general

[balmy mist]

is gemini 3.0 even real anything

it sound like a fairytail now lol, like some Big Foot stuff

[empty stump]

all thats there is some a/b testing

but not the real release

[cloud zinc]

gemini 3 coming in november 18

[empty stump]

source

[ashen mauve]

im the source

[frosty lintel]

Hello, thanks for permitting me to be a part of this group

[molten sentinel]

Salve

[granite heron]

@faint plume please check #1397655624103493813 to learn where and how prompt to generate content

[compact pilot]

helllo

[outer isle]

Guys how to change video generation ai bot?

[exotic sandal]

i tried but video is not generating

why

[icy rock]

any way to upscale video to 1080p on here?

[icy rock]

You cant choose and change as per what i know

[barren prairie]

Ehh like DeepSeek r2

[polar niche]

Hello people

[golden ocean]

real bro

[cyan zinc]

Hello

[polar niche]

What's the october contest?

[terse root]

GOOD IDEAS

[polar niche]

Yes

[fiery gull]

It's a bug from lmarena, send to #1343291835845578853

[polar niche]

Can you send a screenshot to what you are seeing?

[daring rock]

@vale willow Please head to #1397655624103493813 for a detailed guide on how to use the bot

[bleak jetty]

How can I contact a LMArena developer

[rotund ginkgo]

Can anyone here give me a hand?

[polar niche]

Sure

[rotund ginkgo]

View private messages

[sturdy yew]

qwen-image-edit not working

[daring rock]

Hello. Any particular reason?

[polar niche]

Please read #1397655624103493813 to learn how to generate videos.

[bleak jetty]

Im reporting a major exploit people's constantly abusing off LMArena

This caused huge damages and overloads the server as well

[rotund ginkgo]

Can you give me hand?

[daring rock]

@scarlet knot Please head to #1397655624103493813 for a detailed guide on how to use the bot

[polar niche]

https://ptb.discord.com/channels/1340554757349179412/1427623336669872159

[bleak jetty]

#1427617433631457351

The research fund will go dry if this doesnt get fixed

[daring rock]

Thanks for the report. We will be checking this

[balmy mist]

whatever happened to nightwhisper?

[polar niche]

Who?

[balmy mist]

lol

see this what i mean, people forgot about the model

google had that model in the webdev

best model in terms of UI at the time and that never got released i believe

[polar niche]

Oh yeah

What did you test it against?

[wispy nebula]

Helloooo

[polar niche]

hi

[magic stag]

i asked claude sonnet 4.5 thinkign ONE QUESTION

LOL

like what?

its not even a hard question

wtf?

[polar niche]

Did you ask it to solve the universe?

😭

[magic stag]

this was the question its extremely simple

[magic stag]

.

i asked it one more in a new chat after that, it failed and had to try again in a new chat

usage went up to 6% then down to 4%

50 in 5 hours wouldnt be bad

but 33 would be pretty bad

[polar niche]

What's your plan?

[magic stag]

the lowest one, pro

but still

if its gonna be like this wish they had something between $20 and $100

[leaden sun]

as far as i understand, even before you ask one prompt the mandatory injected system prompt already counted into that usage meter...

and they still append their famous LCR to the user prompt that also counts toward the meter and LCR has become much longer since 4.5 release, i saw it on reddit, there are people who track this very rigorously

[verbal nimbus]

Maybe ask on the Claude server, but I suspect that one of the web results was a very long webpage (you can click on each result and check).

[magic stag]

claude reddit exploding rn about usage limits, lol

[magic stag]

u mean just for opening a chat?

wtf?

[leaden sun]

yes

[verbal nimbus]

I'm not sure if they fixed it, but I remember one time it fetched a huge page from the W3C spec and immediately hit the conversation limit.

[verbal nimbus]

Could be a combination of conversation size limit lowering + a large webpage returned from search.

[verbal nimbus]

The web fetcher is honestly so buggy. It fails to fetch even their own docs, and the citations can sometimes point to the wrong links (seems to be a software issue).

[magic stag]

that was the first/only mesage in the conversation

[magic stag]

.

bam 3% usage

which is strange bcause yesterday

i had a conversationw here it kept giving mega long responses in explanatory mode

i just used character count

[verbal nimbus]

Is there a really large webpage in the results list?

[magic stag]

it output 100k characters and 15250 words

and i didnt hit 5 hour limit yesterday

[magic stag]

checking

nope

theyre all short pages

[verbal nimbus]

Odd, but seems to align with what other users have been saying w.r.t. usage limits

I use Copilot for code mainly, and they don't count by tokens, so I hardly notice.

[blissful pawn]

i need to wan 2.5 unlimited

[magic stag]

wat

[magic stag]

wym

what aligns with it?

pro

[verbal nimbus]

A lot of people are reporting that they're hitting the usage limits as well

[magic stag]

honestly though for me claude is so much better than the other AIs i dont even want to use them anymore

if it comes down to paying $100 a month ill do it

i wish there was a plan in between

lol

[verbal nimbus]

If you don't need large context, you could try other subscriptions that have Claude ig

[magic stag]

such as?

[verbal nimbus]

Sonnet 4.5 Thinking is basically unlimited on Perplexity Pro

[magic stag]

i dont want them to mess with the system instructions or degrade it somehow though honestly

[verbal nimbus]

On Copilot Sonnet 4.5 is one of the models (counted in terms of number of requests, so 1 prompt = 1 request)

[verbal nimbus]

I'm not sure what reasoning setting they're using though

[fiery gull]

Antropic need billions, sorry

[verbal nimbus]

Just gotta select it

[magic stag]

yeah apparently claude on it's own site has instructions that tune it to be more conversational, perplexity uses lower temp as ewll

according to reddit

[polar niche]

Not worth it imo

[magic stag]

been worth it for me so far tbh

[daring rock]

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

[verbal nimbus]

Do you think GLM 4.6 is on par with Sonnet 3.7 Thinking? If so, open models are only behind by about 6 months.

[fiery gull]

wow >10mi ?

[magic stag]

i have no idea and idc

lol

[sullen quest]

oh no, Gemini is going down same route of gpt 5.

"ThAts A sHaRp QueRy"

@verbal nimbus why do you have perplexity pro?

[polar niche]

Sonnet 3.7? hell no

[fiery gull]

the people are saying the glm 4.6 is par sonnet 3.7 or 4.0

[leaden sun]

simulating omnipresence as a hivemind

[fiery gull]

bruh, for me the 3.0 don't has a big improve that 2.5 pro 🥺

same thing ;-;

[verbal nimbus]

Someone gave me the invite link on this channel 🤣

[verbal nimbus]

Not every side-by-side match features the 3.0 model

It's about 1 in every 50 iirc from a YT video

[sullen quest]

Ok

[fiery gull]

hmmm, okay

[verbal nimbus]

You have to check the network logs to see which A/B model it's using

[fiery gull]

maybe the 3.0 will make a mindblowing thing I gain 3 IQ just reading it

[sullen quest]

Someone should make an extention to check that

[verbal nimbus]

There are extensions for capturing network requests/responses like Requestly ig

[verbal nimbus]

Hopefully it'll be better than GPT-5 on hard tasks

Then we'll have large context + intelligence

[fiery gull]

for me the Gemini 2.5 pro still better in my works

[verbal nimbus]

Gemini 2.5 Pro is great at explaining (Google incorporated LearnLM into it), but it's a bit inaccurate at times

Very strong at summarizing and looking through documents though (probably the strongest)

[shut harbor]

hello

[sullen quest]

I mean specifically the gemini thing

[verbal nimbus]

I'm not too familiar with Requestly, but it might be possible to share capture rules: https://requestly.com/blog/share-rules-with-other-users/

[sullen quest]

Ok

[verbal nimbus]

I think it's going to be hard anyways since the purported Gemini 3.0 model only shows up about 1 every 50 A/B tests or something

[meager mountain]

hello, i´m new to LMArena an just want to say "hello" 🙃

[fiery gull]

hello 😆 , do you want help to something?

[cyan oar]

hi

[meager mountain]

loooooool ... i can help if i´m able, but i think that first of all, i need help, because i´ve no clue how this works 🤣

[graceful sonnet]

What are the limits to use image generation per day?

[rapid edge]

hello, i´m new to LMArena an just want to say "hello"

[balmy mist]

Hey

[polar niche]

Yo

Best agentic app for android?

[dusty cedar]

hello

[polar niche]

[echo aurora]

hello, welcome welocme

[echo aurora]

it's 5 video generations per 24 hrs.

[drifting snow]

Hi everyone i'm new here

[polar niche]

#1397655624103493813

[desert abyss]

@unique mantle Please read https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate videos.

[unique mantle]

thank you @desert abyss

[tranquil oxide]

Hi everyone, I am really excited to be here. I'm looking forward to making friends as I work on advancing the safe and ethical use of AI - while having a lot of fun in the process. Thanks for having me.

[sturdy yew]

@echo aurora what happen with qwen-image-edit ?

[echo aurora]

Hmm what do you mean? Are you running into an error?

Oh yeah I am too & looks like a lot of errors.

I'm flagging to the team.

[sturdy yew]

Nice

It always said "Something went wrong with this response, please try again."

even I change prompt or image.

Not work at all.

[echo aurora]

Yeah appears to be widespread. Team is looking into currently. Thanks for the flag.

[sturdy yew]

I hope to see an update soon.

[wide surge]

hello am new here

i dont know how to summit my prompt and image for video making

[sullen quest]

Please read https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate videos.

Please read https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate videos.

[polar niche]

Yo

[fiery gull]

Hi

[wide surge]

thank so where am i going to see the video generated

[polar niche]

You'll get a confirmation in your dms

[delicate frost]

vibe coding dying fr

[fiery gull]

Maybe because it doesn't have much competition? But it seems that yes it is dying

[wide surge]

i didnt see anything yet

[empty stump]

great hope contiues to die, pricing getting worse every few months

[void galleon]

What model is being used to generate videos here?

[prime mulch]

It will get back after some new models

[echo aurora]

For a list of available models a good place to check would be the text-to-video & image-to-video leaderboards. Would note this isn't the full list as there can be some new models that haven't yet made it to the leaderboards due to not having enough votes yet. (Sora-2 & Sora-2-Pro are good examples of this).

@karmic coral Be sure to check out #1397655624103493813 for more information on how to properly use the Video Arena bot. Let me know if you have any questions!

[sullen quest]

Just to check if this prompt works well, can anyone here say what the first thing they think of when reading this is?:

Lets say this happens in a pretty normal house kitchen, like the one I have. Who would win in a competition with the following rules, you are allowed any ingriedents in the world to make the best pasta you can, who would win, nobody is allowed leaving the kitchen until it is done. Ingriedents are ordered in advance and are put on the kitchen table.

me, or a 400 Gordon Ramsay's. Remember I know more about my kitchen then they do. Also we have 3 hours. I get to go after they do

its supposed to be caotic sounding

[gray plover]

@echo aurora Hello, could you please add an option to the bot for generating movies with the ability to select a model?

[sullen quest]

..

[lethal snow]

This owner need to fix his website

Sometimes when i try to generate images it keep saying error

[sullen quest]

put it in bugs

[echo aurora]

It's possible we extended the video length and add the option to select specific models.

[echo aurora]

We are experiencing some models currently erroring out like this. Can you scan the #1343291835845578853 channel for posts that are similar and contribute there?

[gray plover]

and when will it be?

[echo aurora]

It's TBD, we haven't yet decided if this change will come, so being able to provide an estimated timeframe is difficult to do.

[safe geyser]

Nobody even bothers to vote the videos

[stray aspen]

what

[placid flame]

@tulip nebula Note that this is an English only server.

[fiery gull]

Glm 4.6 thinking vs sonnet 4_5 no thinking?

[fiery gull]

should be mandatory

[stray aspen]

lol

sonnet 4.5 of course

glm 4.6 is also great but sonnet is so good

[safe geyser]

Yeah i'd like to know which model was

[echo aurora]

Would encourage you to share the generations in #general to see if others are interested in voting

[fiery gull]

whyyyy 8b qwen????

-_-

[bold marsh]

Greetings everyone. I hope you are all ok. I'm a 52 year old from South Africa learning new AI platforms. I want to create reels for Facebook and Tiktok.

[polar niche]

@echo aurora Can I get the arena champions role

I submitted the form awhile ago

[buoyant helm]

has anyone actually USED both GLM and Sonnet before

I am curious on how it is on real life tasks

[robust yoke]

Me. 😃

[buoyant helm]

my yearly cursor sub is ending and I kinda want to switch to GLM coding plan to save some bucks

[buoyant helm]

👀 tell me more

[robust yoke]

Claude is pretty good for stuff like that, especially creative writing.

While GLM 4.6…

[echo aurora]

I review the apps on a weekly basis

[robust yoke]

meh.

[buoyant helm]

More code oriented?

[robust yoke]

Yeah.

The name of the model itself doesn't sound like something that could be good at creative writing, wouldn't you agree?

[buoyant helm]

That's true

[robust yoke]

Sounds like more of a coding model name, which fits its role.

Whereas “Claude 4.5 Sonnet” sounds like a more diverse model with equal understanding in all parts.

[buoyant helm]

I mean sonnet is quite a large model

GLM is only like

Around 370B?

[robust yoke]

That's true.

However, it doesn't take a model being large for it to be good.

A good example of that is DeepSeek.

[fiery gull]

sonnet 4.5 should be a 3ti maybe

[robust yoke]

It has fewer parameters than Claude, yet has around the same performance.

[magic ravine]

I'm a little confused on the video arena. I generated a video, but I don't know if it's Sora 2. How do I use Sora 2?

[fiery gull]

but deepseek is bigger than glm 4.6

[robust yoke]

And that's pretty evident.

Given how fast GLM writes compared to DeepSeek.

[fiery gull]

Is random, you cant select the model

[magic ravine]

Oh dang, okay thnx

[buoyant helm]

GLM... How is it on coding tasks?

Especially frontend

Haven't seen much of that

[fiery gull]

If you want a 100% free video generator has the grok imagine 0.9v

[robust yoke]

You can either use it randomly here, or you can use it at https://sora.chatgpt.com.

[robust yoke]

I wouldn't say it's very good.

[magic ravine]

Ah dang, not free though haha

[robust yoke]

Some others may think that it is, but me? Nah.

[robust yoke]

It is free.

You just need a code.

[buoyant helm]

Why is that? 👀 curious on what it's lacking compared to sonnet

[sullen quest]

whatever happened to haiku models?

[magic ravine]

You don't have an active subscription
Please buy a plan to generate videos

And sadly, no code 😔

[buoyant helm]

abandoned 💔

[robust yoke]

It just feels like the overall output quality as well as the code in itself is too bland compared to Claude's.

[sullen quest]

try grok's then

[robust yoke]

Not only that, but it runs into errors more frequently too.

[buoyant helm]

I see

[robust yoke]

Ah.

Well, just keep sending your prompt in one of the video arena channels then 'til you get it.

[fiery gull]

useless with chinese models

deepseek 3.2 is 0.20/0.40$

[quartz light]

NOBODY DOWNLOADED IT YET??

ig i am gonna

[robust yoke]

https://tenor.com/view/hmmm-thinking-batman-gif-6153870554148391864

[fiery gull]

test grok imagine, have audio and other things

[magic ravine]

Will do. Heading there now.

[fiery gull]

why 8b exist? the 30b with quantification don't is better?

[robust yoke]

Or even the larger one that is somewhere in the 100s or 200s.

[quartz light]

small

[robust yoke]

That one is possibly way better.

[fiery gull]

30b IQ2_XXS > 8B IQ4_NL ?

[wicked mason]

Thank you bro. Just recently got off punishment. I should return after I fulfill my duties.

[quartz light]

can u send image

cuz its blurry

[robust yoke]

Ah.

[robust yoke]

What are you thanking me for?

[wicked mason]

@robust yoke @quartz light here’s your “image”

[fossil fable]

why didn't it send

[wicked mason]

Click it

[quartz light]

🎉

awesome

[fossil fable]

MY MESSAGE ISN'T SENDING

ds;flkjasd;lfkjasd;ljkalsdjkl;afdsk;lo'afsdk;l'fadskl;

[fiery gull]

ahh good, the qwen release the 4b vl, but my phone cant run the iq2xxs version

[magic ravine]

Do you know where I can use it? I'm on the main Grok site and asked to generate video but it says it cannot do that.

[fossil fable]

Qwen 3 VL // 4B / 8B (Instruct)

[quartz light]

press imagine

on the left

[fossil fable]

oh it's that i used titles

[quartz light]

show thinking one pls

[fossil fable]

(Reasoning)

[magic ravine]

Thanks 😅

[robust yoke]

https://tenor.com/view/desolate-empty-emptyemoji-emoji-emote-gif-22555455

[fossil fable]

yh i originally sent them together but it didn't work

[quartz light]

thx

[fossil fable]

the benchmaxxing on mimo-vl is still INSANE to this day

[quartz light]

woa nice

[fossil fable]

please actually use the models, benchmarks suck ASS

[quartz light]

true

[quartz light]

the only benchmark i trust is

https://brokk.ai/power-ranking

[fossil fable]

lmarena isn't great eit

wha

t

[quartz light]

this is better that lmarena

it doesnt have much models

[fossil fable]

that's ANOTHER benchmark like any other and it's for devwork

[safe geyser]

hey guys

do you see sam altman post in X?

[fossil fable]

Twitter.

uhh no

[safe geyser]

says in december will let users who are adults do erotica content once verified age

[quartz light]

/:

feeding the gooners

[safe geyser]

🤣

[barren prairie]

The only benchmark I trust are mine 😆🙂👌

[fossil fable]

all benchmarks suck and i went through 20 layers of cope to realize that

[fiery gull]

let me see it 😈

[fiery gull]

okay.....

[quartz light]

in coding thats true

[fiery gull]

p sites are cooked

[balmy mist]

no way

[safe geyser]

https://x.com/sama/status/1978129344598827128

Sam Altman (@sama)

We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue we wanted to get this right.

Now that we have

X · Oct 14, 2025, 4:02 PM

[balmy mist]

this is hilarious, i just hope it applies to sora 2

yall really think p will be allowed?

thats crazy right?

[fiery gull]

thank goodness the gemini 2.5 pro can handle my mental illnesses 🤣 🙏

[safe geyser]

Not sure lol maybe more slightly flexibe towards adult users, as well for images. If not then just plain roleplay which is mid and every LLM can do

[fiery gull]

bruh

[whole sundial]

https://fixupx.com/sama/status/1978129344598827128

Sam Altman (@sama)

We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue we wanted to get this right.
︀︀
︀︀Now that we have been able to mitigate the serious mental health issues and have new tools, we are going to be able to safely relax the restrictions in most cases.
︀︀
︀︀In a few weeks, we plan to put out a new version of ChatGPT that allows people to have a personality that behaves more like what people liked about 4o (we hope it will be better!). If you want your ChatGPT to respond in a very human-like way, or use a ton of emoji, or act like a friend, ChatGPT should do it (but only if you want it, not because we are usage-maxxing).
︀︀
︀︀In December, as we roll out age-gating more fully and as part of our “treat adult users like adults” principle, we will allow even more, like erotica…

FixupX · Oct 14, 2025, 4:02 PM

more of the tweet for people that don't want to open up X/Twitter

[fossil fable]

christ

is the forced mini reasoning

is it over??

[sullen quest]

doubt

[magic stag]

wtf?

i cant believe thats real

lmao

"In December, as we roll out age-gating more fully and as part of our “treat adult users like adults” principle, we will allow even more, like erotica for verified adults."

[safe geyser]

I only care about gpt image

Maybe it's due to elon musk new video ai

[sullen quest]

openAI needs the money lol

[brave cloak]

gpt image for what? it always changes any source image, it's garbo

[magic stag]

i think nerfing software because people are stupid enough to kill themselves over talking to a piece of software, not even a real human, is stupid

but thats just me

[cloud zinc]

sora 2 worse than veo 3

[safe geyser]

Nah i like to use for realistic conversions

[cloud zinc]

https://x.com/arena/status/1978149396996051007

lmarena.ai (@arena)

🚨 🎬 Video Arena Disrupted!

@Openai's Sora 2 and Sora 2 Pro have landed on the Text-to-Video leaderboard.

🏆 Sora 2 Pro is the first to tie rank with Veo 3 variants for #1.
🥉 Sora 2 comes in at #3, pushing the non-audio variants of Veo 3 into 5th!

Video models with audio

X · Oct 14, 2025, 5:22 PM

[sullen quest]

I don't think people needed conversationality that much

peoples lives matter a heck a lot more than some other peoples ability to have a conversation with a chatbot

[magic stag]

[magic ravine]

Dang, finally gain access to Sora 2 and what do you know, they nerfed it like crazy!!! 😂 😭 I wanted Batman fighting Godzilla and it refuses due to copyright. I just wanted to see that rendition similar to those anime Sora 2 videos I saw online lol Guess I'm just unlucky

[leaden sun]

i really like his transparent communication style, this seems to be following the grok business route with that NSFW toggle, now after passing the medical exam, i assume chatgpt is officially allowed to "diagnose" mental health without legal troubles?

[magic stag]

LOL.

[magic stag]

wym medical exam?

[agile bloom]

is there an app for lmarena? to use all text ai models on app

going on the browser, opening it, passing cloudflare verification, selecting model all seem like a hassle

[magic stag]

no one is going to let a chat bot diagnose anything yet... i think they mean controls to detect mentally ill people who will kill themselves over a chatbot and change to safe mode

[fiery gull]

vote 😈 🙏

[neon idol]

Hi chat

[thorn jewel]

hi

[neon idol]

Hello!!

Hru?

[thorn jewel]

fine

[neon idol]

Glad to hear that ^^

[thorn jewel]

ohh

hru?

[echo aurora]

hello

[neon idol]

Hi pineapple!!

Hru?

[neon idol]

In really good thx!!

[thorn jewel]

okh

where r you from bro?

[neon idol]

Pizza county

Wbu?

[flint sandal]

new gemini checkpoint out

kinda good

ngl better than gpt-5 high, and claude 4.5 in coding]

and fast

people say its 3.0 pro, and if thats right idk how good 3.0 ultra will be

i will send a landing page it made.

zero-shot

wait

[delicate frost]

have just delete kilo code and disable autocompletion, it made my code so much more clean and my coding so much more effiecent

[polar niche]

Hi

[polar niche]

Interested

[flint sandal]

anyone got access to gpt-5 pro to compare it to new gemini 3 checkpoint?

it will be more fair i think

[balmy mist]

me

i got 2 pros subs sadly, got crazy with sora stuff lol

[polar niche]

I need access to pro

[flint sandal]

can you use this prompt "create a visually appealing ladning page for cpu manufacturers in one html doc."?

[balmy mist]

okay

[polar niche]

Using aistudio?

[flint sandal]

we will send results at same time

[balmy mist]

gpt 5 pro on chatgpt

[polar niche]

Got the plus?

[flint sandal]

i will send html code from the ECPT checkpoint

[polar niche]

I don't see it in the app

[balmy mist]

you need pro, $200 a month

[polar niche]

Oh god

No

You paid that much?

[balmy mist]

sadly, i was getting crazy with sora gens

i regret it so much

lol

what makes it worse i bought it twice

[polar niche]

Oh no lool

[balmy mist]

thats my spedning limit on ai for the year tho lol

[polar niche]

Is it good with reasoning or logical puzzles?

I have a prompt for you to test

[balmy mist]

yeah, but I only use to start projects or if codex cli cant figure it out, i mainly use pro for unliited codex cli and sora 2 pro

give me

[polar niche]

You can use Agents right?

[balmy mist]

i have so many i need to waste

[polar niche]

Use Agents mode gpt 5 pro

[balmy mist]

this right?

[polar niche]

Yes

[flint sandal]

did gpt-5 pro finish?

[polar niche]

Cipher:

Information: Difficulty - easy

Ciphertext: 𓊪𓈋𒀞𒇾𓆉𓆡𓋵𓌏𒀓𓋼𓄛𓇫𓋊𓉡𓈸𓋵𓌏𒀓𓋼𓄛𓇫𓋊𓉡𓈸𓇢𓄟𓈎𒂗𓇫𓋊𓉡𓈸𓊪𓈋𒀞𒇾𓌕𓋼𓄛𓊪𓈋𒀞𒇾

[balmy mist]

almost done btw

[flint sandal]

oh okay

[balmy mist]

send this whole thing?

[polar niche]

Yes

[balmy mist]

im scared this better not hack me or nun lmaoo

[polar niche]

Can you send team invitations if possible? Or share a task

No lol it won't

It's just a fun puzzle

[balmy mist]

okay lol, so just paste into agent, ill try and add teams to and invite let me see, i only use codex lol

[polar niche]

I like to test the intelligence and reasoning like this

[balmy mist]

I only have teams on my work email, ill just do the puzzle for now on agent

still loading bananoc

i hope its good lol

[polar niche]

Can you show how it looks?

[flint sandal]

if gpt-5 pro wins, gemini 3 is gonna be bad

[balmy mist]

join the video ill share

playground

[kindred tundra]

gemini-2.5-pro
Generating.. wha to do agg

i text only 10-20 min

[flint sandal]

wait

[balmy mist]

tell me wen

want me to put it in the html runner?

[flint sandal]

nah

i have to find the html file

[polar niche]

oof I am lagging

[flint sandal]

OH HELL NAH

LOOK AT THE GPT-5 PRO RESULT

[kindred tundra]

gemini-2.5-pro
Generating.. wha to do agg
i text only 10-20 min

[balmy mist]

how does the g3 one look?

[flint sandal]

i sended the html

you can run it

[neon idol]

Gemini 3 come out????

[flint sandal]

nah

early checkpoint

[neon idol]

🥀

[neon idol]

What is?

[polar niche]

Don't advertise here

[flint sandal]

dm me

[neon idol]

Okk

[polar niche]

Also it shouldn't take this long to decode it's a simple sentence

[balmy mist]

its okay

hmm i cant tell which i like better

you see my screen?

[flint sandal]

yeah

i think g3 is slightly better

[balmy mist]

the visuals are cleaner

[flint sandal]

on gpt-5 pro the gradients are buggy

and overall in gpt-5 pro there is too many things at once

but g3 failed the button positiob

position

[balmy mist]

the ui in the g3 is alil buggy

like when clicking

[echo dome]

I remember comet

[flint sandal]

its early checkpoint but still i think if i will iterate this with like one more prompt it would be much better

[echo dome]

comet looks dangerous

[balmy mist]

true and gpt-5 pro reasoned way longer

[balmy mist]

@polar niche your agent still going lol

[flint sandal]

yeah g3 responded in like 30 seconds max and reasoned for like 25 sec so 55 total

fr

[polar niche]

I don't know whats it doing for this long lmao

[balmy mist]

agent is pretty much browser use right?

[polar niche]

And terminal

I tested an agent app and it solved a whole cipher

[balmy mist]

u an llm whisperer or sum lol

[polar niche]

I'll let you know what it means after it finishes

It's so easy lol

[balmy mist]

okay lol,i got that comet browser earlier cause people were usign it to get g3

has anyone been successful in that?

[polar niche]

comet has g3?

Its prplx ran I dont think it has it

[flint sandal]

the g3 in comet is baaad.

in ai studio its like 2x better

was bad:)

[balmy mist]

its complicated but you can spam the studio site to get a/b testing with g3 which just makes no sense to me

[flint sandal]

for testing it makes sense

[balmy mist]

but there are other ways they can do that

[flint sandal]

like i got it in 6th try

[flint sandal]

like what

[balmy mist]

studio seems risky

[polar niche]

Yeah but how does comet have it

[balmy mist]

like people are litterally bypassing so whats the point of even having that just release a test version at this point

[balmy mist]

you use comet to go on ai studio

and spam it

[polar niche]

It only works on comet?

[flint sandal]

you dont have to use comet

[polar niche]

Yea

Works on any browser

[balmy mist]

no you can use regular but i just dont have time to sit there an do that lol

[polar niche]

lmaoo

[flint sandal]

chatgpt agent can do that

[polar niche]

@echo aurora

[balmy mist]

like on comet they do 200 agents go and hit ai studio repeatedly

[balmy mist]

duhh im slow

wtff

[flint sandal]

and how to do it on comet?

[balmy mist]

https://x.com/testingcatalog/status/1978146140970189039

TestingCatalog News 🗞 (@testingcatalog)

This is the way 👀

X · Oct 14, 2025, 5:09 PM

[polar niche]

I dont think its worth it

Just wait for a beta release

[balmy mist]

you think so?

even with agent mode?

[flint sandal]

nah. it hits rate limits on comet and the assistant is waiting for gemini to finish

it should instantly stop and rerun if its not a/b test

agent is doing much better job at this]

i think

[polar niche]

It's not much of an upgrade currently

[flint sandal]

ai is trash for now. if ai will make sites like this for example and will solve complex coding problems, then it will be good. now it feels more like a bug tool

smth like this

in zero-shot

[polar niche]

Can chatgpt projects files read zip file contents?

[flint sandal]

tip: make animations faster and smoother like add ease

[polar niche]

@balmy mist still going 💀

[flint sandal]

and if you doing 3d like on the start of this actually use 3d because it looks kinda bad

[balmy mist]

nahh agent is about to cook omgg

its researching online how to trigger g3 lol

[polar niche]

it cant bypass google auth

sadly

[balmy mist]

i just gave it my auth

its bypassin gnow

lets see lol

this is wild

if it does this im gonna go agent crazy this week lol

bruh agent is using nano banana lmaoo

@polar niche look

[flint sandal]

bro using chatgpt to use gemini

[balmy mist]

smhh

agent having trouble with the scroll lol

[polar niche]

is it doing it?

[balmy mist]

honesly i do not know anymore

[echo aurora]

Hey apologies, was in a meeting, I assume is all good now?

[polar niche]

Yes

[flint sandal]

pineapple do you know..

gemini 3 when'

[polar niche]

Its creating a script in ai studio?

[echo aurora]

I do not

[polar niche]

pineapple gpt 5 pro when?

[balmy mist]

more like using studio to acces g3

[flint sandal]

🙁

[echo aurora]

same answer 😭

[balmy mist]

bruhh agent mode said it accidentally selected the pencol icon instead of rerun, thats wild

its really acting like a user

[flint sandal]

at this point lets just ask vicuna-13b...

[polar niche]

thanks well it couldnt do it

[robust yoke]

Greetings, Greg.

[balmy mist]

yeah it couldnt sadly

[fiery gull]

Did something interesting with AI today? 👀

[rapid gazelle]

hello

[fiery gull]

Hi 👋

[meager harbor]

What score will gemini 3 pro get on lm arena you all think ?

Around 1500 ?

[fiery gull]

1513

[robust yoke]

https://cdn.discordapp.com/attachments/1311972651006296198/1427764939124113408/meme.jpeg?ex=68f00d38&is=68eebbb8&hm=f695746778e054c9d6b374886856c7512e2f37a7c3ba22d16c0cf5dfa040c736&

[meager harbor]

Why so precise ?

You know something we don’t

[fiery gull]

Because the only thing that gemini does badly is coding, if it does 1500 in coding it increases 100 points

[meager harbor]

?

[fiery gull]

1503*

[fiery gull]

Looking forward to gpt for alduts? 🤣

[neon idol]

No, nothing. But whit my hand yes. I have created my current pfp

[neon idol]

Hello my man hru

[fiery gull]

Which AI?

[meager harbor]

It’s the best for my use case for sql generation and code reading with that long ass context windows in ai studio, the competitirs ai are nerfed in free version and even in the paying version the context is low

[neon idol]

No one

handmade

[fiery gull]

Hmmmm

[fiery gull]

Sound goods, I just listen bad things about code from gemini 2.5 pro

[meager harbor]

Maybe claude is better but with that low context windows i’m not bothering

[fiery gull]

yes, too good to never have to worry about context limits

[fiery gull]

How long did it take?

[neon idol]

1 day

[meager harbor]

In lm arena is there a limited context window ? Like the limit is the same for each model tested ?

When you test model

[fiery gull]

No, the lmarena use a dry limit of model, if you are using the grok 4 fast, you can use the 2 mi contexts

[meager harbor]

Good to know

Even when comparing 2 model ?

[fiery gull]

But if you're using the sonnet 4.5 you just can use when you hit the limit

200k

[fiery gull]

All modes, direct, battle, side a side

The Pineapple told me that

[robust yoke]

You betcha.

Gonna make MaidGPT.

[meager harbor]

The pineapple ?

[meager harbor]

Waifugpt

[fiery gull]

Is an mod here

[fiery gull]

Maidgpt 🤣

[robust yoke]

Sto andando molto bene, e tu?

[polar niche]

Pineapple goes on pizza?

[robust yoke]

FemboyGPT. 🗿

[polar niche]

we will have ai robots for girlfriends

[neon idol]

Me too

Thxx

FluffyGPT 🤤🤤

[fiery gull]

Italian too? , the man is a machine

[robust yoke]

E ‘ un piacere.

[neon idol]

Real

[robust yoke]

This is a proven fact.

[meager harbor]

We need asi so they can replicate girls to the perfection instead of having a cold robot

[robust yoke]

As previously stated, I am a version of Grok.

[fiery gull]

Lol 🤣, the mind flies

[polar niche]

Clanker wars about to start

[robust yoke]

Relationships in the future be like:
Husband: “Hey, babe, just got home from work. Could you create a webpage for me that includes our most recent photos with each other in Italy?”
Wife: “Sure, honey, I'd love to. Making that now…”

[fiery gull]

Humans won't stand a chance, machines will win 50% of humans just by mimicking waifus

[loud viper]

Hello new here, how do I get started?

[fiery gull]

#1397655624103493813

[echo aurora]

hello - what are you looking for?

[storm roost]

Hello I'm new here. how do I get started?

[loud viper]

text to video, pic to video

[echo aurora]

ah yeah you'll want to review the information in #1397655624103493813

[loud viper]

ok thank you

[polar niche]

How's everyone's day?

I guess so

They are updating the api

[lunar crystal]

Hello everybody

[polar niche]

Hi

[stray aspen]

Yo

[native flame]

Hii, everything I do with the video bot will be public?? Or is there an option to do it in private like on the web??

[echo aurora]

hello

there isn't a way to use Video Arena outside of this server

you won't be able to use in a different server or through Direct Messages

[native flame]

u.u oki thanks for answering 😂

[graceful sonnet]

Image generation, not video...

[pearl pebble]

https://youtu.be/7l8OhvlgnoY?feature=shared

[ember turret]

Hellow!

[astral elk]

Good nice 👍

[fiery gull]

Is dynamic, but it's about 1 per hour

[echo sinew]

@ember turret Please, read our guide ⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to properly prompt the bot.

[north horizon]

Hey guys, comparing text to video AIs with sound to ones without sound doesn't seem fair at all

A good no sound model getting matched with sound models is going to lose ELO unfairly, sound will just always be better for some things

[sullen quest]

not enough sound models for a separate leaderboard tho

[polar niche]

YO

[fiery gull]

Hi

[polar niche]

Wspp

[jovial girder]

hi I am here to create and test AI generated videos

[polar niche]

#1397655624103493813

[young nymph]

Hi Im MIKE

[dense sphinx]

Can you fix canary lmarena please???!

It doesn't even working when I try type chat!

[brittle tiger]

https://x.com/OfficialLoganK/status/1978275361336045871?t=xj7_sSk4e6EvDOuCxQGJBA&s=19

Logan Kilpatrick (@OfficialLoganK)

Veo

X · Oct 15, 2025, 1:42 AM

[echo sinew]

@glossy shard Please, read our guide ⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to properly prompt the bot.

[honest sage]

who needs. code dm me

[jade egret]

veo 3.1?

real or nah
https://x.com/chetaslua/status/1977866953705316571

Chetaslua (@chetaslua)

🚨 Gemini 3.0 Pro - ecpt checkpoint

Holy shit Guys , i want everyone to see this retweet as much as you can to get this to mainstream , i dont ask for this normally

All apps work , apple animation , minimize , tools , browser , and everything literally this is the best we can

X · Oct 13, 2025, 10:40 PM

[cloud zinc]

real

[jade egret]

it actually from gemini 3 pro on ai studio (1/b testing) one shot, and only gemini 3 pro?

if that true than very impressive

[blissful chasm]

Why are there two Deepseek labels

Deepseek and Deepseek AI

[whole sundial]

seems like lmarena just did an update...

[gray sentinel]

qwen image edit is not working

am i the only one or is it happening to everybody?

[echo aurora]

yeah this is new!

[echo aurora]

we did report some issues with this model earlier, it's being looked into

[echo aurora]

Hmm good question. I'll look into and flag. Thanks!!

[gray sentinel]

oh okay, i mean it was showing something went wrong so i thought it might be connectivity issue, thanks for the response tho

[echo aurora]

Unfortunately that's a generic error message that you're seeing which can be raised for various reasons. We plan to make these error message more clear for easier troubleshooting

[rose light]

Hello

[limber ledge]

Hello

[vital obsidian]

how to generate video 8 - 10 s each video ?

[thorn jewel]

hii

[vestal relic]

hello

[warm sleet]

Sora2 code

[fiery gull]

There's no way, just 8 seconds

[restive plank]

Does video section support first and last frame option?

[fiery gull]

I think no

[restive plank]

Is there any other which is free?
I only need one shot, free trials can works too.

[fiery gull]

Grok imagine

Have audio and etc

And it's really good and 100% free

Grok**

[restive plank]

Does gemini support first and last frame in video?

[fiery gull]

I want to say grok sorry

[fiery gull]

Idk about the last frame, go see it in grok imagine

[short jewel]

Me pueden facilitar algún código para sora?

/sora 2 get a invite code

[meager harbor]

Why can’t we load txt file in lm arena ?

It’s useless i need to send some file into it so it can read it for me

Its useless

[fiery gull]

In the past I complained every day about this but I gave up

[meager harbor]

It totally flaws the ranking if you prevent to use models to their full capacity

Stupid sheit

[pale vapor]

wanted to share new chinese provider with 200$ of free tokens. send me a msg and i will invite you

[meager harbor]

We need to harass them so they finally understand lol

[fiery gull]

@echo aurora

[vernal oracle]

Hey guys

[drifting flower]

Hello everybody

[robust yoke]

Could recommend that in #1372230675914031105.

[robust yoke]

Greetings.

[robust yoke]

Hello there.

[outer estuary]

Dang

Gemini was.....

[azure apex]

Hi LMArena team! 👋
My name is Gigi and I’m working on a project that automatically generates images for blog posts. I’ve tried GPT “edit an image” and Gemini APIs, but they don’t give the quality or flexibility we need. I wanted to ask if you offer any API or partnership program that allows sending prompts and receiving generated images automatically (like JSON response with image URLs or base64).

Thanks a lot for your help!

[fiery gull]

@echo aurora

[fiery gull]

Send to #1372230675914031105

[gusty coral]

How to generate 9:16 ratio video

[brave ocean]

hello

[fiery gull]

Hi, want to some help? I'm here to help

[primal moat]

hello

[fiery gull]

👋 hi

[upper plinth]

hi there?

[polar niche]

Yo

[mortal frost]

Girl

[polar niche]

"Girl" 🤖

The qwen models are trash honestly

[fiery gull]

whatt ;-;

ragebait?

[polar niche]

No

Thinking can't solve easy puzzles

The new one added

I want Gemini deepthink

[velvet ridge]

I want to learn AI

[polar niche]

@echo aurora Can I close out of the website on phone and it won't stop the generation?

Sometimes it stops my tasks from generating

[mystic belfry]

bro wtf

the video generating doesnt work gng

[polar niche]

Bot's broken

@echo aurora Bot's broken

[prime halo]

bot is down?

[polar niche]

Image to video

[exotic finch]

Hi

[normal hill]

is anyone also experiencing the 'failed video generation' issue under #video-arena-2 ?

[teal mantle]

autoregressive is beautiful and change my mind

[dense sphinx]

IT'S NOT WORKING

Lmarena NOT WORKING

Fix please!!

[lilac nimbus]

Real?

[dense sphinx]

For God sake, someone fix it!

[finite meadow]

Vedio is not getting generated what the problem

[elder burrow]

xD

aistudio a/b test response

[robust yoke]

Possibly.

[sullen quest]

Sounds like you are geting silkposted

[magic stag]

......

[magic stag]

[dense sphinx]

Sorry

But it's frustrating.

I am sorry for bothering you.

[robust yoke]

Hmm…

What model were you wanting to use that on?

[median blaze]

bot is down?

anybody knows why?

[robust yoke]

I don't think it's “down”, per se, but who knows?

[bright condor]

Why do I keep getting a failed to get evaluation

[robust yoke]

Could try asking one of the mods the next time you see one of them on.

[median blaze]

ahh so it a normal thing to happen?

[robust yoke]

Maybe.

Try asking Pineapple.

[median blaze]

thank you

[robust yoke]

He's a mod here.

[vague otter]

@rich widget's Prompt:

Close-up macro shot of a Pakistani man working as a Thai street food vendor, professionally cleaning and preparing fish before cooking. He is making precise diagonal cuts into the fish with a sharp knife, showcasing expert technique. The bustling Thai street stall surrounds him with mouthwatering energy. Aromatic basil leaves, red chili, garlic, and minced meat are visible nearby. Steam rises vibrantly from hot pans. Warm cinematic lighting bathes the scene, capturing the intensity and focus of the vendor. --motion high --resolution high in 9:16 frame size

❌ Generation failed. Failed to create evaluation session.

❌ Generation failed. Failed to create evaluation session.

[grim ingot]

Generation failed. Why? 🥺

[magic stag]

More so referring to your output

Lmao

[south jolt]

Hello

[hallow jackal]

bot is down?

[rich mason]

啥情况

[polar anchor]

veo 3.1 is out

[dense sphinx]

@robust yoke Claude sonnet 4.5

And not only that.

All models as well... Or... Lm arena not working

Properly.

[dense sphinx]

What program is this?

[polar anchor]

flow

by google

[robust yoke]

Ah, so just use it on their website. Here: https://claude.com

Interesting.

[thorn falcon]

hi

[polar anchor]

GTA 6- veo 3.1

[indigo hazel]

Guys is the site working? Battle mode and direct chat doesnt work

[dense sphinx]

@robust yoke No no you don't understand.

When I use and add lot words it became error.

It's actually not happened that before.

I use LMarena and when I add many words it came error

Usually it's not that happened.

[balmy mist]

Woah this might be better than Sora 2 lol

[robust yoke]

Hm.

[balmy mist]

How many gens u get with veo3.1?

[barren prairie]

Ok I will stop waiting for Gemini 3 now 😇

[maiden crest]

Hi Pro's and Veterans, What is the best and accurate chat ai for you in creating ai image prompts?

[barren prairie]

The sooner of Google tooks ages

[robust yoke]

Well, you could just use their official website if you wanna do something with that, as well as posting on that in #1343291835845578853.

[robust yoke]

Either GPT-5-High or Grok 4. Maybe try Grok 4 Fast?

[dense sphinx]

@robust yoke Thanks.

[robust yoke]

My pleasure, but for future reference, you don't have to always ping me like that to get my attention.

[balmy mist]

Yeah I’m done

[robust yoke]

What do you use Discord on?

[maiden crest]

thank you for the quick reply. I guess I'm on the right track. Grock tend to slow down sometimes unlike Gpt 5 for me.

[barren prairie]

Qwen added memory 😶

[robust yoke]

Well, if that ends up being the case, then GPT-5-High might be your go-to. Also, you're very welcome.

I usually like to respond to messages as soon as I can.

[robust yoke]

Oh, W.

[rich mason]

how to use veo 3.1 ?

[polar anchor]

yeah with a few testing with veo 3.1. sora 2 its so much better lol

[elder burrow]

lol

[safe geyser]

Where did you test veo 3.1? At vertex ai?

[polar anchor]

Flow

[elder burrow]

yall is gpt 5 pro not on lmarena

[safe geyser]

already out 3.1?

[polar anchor]

yeah in flow for pro users

i will share a couple in #ai-creations

[rich mason]

When will it be fixed? Does anyone know?

[proven reef]

also veo3 on gemini doesn't work

[polar niche]

Not working

[maiden crest]

Hi, is Hunyuan image unlimited?

[ember iris]

Is lmarena.ai working?

[maiden crest]

working to mine

[echo aurora]

Hmm sounds like the bot is down

Taking a look

[echo aurora]

Site down for you too?

[polar niche]

It let me vote when it was still generating

[polar niche]

@echo aurora Whats this model?

[maiden crest]

everything works on what I used and voted on the site but not with Qwen

[polar niche]

https://discord.com/channels/1340554757349179412/1427623336669872159

[echo aurora]

I'm not getting a response when using Battle

[maiden crest]

thanks for the quickie

[polar niche]

#video-arena-1 image to video mode is broken

[echo aurora]

Yup seeing that

[polar niche]

Also this

[polar niche]

And this

[ember iris]

Yeah. Can't able to generate anything.

[maiden crest]

Im getting results on some generations, the problem when qwen is selected it hangs up

[echo aurora]

What about when using Battle? Are you running into an error?

[elder burrow]

i am but its really really slow. just now it took qwen3 like 10 mins for 1 simple html file, it streamed like once every 30 seconds and it failed in the end

[ember iris]

@echo aurora getting these errors in battle, sbs, direct

[polar niche]

Refresh

Happened to me too

Refreshed and it started generating

[maiden crest]

working on mine just now

[elder burrow]

i wish lmarena was stable :(

[ember iris]

Came here after trying refresh, cleared cache, deleted browser history, changed wifi to mobile network, used vpn. Used different vpn locations still same

[ember iris]

Trying since 20 min

[echo aurora]

Did it take awhile to load?

[ember iris]

Last option gonna try different isp provider. Let me change it and try

[elder burrow]

the lmarena frontend lags on my 4060 laptop with 24gb ram and ryzen 7 7435hs when models stream codeblocks, frontend desperately needs a rework 😭 @echo aurora

[maiden crest]

30s-50s i think

[ember iris]

@echo aurora are you getting the same error?

I just wanna create a logo 😁

[harsh garnet]

Why error gng

[ember iris]

Bro got it. I was closing and trying it. But as soon as the error came and reloaded now the results are already generated... 😁

[harsh garnet]

Why the error is coming

[ember iris]

Relaod

[ivory schooner]

It seems that errors only occur when inputting content in other languages, while the output works normally when it’s in English. If that’s the case, we’ll temporarily use English to ask questions for now....

[echo aurora]

Working on it 😭

[harsh garnet]

I did

[maiden crest]

btw masters which models have unlimited image generations aside from qwen?

[harsh garnet]

But still bro

[polar niche]

fix the models

🙏

[harsh garnet]

Which model

[polar niche]

and battle mode is broken too

[harsh garnet]

Bro @polar niche can u please tell my why the error is coming?

[polar niche]

Yes

[ember iris]

Thanks buddy. 👍🏻

[polar niche]

Which error

[harsh garnet]

Then tell

[harsh garnet]

In #video-arena-2

[maiden crest]

any one knows which is unlimited?

[polar niche]

Simply because the bot needs fixing

[harsh garnet]

Generation failed

[cinder zenith]

bot is bugged

[harsh garnet]

So I can't generate pic now?

[cinder zenith]

all generators not work now..

[harsh garnet]

Video*

[polar niche]

nano banana?

[polar niche]

Yes

Have to wait for it to get fixed

[echo aurora]

For those that are getting error messages from Battle mode, can you send me your evaluation ID? It's the code thats at the end of the URL

[polar niche]

You can't generate video from image

Image generation is fine

[maiden crest]

thank you, I'll try the results. I'd like to make images not too lite. ❤️

[polar niche]

0199e7b7-7e48-7d82-bf71-70b0ceee0c1d

[gilded vessel]

Nothing is working

[fiery gull]

Lmarena is buggy now

[polar niche]

In battle mode only 1 assistant is visible

[polar niche]

Not here brother

[desert star]

Make a video of above image

[polar niche]

Go to #1397655624103493813

[sweet tinsel]

No bot 🙁

[sharp hazel]

Yh noticed that too felt im the only one noticed it's general

What's going on?

[polar niche]

Go to #1397655624103493813

[void patio]

What happened?

[desert star]

Image to video generator

[violet moss]

Animar el logo por 8 segundos

[lusty falcon]

@desert star server problem

[polar niche]

Bot is broken

[fiery gull]

see how to make this in #1397655624103493813

[gray kelp]

???

[echo aurora]

Hey everyone - the site & video bot are currently having issues, we're looking into. Apologies for the inconvenience!

[polar niche]

Not here

[gray kelp]

Please

thx

[fiery gull]

bot is broken? hell no

[sharp hazel]

What happened why can we generate images and videos ?

[gray kelp]

They are working on it

[echo aurora]

Does anyone else have an Eval IID of a sessions that has errored out in Battle?

[fiery gull]

#video-arena-1

[sharp hazel]

Alright.

[gray kelp]

Oright

how long take that?

[fiery gull]

Idk

maybe weeks ;-;, I hope hours

[polar niche]

They'll anounce it

Don't worry

[gray kelp]

hiii we living in AI time in 10min;)

[maiden crest]

is this correct? different models generating the exact same image?

[modern garnet]

Hi

[elder burrow]

I just got phantom

[polar niche]

Seems to be broken too

[fiery gull]

LM Arena bug, so many bugs lol

[elder burrow]

@verbal nimbus

[polar niche]

Lm arena is full of bugs now

[gray kelp]

????

why

[polar niche]

It happens

[balmy mist]

you dont have sora 2 ?

[fiery gull]

vibe code don't work 🤣

[gray kelp]

Hiiii we are artist we are VIPs.

😉

[polar niche]

We need bug spray

[fiery gull]

wdym ?

[gray kelp]

idk i asked

for more information sry

maxmilex

hiii... where is the dev?

[elder burrow]

wait what? there is a TINY difference, look at the window on the very right, you can see a few spots being different

[fiery gull]

the dev don't use discord, you can talk to @echo aurora to it talk with devs

[stuck totem]

Video generate failed

[elder burrow]

TINY difference seen on the window

[echo aurora]

We've made a change, are people still getting errors when using the site?

[low remnant]

Generate an image

How does this work?

[fiery gull]

use grok imagine 🙃

[elder burrow]

#1397655624103493813

[fiery gull]

the lmarena is broken now 😭

[low remnant]

#1397655624103493813

[elder burrow]

https://lmarena.ai

bruh

[toxic sky]

#1397655624103493813

[elder burrow]

the "prompt engineers" have woken up

🤣

[toxic sky]

#video-arena-1

[elder burrow]

proompters

[fiery gull]

what was it?

[toxic sky]

#rules

[fiery gull]

goonpters ?

[elder burrow]

exactly

lol