]

Honestly, it depends. Sometimes it's A, but 90% of the time (yes, 90% is a random percentage defining the majority) it's B.

[echo aurora]

Good to know, thanks.

[prime umbra]

ok

[prime umbra]

I have a suggestion. Maybe... conduct a survey to see if people actually want this combat mode in direct chat?

[normal abyss]

i wouldnt mind so much if when it came up, it went away and didnt effect the history/flow of your current chat

i dont mind voting and using it, its just when it sticks onto the chat history.

[echo aurora]

It is an experiment at the moment, and we are looking at the data to see how users are reacting to this idea.

[prime umbra]

same.. bro same

[echo aurora]

Yeah it shouldn't be effecting the context like this. This is something we're going to look into.

[prime umbra]

Honestly? Judging by the current feedback after two attempts... it's not going well.

1. Less often, it should be offered.
2. Context must be maintained.
3. It can't get blocked.

[echo aurora]

Thank you for sharing.

[final lintel]

bro why does this stupid captcha pop up every prompt bro

[frigid tusk]

ummm sry but the switch to battle feature is uhhh, starts switching even on the 2nd prompt which ngl annoy me little bit , yall gotta see about that

[normal abyss]

If your not in the UK or America (maybe a few others) it detects you as a bot (i may be wrong) this is just going off what i heard, some countries get this.

But again i might be wrong as i havent tested this.

[normal abyss]

it would also have to not touch rate limits

[echo aurora]

The captcha is going to trigger when it suspects inauthentic use. Any kind of behavior that the system might find odd is what is likely contributing to you getting a captcha. If you're prompting to fast/frequently would be recommended to slow down.

[echo aurora]

Yeah this has been reported to the team.

[prime umbra]

Could you pin this and ping people? We'll see what they think. Since it's an experiment, it's probably worth hearing people's opinions, right?

[frigid tusk]

alr ty

[final lintel]

💔

[echo aurora]

The best way to share feedback about a feature is to use the #1372230675914031105 section. A poll in a Discord channel isn't going to be fully reflective of the LMArena user base and could give an incorrect understanding of how users feel about a new feature.

[raw sage]

@echo aurora I’m a professional AI developer, and I also build high-quality websites. I’d really appreciate it if you could give me a suitable role in this server please.

[echo aurora]

We don't have an AI developer role to give. I may add one if I see others show the same interest.

[blissful jungle]

https://youtube.com/shorts/1RaC3XYgSwg?si=zRHxe3_GxIXcuMqk

[echo aurora]

Sorry for the delay - the difference is the system prompt between the two models.

[raw sage]

@echo aurora Can you give be a high leavel role and What do I need to do to qualify for a high-level role among all of these server

[echo aurora]

You can apply to be an Arena Champion, more information here: #announcements message

[magic ravine]

What happened to the image models?

[balmy basin]

same question

[magic ravine]

Nvm, they're back

[prime barn]

Please after knowing the ai model used to for you video how do you use the same model for your next videos..?

[nova mortar]

hi, i have a problem with video generation on the website
the video button sometimes appears, but when i click it, it disappears and nothing happens

am i supposed to generate videos directly on the site, or only through discord video arena channels?
is this a known bug or a feature limitation?

thanks in advance

[gleaming roost]

[magic ravine]

Well according to what I've heard, it's still in its beta phase right now and they're planning on implementing it in the future. So just gotta wait,

[sour spear]

Why would they? It costs a s***ton of money to run all these models, they have to restrict usage somehow.

[lean pivot]

👋

[echo aurora]

Video Arena is going to be Battle mode only, meaning it's two random models so you won't be able to make a selection for a specific model

Expanding upon file upload options is something we would like to do. I don't have any details to share though about this project.

[echo aurora]

We are experimenting with Video Arena on the site, since it's an experiment that'll mean it's random if you get the feature or not. We do have Video Arena on this Discord server (more info in #1397655624103493813 )

the video button sometimes appears, but when i click it, it disappears and nothing happens
When you got the experiment, if you click on the icon you should be sent to Video modality, which should look similar to how Image modality would look. When you clicked this you stayed in Text modality? Did the Video button remain there or go away?

[echo aurora]

I'm going to be starting a thread for this issue btw, going to have some followup questions in #1458498310150164593

[topaz lichen]

does it make any sense that uploaded prompts on x for nano banana pro doesnt work here?

[echo aurora]

You'll want to use image modality on the site - https://lmarena.ai/?mode=direct&chat-modality=image

[topaz lichen]

well yeah my bad, i was refering to that page, it mostly says violated terms of use for prompts that exist and works uploaded on x

[final lintel]

bropbro is ts funny to yall, spamming captcha

BRO IM LITERALLY GIVING IT A PROMPT EVERY 5m AND IT STILL GIVES ME CAPTCHA\

[high dirge]

is this a bug me got response a and response b even if I was in direct chat

[echo aurora]

We are experimenting with what the occasional Battle happening in Direct mode looks like.

[restive owl]

hello, i have a problem, can´t verify my email

[echo aurora]

Sorry to hear that! What seems to be the issue?

[teal grail]

hi

[stray aspen]

has anyone tried that new movement labs model?

it wont let me start conversation

it just doesnt answer me

[full tangle]

Hi guys how are you today

[full tangle]

What is movement labs model

[restive owl]

When I receive the verification email, clicking on it takes me to a webpage with an invalid URL

"http:///nextjs-api/callback/email?token=......."

[astral blaze]

this llm knows whats up

[echo aurora]

I'm going to followup in Direct Messages.

[acoustic cliff]

lmao

[astral blaze]

look they know there are 3 r's ok please stop asking

[acoustic cliff]

but do they know if 9.9 is bigger than 9.11 though

[astral blaze]

here is your answer

[acoustic cliff]

😭

[astral blaze]

the full response is quite funny

look it knows ok please stop the aboose

[flat rain]

@echo aurora when will the feature of uploading images to claude model be added please? it's really important sir

[granite heron]

@slender mantle hey there, please head here first: #1397655624103493813 message, you will find our guide to learn how to generate content using the Discord bot

[echo aurora]

We are currently working on adding this. Sorry to say I don't have an ETA on when to expect this.

[proven bronze]

I suppose this should fix the infinite generation problem 👀

[robust sluice]

the counting number is gone or something ?

[proven bronze]

They removed it.

[robust sluice]

oh

[peak granite]

hey guys can someone please help? I was using chat for a while but now I cant go back to it and there was so much contexct in that thread I dont know what to do

am I cooked???

it sbeen hours

and only happens in that thread

[robust sluice]

a lot more like inf gen if theres no number

(didnt try yet, just sayin)

[surreal creek]

beluga is responding to like 2/3 of my prompts Amazon wants to benchmaxx LMArena so badly 😂

[echo aurora]

I'd recommend checking out the steps in this article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

For that model the rate limit was adjusted recently, so this is likely why you're running into this error.

[peak granite]

yes but its been like 4 hours

still nothing

I was using it heavily for hours on end the past 2 days

im going to go to sleep hopefully everything will be ok when I wake up

🙁

[peak granite]

There is no thread limit right..?

I really hope not

it will be devestating

I paid for claud now but I dont know how to get all this memory imported its so much....

me and lmarena version of claude go way back these 2 days 😭

[echo aurora]

A way you can check if it's rate limit is by...

Go to Developer Tools, Network tab, search Stream, run a prompt in LMArena that triggers an error, this will then create a file with the Eval ID (random set of #s/letters in the URL), in that file you should see a Status Code, if this Status Code = 429 that means it's rate limit.

If you're seeing a different status code, the best way to get us helpful information is following the 2nd half of what's in that help center article. Basically using a tool like jamdev to then submit more information to that form is really helpful to us.

[peak granite]

hm ok let me try

[echo aurora]

no thread limit right..?
Meaning limits on how many different chats you can have in your chat history? No, I'm pretty sure this doesn't have a limit and confident this wouldn't be the reason you're seeing error.s

[peak granite]

status code 200?

what the heck that mean

[echo aurora]

Yeah that isn't going to be rate limit, something else is broken.

[peak granite]

noooooooooooooooooooooooo

[echo aurora]

(I deleted the image btw)

[peak granite]

can I fix from my end???

[echo aurora]

Yeah if you can submit a recorded video the form in that help center that'd be helpful.

[peak granite]

aw man okay...

[echo aurora]

I don't believe so unfortunately. The steps in the article may help (clear cache/cookies, hard refresh, new browser), but it's hard for me to say for sure. Regardless, this is being escalated.

[peak granite]

oh ok.. I tried with other browser too

its only that specific thread

[echo aurora]

And same result?

[peak granite]

yep

I tried mobile too

also same

Where is help center?

here is the video error 200

[echo aurora]

https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

[echo aurora]

Let's submit that to the form please. That's in the help center ^

[pure swan]

When will you add file send I need to send files to the modules and I am on mobile larger file more than 450lines get cut if file is 600 file I can only copy about 450 rest get cut manila copy is horrible if file is like 3k lines of code it's really simple fix

[peak granite]

😭

[echo aurora]

Hmm that's odd. Can you try in a different browser?

I've never seen that error message before

[peak granite]

so many problems with me nooooo lol

Also I noticed error happend as soon as I made a new thread and sent gemini pro 3 a picture

when I went back to the thread

it started doing that

up until that point everything was ok

same issue dif browser

[proud bobcat]

anything but fixing gpt 5.2

[ocean bison]

But no, for voice cloning https://huggingface.co/spaces/jordand/echo-tts-preview

This boils every other tool.

And it's very indistinguishable with the right seeds!

[acoustic orchid]

anyone know whats going on with this please all day i had this issue and i cleared me cache and history on me browser and tried to refresh what im doing

[echo aurora]

Really sorry to hear you're getting this. This shouldn't be happening in Battle mode. Can you confirm for me that:

1. This is in Battle mode (didn't happen in Direct chat where you were sampled a Battle)
2. What modality this is happening in?
   How often are you seeing this happen?

[acoustic orchid]

its in battle mode and its happened all day for me

[echo aurora]

Which modality (Text, Image, Search)? Can you try this and let me know if it's still not working:

• Hard refresh the website
• Clear your browser’s cookies and cache
• Start a new chat

[acoustic orchid]

text and done hard refresh of the site and cleared browser cookies and cache and tried new chat this morning was going ok but then kept getting the message i posted

i have also tried another browser and still getting same message

[echo aurora]

Okay, can you go to this Help Center article and follow the steps under the If the problem continues... section, use a tool like jamdev to record when the error happens, submit that to the form, and ping me when you've done so?

[acoustic orchid]

website keeps crashing now

can not get to do a recording as website is becoming unresponsive

[wicked sage]

from what i see in the video on the latest attachment, i find the office really nice! i feel like this could be a house

anyways hi guys

[neon idol]

When the output was working, have u closed the tab or closed the browser or just swiped to another tab?

(Talking about lmarena tab)

[acoustic orchid]

i finished what i was doing last night and turned my computer off and i got up this morning did one thing on what i was doing and then left it open went to work and come back and went to do something else and go tthe message i posted

[neon idol]

Thats the problem

When u close the lmarena tab sometimes the output stop working

And u get that issue

[acoustic orchid]

one of my other projects still working no problem

[wicked sage]

minor thing and should not ping pineapple but they accidently forgot to link the channel here

[acoustic orchid]

im having issues with the website not responding

i try again tomorrow see if the website sorts it self out

[wicked sage]

trying to make this (second image) but keep failing to get the vibe (first image)

hi

[stray aspen]

burning memory

lol

[wicked sage]

its everywhere at the end of time but close enough

[acoustic orchid]

how do you close things down so you can carry one with your project another time

[neon idol]

Idk

[echo aurora]

Ah thank you, going to update this.

[wicked sage]

yw! i didnt wanna ping you guys for such minor convininencnen (idk how to say that word)

[stray aspen]

well i guess movement labs aint that bad after all

[wicked sage]

is this supposed to be "thats not my neighbour"
if it is, then this ver is really cool

[echo aurora]

Hey sorry was in a meeting, are you still seeing this?

[obsidian cargo]

more like papers please, which thats not my neighbour was inspired by

[wicked sage]

ah crap my bad,

[obsidian cargo]

how shameful, utterly indefensible /s

(the /s denotes sarcasm)

[wicked sage]

https://cdn.discordapp.com/attachments/1038238583686967428/1457461459364741272/attachment.gif

[stray aspen]

yeah its papers please

idk whats thats not my neighbor

[wicked sage]

coolio

[obsidian cargo]

its a game where you're a guard guy who has to determine if people trying to get into their apartment are evil doppelgangers or not

[stray aspen]

bro movement labs is kinda cooking ngl

despite all of the scam accusations

[wicked sage]

what in the world is movement labs

[stray aspen]

a company

that claims to have a revolutionary ai model

they have been involved ina lot of drama recently lol

[acoustic orchid]

yes its not opening so can not do a recording

[echo aurora]

What are you seeing when you go to the site?

[acoustic orchid]

if i use my windows sandbox i have no issues but only issue i have is if i close my computer down i lose whats on my sandbox

[acoustic orchid]

it loads up and then i get a pop up saying wait or end and when i say wait it goes back to website and then does the same again

[echo aurora]

The site just isn't responding. I'll make a #1343291835845578853 post and we'll use that. Very curious if others are seeing the same.

[acoustic orchid]

im off to bed i check tomorrow and see how it gets on then might have to do it as a direct chat part instead of battle

[wicked sage]

am i on something or is this just claude's ui

[quasi atlas]

@regal flume Please check on #1397655624103493813 to learn how to generate content.

[obsidian hawk]

use hawk max model it cooks

[stray aspen]

yeah it has really nice one-shots

unlike some other models where you have to polish with more prompts

[surreal creek]

It tested under the name “riftrunner” so really just when riftrunner showed up

This is the first ever mention of riftrunner in this server

6 days before release

checks out? they do normally codename test for about a week before full release

[stone cape]

hawk max surpasses claude opus 4.5 thinking model for sure

congracts @swift oyster on the release

[wicked sage]

me when costs money lets goo

wait does it im stupid

[stone cape]

its FREE 🍺

[wicked sage]

NO IM STUPID I need an account

what am i on about

BRO

❤️

i meant to send the heart emoji

[stray aspen]

movement labs is cooking so hard goddamn

[wicked sage]

holy crap THATS SO BRIGHT LOL

[stray aspen]

it gave me a nice sunset scenery on roblocks with just one prompt

while gemini, gpt and claude were struggling

[wicked sage]

for some reason hawk is just completely broken for m

e

[stray aspen]

is it not answering?

it wasnt answering me until now

idk why

[stone cape]

MY EYES 💀

[stray aspen]

yeah that happened to me

it just fixed itself

[stone cape]

reload page or just start new chat bruh

[wicked sage]

wtf

ok thx

[stray aspen]

@wicked sage

try disabling the dark theme

lol

[wicked sage]

no shot

[stone cape]

bruh what is that prompt 😭

[wicked sage]

shut up

i play balatro and i wanna mod

🔥

also holy crap it worked wtf

@stray aspen

[limber panther]

whats this

[wicked sage]

movement labs ai

[stray aspen]

lol

[stone cape]

and make sure u use hawk max

[wicked sage]

you can ask lneduo2en

[limber panther]

movement labs ai?

[wicked sage]

he knows more

i think

[stone cape]

yeh they made a strong come back

[limber panther]

whats that

[stray aspen]

'scam' company

[limber panther]

i dont even know them

[stray aspen]

but they have a good ai model so i dont mind

lol

[limber panther]

really

how good is it

[stone cape]

better than opus

[stray aspen]

its pretty solid

it cooks

[stone cape]

imo

[limber panther]

4.5 opus?

no way

[wicked sage]

i dont care where i use, i care what i use

[stone cape]

yup

look

[limber panther]

gotta be fake benchmark

[stone cape]

at what i made yday

[limber panther]

iquest coder was fake

maybe this is too

[limber panther]

wtf

[stone cape]

they dont even have benchmarks 😂 💀

[limber panther]

broooo

[limber panther]

man why did it add shaders lol

[stone cape]

this is what @stray aspen made

[wicked sage]

when is pineapple adding hawk max fr

[stray aspen]

never

everyone says its a scam

[limber panther]

broo

[wicked sage]

beautiful but OH MY GOD MY EYES

[limber panther]

did u make that

[stray aspen]

yes

[stone cape]

well people still call sam altman a scam

😭

[limber panther]

wait thats roblocks studio

not AI

[stray aspen]

hell yes

[limber panther]

thats roblo tho

[stone cape]

ai made the code for him

[stray aspen]

yes

yes

[limber panther]

for what

the sun?

[stone cape]

the whole thing

he just copy pasted it

[limber panther]

its just roblo studio in baseplate

..

[stray aspen]

it gave me a code that changes your lighting settings so it look slike this

[limber panther]

oh

i thought it made a roblo clone lol

[stone cape]

my minecraft clone was one shot

too

[stray aspen]

nah but the minecraft clone was nice

[stone cape]

i also made chess

[limber panther]

that was veery good

it added shaders lol

[stone cape]

chess was a bit flickery cus of my laptop

but it worked amazing one shot too

[limber panther]

wtf

thats so good

how is it so good bruh

[wicked sage]

f'ing CHESS???????

this is peak bro

[limber panther]

man i need to use it

where are yall doin this

[stone cape]

pineapple needs to add this in arena

movementlabs.ai

[limber panther]

its website?

alright imma see

[stone cape]

use hawk max

[limber panther]

oh ok

best model i guess

[stone cape]

tell us what u think opus killer?

[wicked sage]

literally what i said lol

[limber panther]

imma try a prompt

but im mobile so idk

[obsidian hawk]

try asking to make a game for mobile

🤓

[limber panther]

hmm

what game is hard to make

for mobile

[obsidian hawk]

minecraft mobile 💀

😂

[limber panther]

oh damn

imma try that

[obsidian hawk]

bro if that works 🫃

wana see what it cooks

[stone cape]

whats ur prmpt

[limber panther]

is good?

[echo aurora]

When it comes to models that are/were using codenames I'm unable to give out details that confirm nor deny information about them.

[echo aurora]

I couldn't comment on the specifics, but overall labs are encouraged to reach out to our team if they're interested in evaluations.

That may happen, but just depends. But overall, other information related to it I'm unable to share specifics on.

[obsidian hawk]

anything ye?

[proven bronze]

Did they add the filter to all accounts now?

He knows that that garbage throws up false positives, but they don't give a damn.

[limber panther]

i got this

pretty good graphics, it has chunk generation tooo

fly mode

and controls are good

[obsidian hawk]

no fkn way

[limber panther]

its better than opus somehow

[obsidian hawk]

does it work tho

the game

[limber panther]

yeah no bugs

i did everything

[obsidian hawk]

[limber panther]

its kinda wild

it dont have issues

[stone cape]

https://tenor.com/view/cat-no-nonono-noooo-cat-no-gif-13597132752790194338

[limber panther]

im not on my laptop right now

i cant test big stuff

[limber panther]

is it good for u

[stone cape]

can you paste code here

so i can try it on my laptop

[stone cape]

@stray aspen

[limber panther]

its quite big file

i have to make a .html file

[stone cape]

i made a amazing version on laptop

i wana try more prompts

im on laptop so any ideas

[stray aspen]

insane

[limber panther]

its my version i said to make it mobile friendly

it doesnt have strong memory consumption

pretty much better than 4.5 opus right now

[limber panther]

can yoi try "an impressive black hole simulator in three.js"

[stone cape]

yeh

[limber panther]

thanks man

[limber panther]

@stray aspen what are u making

[stray aspen]

nothing

i ran out of prompts lol

[limber panther]

make new account

[limber panther]

damnnn

ts is so good

if i tried this on my laptop

it would fry itseld

[stray aspen]

thats actually insane lol

[obsidian hawk]

movementlabs hawk model got overloaded yesterday

[stray aspen]

its so cool

[obsidian hawk]

maybe same happened today

its still a new company

[limber panther]

the model is so good so fae

i hope they get better

[stone cape]

i dont know what they feed to it

@echo aurora srry for tag u guys gona add this?

[echo aurora]

I won't be able to go into details, but overall labs are encouraged to reach out to us if they're interested in evaluations.

[stone cape]

ah ok thx bro

[stark arch]

<@&1349916362595635286>

[stone cape]

scammer here

[stark arch]

that was fast

[stone cape]

my guy pinged everyone

[obsidian hawk]

@safe sleet and his crew doing chargeback fraud on movementlabs 😂 this was from a few days ago thought i'd post it here for some laughs

[burnt sinew]

what is even the point of bernie

[wind shore]

Ernie is so sweet. Such an awesome bot.

[burnt sinew]

worse than a 10 month old model

[burnt sinew]

the duality of man

[stone cape]

is this real? and they posted this publicly?

[obsidian hawk]

yeh coming from the same people who called mlabs "scam" 👴

its in OR group they banned me there, the admin ( @tall vapor ) there is in favour of the guys doing public fraud on his server .

[stray aspen]

🥀

well those 15 hawk max prompts didnt disappoint

[obsidian hawk]

better than claude limits 😂

4 prompts then wait a week

🥹

[obsidian hawk]

glad to see you on the good side

[stone cape]

server owner allowed this?

danmm

[safe sleet]

Why am I being pinged out of nowhere to be accused of fraud? Lol

[tall vapor]

you continue to troll, not going to engage. The user you’re talking about was in contact with Stripe and they told him to dispute with his bank. You got banned for throwing around insults and existing solely to promote movement labs.

[safe sleet]

To be clear, I have never made any payments to the service mentioned

[obsidian hawk]

@tall vapor will be reporting to discord too

i have documented everything.

[stray aspen]

holy drama

[obsidian hawk]

"This user" teaching fraud on your server got to stay? and i got banned for being a troll?

[obsidian hawk]

stripe told him to dispute? stripe tell you to contact the provider first, stripe does not like chargebacks 😂 FYI

[echo aurora]

Hey going to ask we move on from this conversation. Seems like there is some off-server drama that is spilling over here. This isn't the place to be discussing it.

[keen lake]

https://tenor.com/view/surprised-sorprendido-shaquille-oneal-gif-7318718906883567382

[keen lake]

when on arena?

[stray aspen]

probably not soon

[burnt sinew]

@echo aurora did someone do an oopsie with the graphic?

[stray aspen]

theres too much drama surrounding them

[burnt sinew]

text to image... seedance... hmm

[echo aurora]

Removed the announcement and will look into. Thank you.

[echo aurora]

Good catch. Thanks again.

[jade egret]

what model is that?

[keen lake]

hawk max on movementlabs.ai

[jade egret]

oh, is it good?

[keen lake]

lets try it together

in 8 min

[marsh hedge]

bro why is this bot so strict

im not even generating anything nsfw onb

[obtuse smelt]

oh yeah i look grok in website is down

[jade egret]

so is it good?

:0

better than like gemini 3 pro, claude 4.5 opus, or gpt 5.2?

[zenith berry]

Why do I keep getting errors and limits from using Claude guys

[wind shore]

Claude is unfortunately limited.

[zenith berry]

Are there any other good AI recommendations to help with coding?

[keen lake]

heck YESSSS

ITS BACK UP

[manic current]

use https://antigravity.google/

[zenith berry]

do i have to download to use this?

[terse shuttle]

Yes, it's IDE and has free quota that resets every 5 or 7 hours (I don't remember exactly)

[jade egret]

is there any official benchmarks? Also it only good at coding right?

[plain adder]

hello

:D

[rose sky]

Umm... should I put my hand in there???

[radiant heron]

is hawk model good

[radiant heron]

goat software

[radiant heron]

make sure you download from actual legitimate site and don't download anything fishy

LMarena being valued at 1b< is crazy

[hexed drift]

hello

[rose sky]

😂😂😂

[cloud oak]

-# there are guys called Callum too.. 😶‍🌫️

[rose sky]

I made this

[harsh wolf]

Guyz, why when i try to generate claude opus code after 1-2minutes "something went wrong"

[wicked sage]

hi guys

[wicked sage]

what is wan on about

[wicked sage]

real
cursor is much better if you pay for it (im pretty sure)

[wicked sage]

mostly rate/usage limits

[rose sky]

Why is it SO obvious that these are AI-generated? Because below the images looks like AI text-to-image prompts, because of the way they're structured. Am I right?

[lapis imp]

Hi anyone know good websites for image to 3d model?

[tall tulip]

Guys

[keen lake]

hawk max

[tall tulip]

When u get this results, is this reducing your limit quota? I usually get up to 5 image per hour, now probably only get 2 and other 3 is like this

Help?

[quiet ginkgo]

👋

[wicked sage]

<@&1349916362595635286>

[stark flax]

Hey how do I submit a model to LMArena, I've made a model which combines every model, and it outperforms Gemini 3 Pro by 5.5% on my HLE tests

[silent orchid]

hi

[stone cape]

"i made a model" - comical

nah movementlabs is goated now

its outperforming opus

#general message

#general message

#general message

#general message

[placid geode]

https://tenor.com/view/ace-ventura-jim-carrey-raised-e-yebrow-gif-6338947979135718753

I want benchmark

[stone cape]

run your own eval

[exotic musk]

#ai-creations

[stone cape]

iquest coder if you want benchmarks @placid geode

[placid geode]

Why you need to say iquest coder

[stone cape]

just to prove how a crap model can benchmaxx srsly do people follow benchmarks?

i would rather use a model with no benchmarks that does not f up my code dude

💀

[keen lake]

I can vouch that MovementLabs is indeed an Opus competitor. Not sure about the other line, but the Hawk series oh my god, it's incredible

[hushed yoke]

hi

[stone cape]

made with hawk

[rigid copper]

sup guys

#january-1st-contest message

<@&1349916362595635286> someone posted something that isn't related in that channel

[placid geode]

Bro who have dataset 💀

[astral elk]

Hi

Good

[stone cape]

g3 pro or hawk?

wana hear ur thoughts bunn

[placid geode]

I mean dataset
You don't know what is dataset?

[stone cape]

github

most of them are public run eval bro

if you want benchmarks

[placid geode]

All models all trained on these dataset

so useless

[placid geode]

also you need paid bit money for this

[safe sleet]

#1397655624103493813

[cloud oak]

this server definitely needs more mods.. 🥺
-# eatventure has 1/8th of the members here and 22 mods in total..

[hollow ivy]

I believe, in the future, AI/AGI-mods will be even better/faster than human mods.

[golden ocean]

this already so easy to do right

but now it costs 943873982398902398298$ because constant token input in an environment like discord

in future less i hope

[fiery gull]

wth 5k problems

+5k

[shy saffron]

how to create more than 5 videos?

[ocean vortex]

Have you actually ran the entire HLE benchmark?

You are not gonna get featured on lmarena obviously, but custom router implementation still sounds kinda interesting

[dreamy ivy]

whats up big bois

[short stirrup]

how make tex to video ?

[dreamy ivy]

what is hawk

[dreamy ivy]

i don't think everyone has that feature. i also don't have it, it doesnt show up to me atleast

[short stirrup]

😒

[ocean vortex]

#video-arena-1

#1397655624103493813

Mostly using this for compiling the leaderboard currently, although I'm not quite sure how they weight the number of votes

Since you can have 1 user voting vs 2-3 users voting for the same video

[dreamy ivy]

where is it in the browser one i don't see text to vid in lmarena options

[ocean vortex]

It's not there, not yet at least. In the testing stage not a public feature for now

Now it's through the bot

[dreamy ivy]

oh

[ocean vortex]

Most likely each individual is considered as a separate battle come to think of it. Treating it equivalent as different users inputting the same prompt and getting the same outputs but voting individually for separate chats/battles

[dreamy ivy]

whats automaticaly detect modality

[ocean vortex]

That will select auto what it thinks you need. If you say 'generate an image of...' it will enable image modality etc

[dreamy ivy]

hmm okay

[keen lake]

opus killer

[keen lake]

try it movementlabs.ai let us know ur thoughts

[wheat coral]

https://cdn.discordapp.com/emojis/1102784149808484433.png?v=1&size=48&quality=lossless

Hello

[safe sleet]

It's complicated™

An AI that does this will inevitably be rated lowers by humans and likely lose accuracy overall by erring on the side of caution if it's not sure

Would love an API model from someone focused on hallucinating less, though

[loud crag]

Phrasing is everything. If instead of saying no, it could internally rate itself per query and give you a confidence score, it'd be a lot better.

A confidence of like 5% would be better than a straight no, so you'd still get an answer but you'd be able to safely ignore it

[ashen epoch]

Hey guys, question, is there a way so I can train my model so it stays consistent?

[ashen epoch]

@hardy swallow @quasi atlas ?

[thin current]

Guys, when will you add the ability to upload code files to avoid cluttering the text chat?

[worthy bluff]

this happening to anyone?

[rigid copper]

sup guys again

[dense sphinx]

My lmarena not working.

Hello everyone

[vapid lake]

Hello! I've been using LMArena for a while now, and am still discovering new stuff about it.

[surreal frost]

I absolutely love how their service/website here is such a mess, breaking left and right, doesn't even have a proper history trimming system when a model runs out of tokens, has this insanely aggressive captcha system that puts u on a loop for 5 hours straight and these guys are here, sitting around, celebrating, clanking beer bottles together like their service isn't on fire

I'm sure I can't be the only one

[wicked sage]

not to mention, its more cheaper than like any ai

its like 6.5 euros i think

[wicked sage]

rate limit possibly

ello ananas

[keen lake]

cheaper

[wicked sage]

thats yearly i think

[echo aurora]

Can you check out this help center article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message There are some steps that may help with this error, along with a way to report more information about what is going on. It's most common that this being caused by the rate limit, you'll want to not prompt as often.

[wicked sage]

lets see monthly

give me a second

[wicked sage]

wait no im stupid this is only monthly

[wicked sage]

i was talking about the plans 😂

[echo aurora]

What seems to be the problem?

[keen lake]

$6 monthly 😂

[wicked sage]

to be fair, for a opus competitor thats insane

[keen lake]

thats practically free

[dense sphinx]

@echo aurora Sorry. Last moment lmarena not worked

Let me check

[wicked sage]

you can get a claude pro plan for the price of like almost 4 movementlabs pro plans

[keen lake]

i just purchased $300 in api credits

[wicked sage]

absolute cinema

[keen lake]

im going to use hawk max to build an IOS app

[wicked sage]

gl bro, i believe in you

[dense sphinx]

I've encountered 404 error then it worked.

[wicked sage]

i hope you get money from the app (if you finish it)

[dense sphinx]

Sorry for bothering

[keen lake]

im gona make a calorie tracking app with ai integrated hobby project

[wicked sage]

in this time? a calorie tracking app is perfect

you should also add an ai button so you can talk to the ai and stuff

wait idk why i said that i see ai integrated

Lol

[echo aurora]

Bugs/issues with the site, along with developing new features, is something the team is very focussed on. We are striving for a great experience when using the site. Regarding the captcha this is something we're continually monitoring and will make appropriate adjustments to. I can't go into details, but overall this is going to happen if the system detects any behavior it finds to be inauthentic. Prompting too fast/often can contribute to this. If you are doing things that the system thinks is inauthentic use this is likely contributing to why you're seeing a captcha.

[wicked sage]

btw do you have an idea in mind

like a ui idea, or something

[keen lake]

is this got to do with lm arena to api wrappers?

[echo aurora]

I don't have an ETA I'd be able to share, but this is very much on our to-do. It's clear to us why more file upload options would make the overall experience better.

[dense sphinx]

Thanks pineapple

[echo aurora]

It's working again? What was the problem in the first place?

[echo aurora]

Sorry I'm not following, can you elaborate?

[wicked sage]

hey pineapple i got a question, i think this is internal but how do you guys get the money for the ais/api tokens
and also i randomly found this

[dense sphinx]

Yes first time it doesn't worked I thought it was technical issues. But moments later it worked. Was it maintenance? @echo aurora

[echo aurora]

In Direct and Side by Side those chats should remember the context.

[keen lake]

oh on github i saw some stuff where people use cookies to forward requests to openai compatible endpoints buring your usage quota

[echo aurora]

No not that I'm aware of. Can I get a better idea of what wasn't working? Were you seeing an error message?

[wicked sage]

nvm im stupid idk why im asking

[keen lake]

@echo aurora

[surreal frost]

I hate to be that guy... but I've made my own chatrooms before (I'm known among my friends as the AI guy) none of the issues you guys have, are anything huge and can be fixed very easily, directly in javascript, in under like 5 minutes.
The history trimming, the captcha error, all can be fixed easily. In fact the current chatroom that I've made myself doesn't have this issue.
And yet this is an issue nobody has fixed even after months. Maybe years. And you guys have the audacity to celebrate your funding..

[keen lake]

these people have put in a lot of work to reverse engineer , shame becuase its affecting normal users like me

[dense sphinx]

@echo aurora It's just battle button not worked but that issue is fixed. And also models disappeared. But again it works perfectly. Sorry for bothering.

[echo aurora]

Have you been noticing unmoderated situations? My experience this hasn't been the case. But very curious if members of this community feel differently. We can scale up more moderations if needed, but in my experience we just haven't needed this yet. But again would love to hear what you all think of this.

[echo aurora]

Can you point out where you found this?

[echo aurora]

It's no bother at all. Keep me updated if things stop working again.

[keen lake]

i'll dm you cus dont want people to abuse

it works i tested it myself

[dense sphinx]

@echo aurora Definitely. Thanks.

[keen lake]

sent

[wicked sage]

wtf, how?

[keen lake]

yea basically unlimited free usage piggybacking of any model on lmarena

[wicked sage]

0_0

..ok

[keen lake]

lmarena pay while people vibe code

models like opus could burn thousands of dollars a day

im a cyber security expert and i stop things like this

i helped patch cursor ide to api endpoint

[wicked sage]

well, thank you for helping lmarena!

[keen lake]

they used mitm and sharksniffer to capture ide requests and mimic locally and expose local running npx to server open requests world wide using 2 packages.

[wicked sage]

what in the world is mercury

What.

[echo aurora]

history trimming
This is an interesting callout. I don't believe this has been recommended before. I wouldn't be surprised if the devs are exploring this already, but regardless I will pass this along. Anything that could help reduce errors and make the experience better we'd like to implement.

the captcha error
Our captcha system is very important as it helps prevent bad actors from using exploits to cause harm. This is a system we're going to continually make adjustments to over time and use data and community feedback to drive what changes we make.

[wicked sage]

i found another one, im not even sure where you can talk to this model since its not in direct chat

i think its not there

im checking

[echo aurora]

Can you send our team an email at [e-mail]?

[wicked sage]

:shrug:

[keen lake]

WOW, you guys eval this kind of stuff?

then whats the controversy around mlabs😅

[surreal frost]

history trimming, usually when I send large amounts of texts to ur site the models fail to respond and refuse to respond anymore till I make a new tab. Me and a couple friends assumed it was cause the history length was too long.
Other sites cope with this by history trimming. Which till now I still don't get why you guys don't have.
About your captcha system, it stays around even when I take rest before making another request. My chatrooms have this as well, but the difference is my system isn't as aggressive as yours. There is simply no reason for it to be this aggressive either.
I also see a bunch of complaints to u guys about these very often, and I also see you guys usually giving them an AI generated response and calling it a day or something generic like refreshing the page.

[echo aurora]

Yeah check out this blog for more info: https://news.lmarena.ai/ai-evaluations/

[echo aurora]

large amounts of texts to ur site the models fail to respond and refuse to respond anymore till I make a new tab. Me and a couple friends assumed it was cause the history length was too long.
Yeah this is a great callout. I've shared this with the team already. This could be a really helpful tool to keep the experience more stable.

[thin current]

Thank you

[surreal frost]

If you're actually considering that, can you consider adding an option to edit messages and regenerate so I don't have to make a new tab everytime, or need to make a new message anytime I type something wrong in my message

[barren sorrel]

Any fix for this? Stuck like 4 hours ago

[surreal frost]

there is no fix, unless you haven't tried reloading yet. otherwise this means your tab is bricked now.

[wicked sage]

https://media.discordapp.net/attachments/1006341705924087919/1200758925205708870/caption-2-2.gif

hi guys

[echo aurora]

edit messages
Yes, this idea has been floated before. It's an idea that's on our radar, but I don't have any specifics I'd be able to share for if/when something like this would be implemented. I imagine a difficult challenge is figuring out how something like this would work in Battle mode, or in the other modes when changes from one model to another. I do think we have unique challenges seeing we do side by side comparisons, but it's an interesting idea worth exploring.

[barren sorrel]

I've already rebooted it a dozen times, and I had about the same problem during the day. But she somehow corrected herself.

[surreal frost]

just regenerate the 2 message bubbles again whats so difficult

[surreal frost]

definitely bricked now, either wait about a day for the devs to notice that ur tab got bricked (or make a report) or simply start a new chat

[echo aurora]

Really sorry to hear you're getting this bug. I have seen cases where hard refresh will help. However, that's not always the case. Sometimes creating a new chat is your next best option.

[keen lake]

OPUS 4.5 vs HAWK MAX what a battle shame i had to use my own credits

Edit: Hawk Max actually made a 3d game, opus was lazy and went for the 2d style.

[barren sorrel]

There's a problem here, you can only upload images on your site. I have written a text in about 1 million characters. I was able to download it via TamperMonkey and an additional screech for LMArena. And then what? I could create a new chat and upload a file with the history of correspondence, but you can't do that. That's the big disadvantage.

[wicked sage]

hawk might genuinely be better than opus LOL

btw how many credits did you use for this

[keen lake]

hawk max free, opus $0.13 approx

[wicked sage]

ah

also i thought the hawk max part was where it used the credits

[keen lake]

nope, its free on frontend i have claude max sub they charge opus not free on claude its in paid teir

[stray aspen]

both cooked

but its nice the movement labs made it 3D

[hot herald]

Just saying hello and happy to be around here

[fleet lintel]

Lot of statistics about AI traffic from https://x.com/Similarweb

They found out that people are really interested in AI stats and milking the engagement

Similarweb (@Similarweb) on X

The World’s Leading AI-Powered Digital Data Company. Listed on @NYSE: $SMWB

X

[echo aurora]

hello!

[keen lake]

id love to battle these two beasts in the arena

[sour spear]

GPT-image is driving me insane with its piss-filter-max-film-grain-aesthetics. Like, it adds yellow and tons of grain to almost every photorealistic image. 😩
Not to mention the makup and lipstick that's not supposed to be there.

[echo aurora]

I agree. Starting a new chat isn't really a "fix" if the stuck chat has a lot of context that's necessary for continuing the conversation. This problem of the model getting stuck is a big problem. Something like a Stop Button could help with this an is something I've been sharing the feedback on with the team.

[rigid holly]

The battle modes dual anonymous response is invading direct chat

Again

Oh how i did not miss that

[echo aurora]

This is a project we're currently experimenting with.

[rigid holly]

Can not say i appreciate it

It kind of sucks being barged by some random bot model in while storytelling in creative writing and it messing up the consistency it had

[echo aurora]

I am sorry to hear you're not a fan.

and it messing up the consistency it had
We are exploring this potential bug in #1458498310150164593 . I'm going to ping you there to learn more.

[sour spear]

On the other hand, this service is primarily intended to test AI models and give unbiased feedback, not as a means to have free access to everything.

[rigid holly]

Is that not the battle modes goal then? Like it already exists

Why mix em when the direct modes goal is direct communication with one singular model of ai at a time

[wicked sage]

hey pineapple. does gemini-pro actually exist in your database? or is it just NOT direct mode available

[echo aurora]

Our goal is to improve models through real-world community evaluations, so we are going to explore new ways that preference and feedback to be provided.

[echo aurora]

What do you mean by this? I'm not following.

[wicked sage]

@echo aurora

i was checking leaderboards

[viscid venture]

Which AI tool is the best for Virtual Try-on images?

[coarse current]

Nano banana pro

[viscid venture]

I tried, but it does not generate using the same outfit.

[coarse current]

Maby is your prompt

It's*

[viscid venture]

Should I use Nano Banana Pro in Gemini or Whisk?

[golden ocean]

https://media.discordapp.net/attachments/1006341705924087919/1200758925205708870/caption-2-2.gif

[stone cape]

https://tenor.com/view/brain-hurts-gif-12096381820083648834

[swift oyster]

Very nice work!

[cloud oak]

please understand that i did not write that as a criticism.. 🥺
i merely wrote that as a suggestion with concern of the workload and stress of the mods in mind..
sometimes i only sees 1 mod online and they have to take care of so many channels and all their needs..i just thought that more mods might mean the workload can be shared and the users' issues be resolved sooner.. 🙏

[random flame]

Is there a solution?

[clever radish]

how long does a video take to generate

[echo aurora]

I appreciate the concern and feedback. I wasn't interpreting it at criticism. We do want this server to be a well moderated space so hearing form the community if it's meeting those standards is important to me.

[echo aurora]

This error can appear for various reasons so it's difficult to say with certainty what is causing this. However, you will want to check out this Help Center article for some steps worth trying along with a way to report further details - https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

[echo aurora]

It should take around 3-10 minutes.

[random flame]

Yes, I looked at it and the problem was solved.

[random flame]

This solution worked
First, try these quick fixes...
If you see this error, please try the following steps:
Hard refresh the website
Clear your browser’s cookies and cache
Start a new chat
These steps often help, but they may not fix the issue every time.

[random flame]

Thank you

[echo aurora]

Excellent, very glad to hear it.

[golden ocean]

https://tenor.com/view/salute-cat-cute-yessir-gif-17449845613192073437

[tribal hamlet]

They are generating now. However, I do have one that I'm attempting to get a jumping jack but both are really bad. A basic jumping jack is starting upright with feet together and arms down, then simultaneously hopping feet out to shoulder-width while swinging arms up overhead, and immediately reversing to the starting position, emphasizing a light landing and continuous rhythmic motion. Do I need to resubmit this one with this prompt?

[marsh hedge]

im here once again

to complain

about the flagging

of perfectly appropriate prompts

[echo aurora]

Can you share the prompt (copy paste + include the image uploaded if there is one) in #1447983134426660894 ? That'd be a big help.

[elfin falcon]

hello

[echo aurora]

howdy

[echo aurora]

Yeah if you're unstatisfied with the results you can try again. I would recommend making minor changes to the prompt if you think that'd help.

[marsh hedge]

oooooh bet bett

[loud verge]

Guyz

Does the censor have to be this soft? 😭

[echo aurora]

If you could share suspected false positives in #1447983134426660894 that'd be ideal. I've documented this one, but if there are others flagging in that thread would be great.

[golden ocean]

https://tenor.com/view/get-blown-up-dog-dog-explosion-sticky-notes-gif-17071604041793579031

[loud verge]

I didn't read the tos but I'm assuming calling the model stupid shouldn't be a violation. I guess I'll read the terms now.

[echo aurora]

I won't be able to give interpretations of our Terms of Use, but it's possible that this filter is acting overzealous and blocking content that's a false positive, we are collecting these suspected examples and plan to make adjustments to the filter over time.

[hushed gyro]

BTW don't ping everyone

[echo aurora]

@lucid kite I'm going to remove that message for the @ everyone ping, but can you please provide this feedback in #1372230675914031105 ? I would ask that instead of having 1 thread with a bunch of different feedback, put specific feedback into the appropriate thread.

[echo aurora]

Why

[lucid kite]

My phone went crazy and I accidentally deleted the message from the post.

[acoustic orchid]

is there a way to swap things from battle to direct

[loud verge]

Also, is the option to share threads on the way?

[spark python]

pineapple

can u remove the random as thing

where it randomly gives u 2 models to choose from

but u are in a direct chat

[granite igloo]

I can't find video generation ai's on the website?

[echo aurora]

Battle -> https://lmarena.ai/
Direct -> https://lmarena.ai/?mode=direct
Let me know if you want a different modality

[acoustic orchid]

im on about moving my stuff from battle to direct

[echo aurora]

This is an experiment currently to see what the occasional Battle in Direct looks like.

[echo aurora]

It's currently an experiment meaning it isn't yet rolled out to 100% of users. It'll be random if you see the feature or not.

[gleaming roost]

[hoary verge]

Yours know AI for TEXT-TO-SPEEAK, for video long in english to portuguese speak?
i have vídeos in inglesh, but i need for Voice video in português
Free please
?
🤓
Very Very vídeos long (1 h)

[echo aurora]

More information about how to use the Video Arena bot can be found in #1397655624103493813. There isn't a way to make the length 1 hour, it's going to be closer to 6 seconds.

[stray aspen]

1 hour is crazy lol

[surreal creek]

I put in a serial number of the dollar bill I had in my wallet

WHAT is Grok talking about 💀 💀

[stray aspen]

LMAO

[exotic belfry]

hello

[hollow ivy]

careful.. CIA/NSA might track you now
/jk

[surreal creek]

Pfft 😂

[opaque current]

I've experience in building Dapps, Smart Contract, Dex Development, NFT platforms, web3 games, AI Agent, Mobile apps..
If you're building such a project, please let me know..

[half plank]

Security footage view. EVENT: Person slips slightly then suddenly looks behind with exaggerated reaction, then shrug and walk away. 6 seconds, comic timing.

[fiery gull]

wtfffff lolll

[loud crag]

I am so extremely good at using a prompt that requires 2-3 images, and then half a second after hitting send, realizing I forgot to upload those pictures and wasting an input... 😅

[round oracle]

@echo aurora Please remove this.

[echo aurora]

Sorry had to delete this, was a little too NSFW for this server.

[echo aurora]

You'll want to review the info in #1397655624103493813 for a better understanding on how to use the bot.

[hazy forge]

are claude opus 4.5 coding and gemini 3 pro coding broken?

mind you, this a recommended test prompt

[echo aurora]

I can't say I'm seeing these models error out at a higher rate than usual, so doesn't appear to be widespread. Can you try the steps in this article and keep me updated if it helps?

[hazy forge]

sure, give me a sec

@echo aurora hope this helps

[echo aurora]

Thanks, I do have the jamdev but wanted to remove so others don't have access to that.

Okay yeah so this looks like this error is happening because of rate limitting. Essentially, the model is being used too often and is going to result in this.

[hazy forge]

mb you're right, thanks for that, next time ill send you a dm or write a ticket or whatever system y'all have here

[hazy forge]

understood, thanks a lot

have a nice day

[echo aurora]

Sounds good, would note that in that same help center article is where you'd want to submit a jamdev. There is a google form towards the bottom.

[echo aurora]

Hope the same for you.

[brittle garden]

hi

[maiden fulcrum]

how can I use direct chat for video models on lmarena?

[stray aspen]

You dont

[quartz light]

💀

[echo aurora]

Video Arena is just going to be in Battle mode

[fickle venture]

Is it fully out now? And do you expect to make it direct in the future?

[echo aurora]

No it's still in an experiment, we will put out an announcement when it'sfully launched.

It's possible we add Direct Chat, but can't make any promises.

[modern trellis]

Looks like you've hit today's video limit. Check back in 23 hours.

After creating 2 videos lol.?

[echo aurora]

Hmm looking into.

[modern trellis]

I just voted 2 videos . And I got limited

[echo aurora]

I'm only seeing 1 prompt request from you. Can you link me the message where you're being told you've reached the daily limit?

[modern trellis]

Im on the website

Im talking about the website

Any updates about this? Is there a limit of 2 generations per day on the website or its just for me?

[echo aurora]

Oh gotcha. It's limitted to 3 per 24/hr. I'm not sure why it was only 2 for you. Are you able to send a screenshot of this?

[modern trellis]

Ok

[gritty arrow]

is it 3 on the website and 4 on the discord?

[echo aurora]

5 for Discord

[gritty arrow]

so 8 in total?

[stray aspen]

hell yes

[tight oriole]

hedllo

[next thistle]

Vidu i think mochi API right?

I think Mocha API right?

[next sinew]

how do i generate a video longer then 11 secs

i want a 1 minute long video

image to vid'

[stray aspen]

You dont

We are limited to like 6 or 8 seconds

[next sinew]

ugh

[thick lantern]

Yeah. Its just so unsatisfying. Feels like im playing guessing game.

[rose sky]

Image -> 3D Gaussian Splatting

[gusty rain]

Forex

[echo aurora]

Going to share this in #ai-creations hope you don't mind

[royal coral]

Hi

anyone know why my nano bananas image quality got worse and worse after each prompt ?

from here to here now 8-9 prompts later

[night moat]

Why bot can't generate zombies video?

[lofty frigate]

Did the limit of gemini 3 just became shorter?

[whole sundial]

it's sending the same image that it outputted back to itself to edit, that's why it has the weird watermark. also those green spots are google's "invisible" synthid watermark, but since you used the same image many times the watermark applied onto itself multiple times, causing the watermark to be plainly visible.

[wicked tapir]

is it just me or is the title of the lmarena.ai website changing from reload to reload?

sometimes it's "LMArena", but sometimes it changes to "AI Chat Arena"

are we looking at a rebrand or something?

[wicked sage]

because bread tastes better than key

[wicked sage]

why is it in japanese???

[whole sundial]

i think grok 4.1 mini is heavily trained on japanese data, it can do its thinking in japanese as well which is probably why it answered in japanese

[timber berry]

how to solve this?

[whole sundial]

try to clear your cookies? maybe try another browser?

[bitter mason]

Hello.......Good Morning friends.

[near kindle]

hello, happy new year

I am expereienced developer and currently looking for opportunities I can apply my skils

if you are interested , feel free to reah ou anytime

[wintry jackal]

What username am I supposed to put into Copilot Arena? I need some help, I tried like 20 different usernames

[whole onyx]

Will video arena ever remove rate limits as long as ai inflation died down?

[wintry jackal]

I Copilot Arena still a thing? I mean is there a way we can still use it or is it a dead project?

[empty stump]

How can we remove these watermarks

[whole sundial]

not use the gemini app to generate images, or use third party tools to get rid of them

or maybe try to remove the gemini watermark with another model

with synthid, it's harder because it's meant to withstand compression but avif compression may do damage to it

[empty stump]

The steganography can probably get removed by like qwen image edit

But its done in such a way that image quality will be destroyed because need a model of same level

[meager oxide]

@echo aurora hey is there any way i can rename the chat?

[whole onyx]

@whole sundial any ideas what to do with ai image aside from ai comics?

[sour spear]

Open the image in Google Photos and use magic eraser. Bit clunky, but works like a charm.

Orjust crop the image a little at the bottom.

SynthID isn't an issue, unless you want to spread misinformation. No shame in admitting an image is AI generated. 😉

[whole onyx]

Any fun ai image prompts aside from making a comics?

[light hinge]

helloo

[sterile tartan]

That's What i exactly do

[lofty frigate]

Can gemini 3 have the same limit rate as reve 1 cause bruh i just generate 4 times and hit the limit

[ionic stump]

Hi

[wide vector]

hello all!

[obtuse smelt]

hi

[surreal frost]

what the hell

is this not an AI chatroom server

the hell are people doing

are these guys lost or just bots advertising random things

you seem to be one of them

[rose sky]

wdym?

[surreal frost]

what are u sharing

[rose sky]

I'm just sharing videos

[surreal frost]

and modpacks apparently

or shaderpacks

ig

this is not a minecraft server

idk if u noticed dude

[rose sky]

Yeah, I now

know*

[surreal frost]

why are u advertising ur shaderpacks on an ai discord server

[rose sky]

I'm just sharing them, btw, I already removed them

[surreal frost]

you confuse me dude

[rose sky]

Yeah, I removed them because your reaction sounds like that it's wrong

[surreal frost]

People come here for AI related stuff not your shaderpacks

[rose sky]

Yeah, ok, I won't share 'em again

[golden ocean]

thanks that was pretty interesting
i like your Real-time RT (Ray Tracing) with a 128x resource pack ngl
I just came to this chat for that

[surreal frost]

lmao fr

[high ginkgo]

Nice shaders I like them

[rose sky]

Sorry for the late reply, I was having dinner, but, thanks

[lofty frigate]

Am i the only one that lmarena just randomly exits the site?

[misty vault]

cool shaders bro
im happy i stumbled accross this footage

[magic pond]

Is there a way to pay for more credit ?

[whole swallow]

What was the URL for LM arena but for medical papers/research

[tidal grove]

Why is GPT 4.1 so high now?

[meager oxide]

is there any way i can rename the chat in LMarena?

[tidal orchid]

There is no way to do it. I think so.

[placid skiff]

a large model and a small model are given 60 seconds to reason, which develops the better result usually?

assuming the small model’s tps vastly outweighs the large one, thus generating more reasoning tokens overall

[stray aspen]

is that a woman with a beard

[tidal orchid]

Only 5 images generation on nano Banana?

[golden ocean]

among us

[robust sluice]

why the counting number was removed ?

[tidal orchid]

Means don't understand bro?

[delicate cloak]

lol

[golden ocean]

lol

[steel dune]

Hi everyone, please is ChatGPT down by any chance? I can't seem to access it from my end.

[echo aurora]

There is not

[echo aurora]

What seems to be the issue? Is the model missing?

[echo aurora]

There is not, but overall improvements to the chat history are improvements we plan to build out

[topaz stratus]

helllo

[wintry jackal]

Copilot Arena for VSCode still works ?

[frail bluff]

WHERE IS IQUEST CODER?????? HUH???? IT SHOULD BE RIGHT AFTER CLAUDE..... SO MUCH FOR UNBIASED BENCHMARKS....

[obtuse smelt]

?

[frail bluff]

check this on 24.00 https://www.youtube.com/watch?v=7kPNA86G_GA
model is here https://huggingface.co/IQuestLab/IQuest-Coder-V1-40B-Instruct

[obtuse smelt]

hmm is look Something went wrong with this response, please try again.

[steel dune]

It’s not responding to prompts

[echo aurora]

Can you try hard refreshing the page and clear cache? This may help move it along.

Unfortunately this is a known bug.

[echo aurora]

I’ll flag to the team. Would note #1372229840131985540 is a great place to make these requests.

[steel dune]

Thank up

[marsh hedge]

im hungry dawg