general

[knotty fable]

That's another reason to leave, but it has been useful.

Too bad, but I am distancing myself as much as possible from them.

[keen beacon]

You won’t be able to avoid it soon in the future

I bet you in the future you’re gonna need to identify yourself before you get online

[knotty fable]

Well there's one European alternative which I have signed up on now.

[keen beacon]

Imagine v10 of captcha

“please scan you iris”

[knotty fable]

Google have tried that on the phone already, which is the reason I bought a new google free phone.

[keen beacon]

I mean logically it makes sense.

I mean ai is going to advance and agents will become hard to tell from humans

From a security stand point it does make sense

[knotty fable]

No free tab offered at a reasonable price, so I had to make quite an effort to get the crap out of mine.

It runs on Linux now.

[keen beacon]

I have iPad

I need to get off the ecosystem

But moving is so much work

[knotty fable]

Yeah, that's what I'm doing, you would be hard pressed to find a single photo of me online.
There is, if you know my full name - from the University pages, but those I accept to be seen.

[keen beacon]

This one’s a little better

[keen beacon]

Yeah I need to step up my opsec

[knotty fable]

Time for me to work, have a nice day now.
👋

[steel plinth]

Aloooh nice to meet you all:3

[thorn swan]

Hi

[keen beacon]

Same pleasure talking to u

[dawn ivy]

Hi I am new here I want to create longer videos for about 3-5 minutes long but video arena only create videos 11 sec maximum tell me how to create longer videos here thanks

[worldly garden]

Anyone got sora 2 invite code? If yes dm me pls

[oblique glen]

Hey all! Anyone have a Sora 2 invite code they're willing to share?

[worldly garden]

Codes?

[vivid sandal]

Is it just me or I'm getting spammed captchas whenever I send a message...

[wicked sage]

hi

[patent ridge]

hi

[cloud marten]

does Image to video arena is only In Discord or it is available on LMarina.ai website

[raw turtle]

I have made two videos now and it is being said here that your daily limit has been reached out. Please tell me that I am getting five videos a day.

[boreal saddle]

It already started with age verification.

Soon your phone will be bricked if you refuse to provide any personal info, lol

A unholy combination of brave new world and 1984.

I mean, remove it from the lmarena website.

[keen beacon]

You can’t cause I think you needed to use the cloud

Otherwise, Google won’t rank you in a whole bunch of other things from what I understand I could be wrong though

Well, people start to get clever and started to build bots

I’m guessing once again this is speculation

[boreal saddle]

Bots are a preferable alternative to... this. Especially with how much it disrupts your chatting.

[keen beacon]

Agents bots would hog up all the API on their own and people will just abuse the hell out of it

[boreal saddle]

Imagine asking an AI the question only to be met with a "Please verify your identity!" pop up every time. Even if you already did it before

The whole tool becomes completely unusable.

[keen beacon]

Well, dude, nothing is free

If you want the convenience of and the luxury of not having that the alternative do pay subscription to AI service

I totally feel your pain, man. I do I run the same thing.

[boreal saddle]

I mean, LMArena doesnt depend on constant ReCaptcha popups to work, so this is a nonsensical argument.

Like I said, its enshittification.

[keen beacon]

I’m guessing that it’s the only viable cheap alternative as of now

[boreal saddle]

Reminds me of how elon musk bricked Twitter by imposing extremely draconian data and post caps and the fallout was so terrible that he was forced to roll it back.

[keen beacon]

Have you ever heard the term techno feudalism?

Or surveillance capitalism?

https://en.wikipedia.org/wiki/Surveillance_capitalism

[boreal saddle]

Of course

[keen beacon]

https://youtu.be/xFy41HbYbg4?si=F97sGsGQhi7NOZxj

[hazy kernel]

wym completely?

[keen beacon]

Well, in all honesty if people who genuinely oppose the systems and find them oppressive and detrimental I don’t see how one could even use AI

Since it is the evolutionary next step to much broader and more restrictive systems of controlling information and access as well as powerful tool for that extraction and optimization

[errant pagoda]

#video-arena-1 is unlimited?

[keen beacon]

https://youtu.be/-gGLvg0n-uY?si=1tUWw0IefB6_8Vly

The more I live the more experience, the Internet, the more I see in the more witness and the more I interact with them more. I am starting to think this may be a possibility. lol 😂 (don’t watch it if you want to stay in bliss )

[sinful pebble]

hello

[halcyon raptor]

Hello

[stone kiln]

hi

[dim basin]

Anyone here work with n8n?

[visual hull]

I think I broke the AI

Anyone having this issue?

It has been like this for hours now

And no matter what I do it just keeps getting stuck like that

whether I reload the page

reopen it

it's just like that

[keen beacon]

Rate limit

[visual hull]

Will this be gone tomorrow?

Or do I need to open another context window

[keen beacon]

I’m not sure I never tested it against that but I think what people are saying it’s you get a few messages every 24 hours. I think it’s four or something like that.

Maybe 12 hours or 6 not sure

[visual hull]

Aight

I'll patiently wait I guess

You know any alternatives offering claude?

[keen beacon]

No. A subscription lol

If you find value in it and use it often you should consider a subscription 20$ is a bit spendy but it’s not to bad of an expense if u find ur self using it often

[ionic shard]

I’ve been trying to test Claude 4.5 Sonnet, but I keep running into this error. Is there a fix for it?

[keen beacon]

Rate limit

[ionic shard]

How long do i have to wait

?

[fiery gull]

Just change for opus 4.1

[obsidian shell]

Anyone knows what gpt-5-high-new-system-prompt is?

[ionic shard]

Still the same

Error

[obsidian shell]

Is it your first message even after a reload of the page?

[ionic shard]

Yup

[obsidian shell]

There is nothing you can do from your side unless you change a device, they will fix it in a few hours.

[ionic shard]

Oh alright

Thanks

[obsidian shell]

No problem.

[thin current]

Same

[ionic shard]

Also ive tried other models it worked perfectly fine

Expect for claude models

[obsidian shell]

Its anthropic's way of saving money.

[hazy kernel]

what is this

why does it contain an exe file

[viscid venture]

I want to generate images and videos

[fiery gull]

generate images in lmarena.com and video is #1397655624103493813

[golden ocean]

sora

[bitter bloom]

Hello everyone. I just joined thus server. I am completely new to video generation world but dream of working on in this field

[quartz pike]

YALL is minimax finally stable?

minimax m2*

[south kiln]

hello

[keen beacon]

All right gang I’ll see you guys later. I got work. Be back in a day or two.

[fossil fable]

m2 is garbage - don’t use benchmaxxed models

[quartz pike]

ok

[fossil fable]

grok COULD ironically be the worst mainstream example of benchmaxxing so don’t trust that either

[ember iris]

Any model to create realistic images like seedream4-highres. Not fal.

If any sites we can use it for free to try it drop msg. Missing that model.

[daring rock]

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

[distant horizon]

hi

[pulsar saffron]

IS THIS SAfe?

[sullen quest]

always put it into virus total first

[pulsar saffron]

okay

[odd stratus]

yoo hi

[daring rock]

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

[balmy mist]

looking for beta testers for this prompt gen app for video models I have been using to make ai vids, let me know if anyone is interested, just need to see all the ways it can break, gonna make vids on how to use it fully, only need a few more testers(3), dm if anyone interested, a gemini api key is required as well

[hexed rune]

Guys what's the latest ai tools
Trending one please? 🙏

[eternal spruce]

🤨

gemini 2 5 flash image nano banana error

[modern jewel]

Hey, which AI is currently recommended to make Powerpoint presentation?

[wicked sage]

its a usage problem

claude ai has a setting called usage(s) where you can check how much you used on claude

(i think this shows to pro plan members? idk)

[granite heron]

highly recommend not to download files that were posted out of nowhere, nobody requested this user to share anything, he just posted it

[leaden cove]

BRUH

lmarena where the claudes models at?

are they're gone?

@echo aurora

[stray aspen]

no

they arent gone

[leaden cove]

where is they

i cant select any of them

i dont see them in directly chat

[halcyon nimbus]

oh yeah looks like direct chat to any llm is down

im only seeing the image option

[stray aspen]

its not lol

[halcyon nimbus]

refresh the homepage

[stray aspen]

still working

[halcyon nimbus]

oh weird, i have to x out the image selector down the bottom then text appears, never had to do that before

wasnt it just the option of text or image before?

i could be wrong lol, anyway, claude is there fine \o.0/

[echo aurora]

Hmm they're missing?

Yeah I'm seeing the models on my end as well.

[young magnet]

hi

[echo aurora]

@jade owl are you not seeing the models?

[gray sentinel]

is qwen image edit not working?

[echo aurora]

Are you trying to do an image edit with it or just text to image?

[gray sentinel]

i am trying image edit only

but its not working in the direct chat

but its is working sometimes in the battle mode

is there something wrong from my side?

now it isn't showing up in the battle mode also

[echo aurora]

How often were you using the model before it started to error out? Did it happen right away?

[gray sentinel]

yes, when i select it in the direct mode, and give any image with prompt, it straight away shows the error

i dont use the model too often, i use it like twice or thrice a week

[echo aurora]

I assume the same is happening in side by side? You aren't seeing the same with other models?

[gray sentinel]

yeah, it is showing the same error in the side by side mode too, no other models are showing that error, the rest works fine the image editing ones like seedgram and nano banana

[echo aurora]

Can I have an Eval ID?

[flint sandal]

i always have this problem with gpt-5-high-new-system-prompt in coding, that the code isnt in a codebox, its just plaintext and it like in %80 doesnt work well.

[surreal creek]

final leaderboard update of the month coming today?

M2 should have enough votes now :3

[echo aurora]

I normally don't share specifically when an update is coming.

It'll be updated when we've collected the votes and validated the data.

[obsidian cargo]

did you guys know the wait for gemini 3 is one day shorter than it was yesterday?

[echo aurora]

This is true.

[magic stag]

[surreal creek]

oh, interesting! just figured it’d be today or tomorrow since the MiniMax Twitter post last week said it was coming this week, been a lil bit since the last one 😄

excited to see where it lands!

[sullen quest]

Nah its like am iceberg

[knotty fable]

The rare Platitudinous pop up with its head.

[sullen quest]

Every time it takes 2 steps forwards it moves 5 steps back

[echo aurora]

It normally takes about a week, but can depend.

[sullen quest]

Pineappe knows

[echo aurora]

Full honesty I do not lol

[warm zodiac]

I don't even think most people internally at GDM know

[obsidian cargo]

Nah my uncle's wife's girlfriend works at Google and she said so

[golden ocean]

https://tenor.com/view/monkey-thinking-monkey-thinking-think-money-gif-9734025314392564509

[knotty fable]

We intend to keep you anyway, even if you're not omniscient. 😺

[echo aurora]

I appreciate that

[unkempt smelt]

hola

[knotty fable]

Hola.

[rich cipher]

guys

heko

help

i dont want build mode in webdev... how i switch it

[echo aurora]

Can you not unselect the modality?

Yeah yeah

[cold oracle]

@keen beacon @keen beacon

[echo aurora]

Then we'd still be one day closer compared to the day prior

[rich cipher]

where

[echo aurora]

Ah, that site is just webdev, you'll want to go to - https://lmarena.ai/

[rich cipher]

i wanna choose build mode without going in webdev

wh

@golden ocean

do you know

[echo aurora]

You can go to https://canary.lmarena.ai/ and use webdev with other modalities in the same area

[rich cipher]

bru why is there so many websites for lmarena

?

[echo aurora]

Other arenas were created at different times, but we're in the process of brining it together

[rich cipher]

oh ok

thanks for your help...

[echo aurora]

No problem, let me know if you have any other questions

[rich cipher]

no for now

[snow thunder]

any space to see our videos generated without scrolling?

and only 5 creation per day?

[echo aurora]

Yes, 5 generation requests per 24/hr. And you can find the generations you prompted by either:

• Open up Direct Messages with the bot
• Search your username in the search bar but use the mentions feature

[frank night]

i need one help can anyone help me ?

[fossil fable]

targeted benchmaxxing might be worse than general benchmaxxing

[echo aurora]

Whats up?

[frank night]

i want to type one json prompt but when i am pasting it becomes txt file

[fossil fable]

i feel like it's pretty self-explanatory

i gtg too

[frank night]

can anybody tell me ??

it becoms like this and i am unable to generate video

[rich cipher]

its discord problem lol

[rich cipher]

you should sent it as file ?

.json file

[frank night]

will it generate video ?

[rich cipher]

ye do not paste the text in the textbox

[frank night]

unable can you guide me

[autumn summit]

Hello

[lapis imp]

Question

Let ssay you hit a ratelimit limit on a long like 4 days chat worth lot of creative writing and you want to continue will the rate limit go or will I have to restart the burden aka the chat all over again?

Apologies if troubling

[echo aurora]

Hmm sorry to say I'm not following the question, would you mind elaborating a bit?

[echo aurora]

You shouldn't have to restart with a new chat, justing waiting for that rate limit to end should be all it takes to start using again

That being said I have seen some cases where very long chat conversations can become a little slow

[lapis imp]

So an hour or few hours?correct (sorry for troubling you)

Also suggestion to add a timer to the ratelimit to see.

I appreciate the help

[humble flame]

how to use direct video generation messaging?

[frank night]

look when i paste the json promt into chat it gets converted to txt file which is not accepted by ai is there any other way to paste large json prompt ?

[echo aurora]

It's no trouble. And it can be a bit complicated so apologies in advance. There are rate limits imposed by LMArena, where you will see an error message that says ~"You're being rate limited, and it'll expire in X minutes". The model itself can also impose a rate limit, and this is where you see the generic ~"Something went wrong" error message, for this case you won't see a count down timer. OVerall we are looking into ways to make this experience better for the user as it's not always clear what the issue is if you run into this error message.

[echo aurora]

You're unable to use the bot in Direct Messages, you can only use in the Video Arena channels ( #video-arena-1 #video-arena-2 #video-arena-3 )

[echo aurora]

Increasing the file types LMArena can accept is something we want to increase in the future, but unfortunately it's just images at the moment.

[gaunt spade]

lol what kinda name is that

[frank night]

shortform for jewellery

[gaunt spade]

lol fr

anyways did u guys test the new model

its called "mist"

idk if its from google

[stray aspen]

no

[gaunt spade]

#codename-discussion

[humble imp]

ummm. how many videos i can make ?

photo to video's

[humble flame]

how to handle text spelling errors in video?

and which video ai has that soud effects?

[knotty fable]

Not much to do, most AI's add warbled words. [Few exceptions: DDG actually do add correct words on lets say advertising that appear from off screen when the camera zoom out in a scene - and it the right language when stated. SeaArt also appear to handle text quite well.]

[echo aurora]

5 generation requests per 24/hr

[echo aurora]

You can see the list of models available on our leaderboards - https://lmarena.ai/leaderboard/text-to-video we don't list out which have sound or not, but iirc: veo, and sora models have sound

[humble flame]

thanks guys!

[knotty fable]

My test today of the AI's here will be to see how these AI's handle a scene with Butoh dance - I expect many fails. 😺

[echo aurora]

TIL about Butoh. Please be sure to share with us how they perform!

[tulip tree]

@echo aurora how do you not have stickers?

I just noticed

[pastel adder]

Has this been updated for all models? I havent used LmArena this week and havent gotten that message yet

[echo aurora]

For the server?

[tulip tree]

Yes

[pastel adder]

I'm still using Sonnet 4.5 32k thinking as my default model btw

[echo aurora]

Each model is going to have it's own rate limit, which will vary from model to model

[echo aurora]

Hmm yeah I don't think we've ever really used that feature.

What stickers would you like to see?

[tulip tree]

Are you old and don't know anything about discord

[tulip tree]

Anything

[pastel adder]

I see, and would that be released publicly? or do we just need to find out?

[pastel adder]

Yo chill bruh, not everyone is a Discord nerd Lmaoo

[echo aurora]

I am old, but I did work at Discord for awhile prior to LMArena

[tulip tree]

I can upload stickers if you want or give you them in dms

So you don't need to search

[pastel adder]

Why do you care about Stickers anyways?

[tulip tree]

Because they're cool?

[echo aurora]

I know how, I'm asking more-so for what stickers you'd want to see added. TBH I don't see many people using them so it's something I just never added.

[tulip tree]

Something funny

Like this

[pastel adder]

I agree with him, stickers are rarely used

and if you want funny stickers, there are other Discord communities for that

[echo aurora]

The best way is to use the model and find out. We don't currently have this info listed somewhere, but we do think it'd be helpful to have so we may add a solution one of these days.

[echo aurora]

Yeah if I were to add stickers it's be LMArena related, I don't think we'd add stickers that aren't related

[pastel adder]

Sounds good, I really do think the Claude sonnet 4.5 32k model is a heavy token usage. Been noticing needing to re open chats after talking to it for some quite a while

Well vibe coding really Lol

[polar niche]

Gday people!

[echo aurora]

hello

[polar niche]

pineapple the best mod

[echo aurora]

[gaunt spade]

fr

[sturdy mica]

Claude 4.5 Opus is out

[vestal hull]

Where I see side by sides where the model was actually revealed? How am I supposed to choose if I don't know which one was the better AI?

[sturdy mica]

its shown at the top

[vestal hull]

Hi. Thanks so much for your reply. Can you be more specific please?

[split spindle]

Hello, I am here to learn and share

[polar niche]

Where?

[knotty fable]

Butoh test: Winner: VEO, second place DDG - which got the place not for understand the prompt of interesting movement but for fludity and beauty. Bottom of the list. WAN and SeaArt - with techically good image but not understanding Butoh in any which way.

[echo aurora]

Are you referring to the generations in Video Arena?

[sturdy mica]

mate

[knotty fable]

As earlier I included the two AI's I generally generate with for my own stuf. Where one of those ended up last.
Today one arena video won. 😺

This is Butoh IMO.

[echo aurora]

That's awesome, thank you for sharing! Can you share the link to the generation? I'm curious what the prompt was.

[vagrant sandal]

anybody have a sora 2 invite code by chance please

[vestal hull]

I don't know where the videos are that have enough votes to see what engine was used

[echo aurora]

For the video generations to reveal what model created them, it'll need **at least 2 votes ** before that information is disclosed. #1397655624103493813 goes into a bit more detail about what that looks like.

[stuck orchid]

@echo aurora What about adding Google/Suno audio models to the arena? or should they ask to be added?

[vestal hull]

Correct. Where can I see videos where it was revealed which engine won?

[echo aurora]

#1372229840131985540 is the best place to put any requests for new models to be added to the platform.

[stuck orchid]

Okay

[echo aurora]

The best way to see this would be simply using the Discord search function and search their name.

[brittle ivy]

Good

[keen beacon]

Nice privacy policy lol

[echo aurora]

I'm sorry to see you're running into this error. Are you seeing this on multiple browsers? Were you able to get to a chat, or are you unable to access completely?

[keen beacon]

I have no issues everything works fine. But a limbo since i failed to accept the agreement and the arena still working with no captchas ?

How is that possible? That may be a back end issue maybe on your guys back end or the api with Google cloud? I’m guessing Google will be fingerprinting and then you guys also get the data as well as the actual company of the model we are using?

[echo aurora]

I have seen (it's rare) that Failed to accept terms-of-use error message appear. However, it doesn't actually fail, the error message is being triggered as a false positive.

[keen beacon]

I figured that much but just thought I let u know

[echo aurora]

But a limbo since i failed to accept the agreement and the arena still working with no captchas
For my clarity -> you received this error message, but everything works fine? If you reload the site, do you still see it or is it random if/when it happens to you?

[keen beacon]

No captchas period

It actually have to do more with Google now that I think about. Since i have all my accounts linked and use many Google services through out the day. Theoretically it’s the same captcha system across most web traffic

Using vpn triggers it heavy though

[storm needle]

@echo aurora bro i still keep getting infinite recaptcha loops

[keen beacon]

Use a better browser like duck duck go?

[hollow imp]

Bro

[storm needle]

complete the recaptcha then the model fails to send the response and if you retry it it just sends another recaptcha

[hollow imp]

Oxaam

[storm needle]

it ends up in an infinite loop

[keen beacon]

Dang

[storm needle]

It's not a browser issue

[hollow imp]

@storm needle

https://cdn.discordapp.com/attachments/1424855150517813399/1427269120818286633/image.png?ex=6904a8f3&is=69035773&hm=bb273a02e207cae7303ecbab715b2373d4459eced5009fb3e001c858aab50fec&

[storm needle]

oxaam is garbage you can't use any of their models on their account if you don't pay

[hollow imp]

Bro

My friend got access to their workspace

Like they accidentally gave away the owner account of the workspace

He added his email made it owner and kicked oxaam's email

😭😭😭

He's not sharing

[storm needle]

so what's the point of telling others this

[hollow imp]

What happened to your other account

[keen beacon]

U get free ai with duck duck go too

[storm needle]

what do you mean

[hollow imp]

Give free kagi api

[hollow imp]

"He threw a piece of bread at me"

[keen beacon]

For 10$ u get other models plus vpn

[hollow imp]

I fooled protonvpn with a fake cc and got 14 day trial

[keen beacon]

Haha

DuckDuckGo is cool for offering free ai

Older models but still

My favorite model to this dat is 4o mini

[hollow imp]

Bro

Lmarena doesn't?

[keen beacon]

They are I’m just saying it’s cool they do

[hollow imp]

Also poe exists

[keen beacon]

Ye but it’s convenient it’s put into the browser

Plus there comes a point in time where leeching isnt worth it I might as well just get a subscription. At least that was my thinking.

Especially if your someone who uses it daily for long periods of time

[upbeat ivy]

Hello! Happy to be here. I've been playing around with Aleph AI to make halloween effects, with mixed results. I'd like to try more options and hear recommendations

[echo aurora]

Really sorry you're still getting this. Do you have an Eval ID by chance?

[keen beacon]

Oh yeah I don’t have a logout button option on my other browser lol

[hollow imp]

@echo aurora yo what about the end of October contest and 1 month nitro winner

[keen beacon]

Deepseek v3 is really smooth

[echo aurora]

It's on my to-do, should be very soon.

[hollow imp]

Also if they don't want a 1 month discord nitro, I think you should give them the equivalent money in the form of their choosing

[echo aurora]

It's possible we make changes to future contest rewards, I'll keep that suggestion in mind.

[storm needle]

https://lmarena.ai/c/019a36e4-c7ac-7075-8943-7c20ee0411ce

[bleak jetty]

Fix the exploitable system before your crowdfund goes dry, I am surprised that the team hasn't came up with a solution yet

The recaptchav3 was working for a few hours before being reverted back for nothing

[worldly gorge]

hello

[gaunt spade]

thats dumb, making everyone suffer from captcha for every single prompt, just to ensure everything's "safe"

[gaunt spade]

you get a captcha for every SINGLE prompt you send

we dont need ts back

[echo aurora]

Okay, thank you.

[bleak jetty]

Not what I meant

[hollow crypt]

@echo aurora any issue with leaderboard ?

[bleak jetty]

The captchav3 I mentioned was just additional field in the rest api request payload

It's not the captcha that's been popping up

[echo aurora]

Reminder: ✅ Treat others with Respect.

[bleak jetty]

It's still in the payload whenever you make the request to the server but it stopped checking you can just put null or nothing

[echo aurora]

What seems to be the problem?

[hollow crypt]

Text leaderboard stuck on 16 Oct

[echo aurora]

They just haven't been updated recently.

[echo aurora]

Tor
What's that?

[rare lagoon]

hello

[echo aurora]

I am not sure, but will pass this along to the team.

[hollow crypt]

I've had hard time finding anything of value there

That yes but I was looking for other stuff

I'm against censorship as well

[gaunt spade]

e

[solid radish]

Hi, I sometimes have this problem. Anyone knows why I get this message after sending the next prompt?

And when i try refresh it happens again... and again

[undone saffron]

How many messages do you have in that chat?

[solid radish]

9

[gaunt spade]

u probably got rate limited

[undone saffron]

Are they very long?

[solid radish]

so i can send prompt only 8 times?

[gaunt spade]

how many chats did u have with it so far

in total

[solid radish]

one of them yes

[gaunt spade]

it depends on your total prompts

to the model

[solid radish]

2 chats

I mainly used it without logging in

[solid radish]

18 prompts in total

[inner trellis]

First time here. New to all of this, saw a video about it seemed cool and interesting, so I figured I would check it out. So hello to all.

[echo aurora]

hello

[inner trellis]

Hello, appreciate the response.

[wintry tinsel]

October needs to be over already

October is not the month of Gemini 3 so get it out

November on the other hand

Im a hoping!

[celest dragon]

guys

the best code LLM is claude sonnet 4.5?

[burnt sinew]

Yes practically. Though gemini 3 in ai studio with ab test is much better. I can get only a few requests per day

[sullen quest]

you can still get ab tests?

[burnt sinew]

I got one yesterday, not today yet though I haven't tried

Or i mean 5 yesterday i think

[sullen quest]

ooh

[burnt sinew]

What do you think i should use it for?

Definetely nothing that requires fixing code etc because it takes long to get another

Though maybe if I'm using like 10 accounts

[sullen quest]

so the method is just spammnig prompts for 2.5 pro?

[burnt sinew]

No, for 2.5 flash lite with thinking enabled.

[sullen quest]

why 2.5 flash?

[hybrid nymph]

How to change the ratio?

[burnt sinew]

Lighter rate limits

[burnt sinew]

Of what

[hybrid nymph]

ভিডিও জেনারেট করার

[hybrid nymph]

To generate video

[undone saffron]

That could be the reason
For example, with Claude, depending on the amount of text, there comes a point where, when I do PUT to the AI, it can't process the huge amount of text, causing the same error to appear no matter how many times I try again
The number of messages is completely irrelevant, since what really causes the error is the amount of text between you and the AI
Check this: #1417174113092374689 message

[burnt sinew]

Idk

@echo aurora I'm curious, will the pre release gemini 3 models be put on leaderboard

[sullen quest]

fair

[undone saffron]

Between claude-sonnet-4-5-20250929-thinking-32k and glm 4.6, which do you consider to be the best in programming?

[sullen quest]

I'd say claude is better

but glm is more cost efficient

[undone saffron]

I would say it's a tie, since the glm model has no chain-of-thought reasoning

🤔

Lol

I ask the AI if it can think
All its thoughts:

Claude thinks less than that

[burnt sinew]

Wdym

Literally?

[void brook]

Oj

[burnt sinew]

What does ability mean

[hazy kernel]

Ask chatgpt

[burnt sinew]

Like ability to use tools?

Then how is that any different from performance

[void brook]

1right

2false

3true

[burnt sinew]

How come? They all share the ability to produce text right

[void brook]

4true

5,6 true

7,8.9,10 false

11 true,12 false

[burnt sinew]

Most of them are super vague

[hazy kernel]

Article promo

[burnt sinew]

Reasoning and non reasoning llms have the same ability structure

Super ambiguous

[void brook]

How come true

How come,true?

[burnt sinew]

Send it to me, I'm guessing that it's a slop article though

[hazy kernel]

Do reasoning is the same as prompt enchantment

[burnt sinew]

What does this mean, thinking models can do things non thinking ones literally cannot

[hazy kernel]

Stop playing word this is confusing to me

[burnt sinew]

Ability structure??

Just making it vague

No, but im saying you could try to make less vague questions rather than engagement bait on general

Send me your article

[burnt sinew]

.

[polar niche]

Come on where is gpt 5 pro

[burnt sinew]

Where is gemini 3

[stoic swan]

yeoo

[sullen quest]

hi

[burnt sinew]

Hi

[undone saffron]

Wise

Well, I got that from the official glm website

Yes, I know

[hazy kernel]

why are there 2 types of api

and the free one with 205K context

nevermind I just checked the APIs they got different latency

[polar niche]

Yeah im gonna rinse my wallet

[odd arrow]

HELLLO FRIENDS

[jade egret]

😭

[polar niche]

ahahah

new fonts wow

[jade egret]

?

[polar niche]

look at my profile

[jade egret]

:0

[opaque crescent]

Guys, opus 4.1 has a limit right? So what's the limit, if u guys don't mind me asking

[verbal nimbus]

One is free (usually with logging for training), the other is paid. Providers may be different for each.

[hazy kernel]

Owhke,that's true

with prompt logging

[wheat venture]

testing

[brave cloak]

hey people

is image generation down for anyone else?

[halcyon nimbus]

seems fine

[noble lagoon]

Hey!

Anyone of you can please tell me how do I view the name of the video model? After voting it doesn't work. And overall what is video arena 1 2 and 3? Difference and also what video models are generally used and how to change aspect ratio of more specific things?

I am new here. Please tell me what I need to know.

[oak eagle]

"Hi, could you please help me? Every time I create a video, it turns out in landscape mode. I want it to be in portrait instead. How can I change that? Thank you so much!"

[empty stump]

What llm is best for brainstorming and planning

yeah i really need gemini 3 right now

[noble lagoon]

Same!

[prisma sparrow]

This is great community!

[boreal saddle]

Man, the captcha still hasnt been fixed?

[polar niche]

ADD GPT 5 pro already

whats taking so long 😭

[robust sluice]

captcha show up everytime I generate image, is that normal ?

[boreal saddle]

Lmarena is currently being crippled by a captcha popup on every single prompt, no matter how much you completed it, so right now you just gotta move elsewhere.

[halcyon nimbus]

it does with battle but not direct

[vast scarab]

Hey 👋

[feral lichen]

Guys whats Best ai for coding Python?

[eager stag]

hi sir

why lmarena don't adding video option to their official website?

[dawn ivy]

How to make longer videos

[dawn ivy]

I think you should command him for YouTube short

Anyone please tell me how can I make longer videos

[whole sundial]

the thing i don't really like about this server is all the confused people putting things in the wrong channels because they don't know what each channel used meant. maybe the names will have to made clearer. #ai-creations is being treated as a channel to create images/videos, same with #share-prompts
#codename-discussion is apparently now a place to share sora 2 codes?
#ai-memes is now an AI meme generation channel I guess
i don't even what is up with #leaderboards
i think the channels should be renamed like this:
#leaderboards to #leaderboard-discussion
#ai-creations and #share-prompts should be folded into a channel called #share-creations
#codename-discussion i'm not sure about, i don't think it's too much of an issue because people are asking for sora codes in other channels too
#ai-memes could be #share-ai-memes

[glossy umbra]

No this only if you spam it

[rough hornet]

Hi, I’m from the UK, I’m new to AI and I’m learning new stuff everyday. I find it totally fascinating, both AI Video and AI images. Great stuff

[teal mantle]

claude is not working

rate exceeded?

[hazy kernel]

At that point

just be

A mod

[teal mantle]

yeah some of us may want to

but where to apply lol...

[whole sundial]

yeah they have never opened any type of mod application since the server's boom started, instead they have like 20 different mods that came from... somewhere.

[teal mantle]

oh btw

I want some good general chat models that are smart enough

I have abandoned hope with OpenAI

Claude is good

best non thinking so far

at least Claude API did not break

but their website did

[whole sundial]

kimi k2

[fleet lintel]

why is Google not releasing Gemini 3? They have already done so much testing.
Few months back they were release half-baked models left and right .
Are they not feeling the heat from OAI anymore ?

[blissful chasm]

Because LLM itself is stagnating.
Constantly shipping in pursuit of short-term SOTA status is no longer important. Just like mobile phones, a new generation is released only annually and only a small percentage of people actually upgrade.

[hollow imp]

Google is known for slow and steady wins

Gemini 2.5 pro final version totally beat everyome

At the time being

[fleet lintel]

LLM is not stagnating.. far from it. Major improvements (especially agentic) will continue to happen for next 3 years for sure (probably longer)

[fleet lintel]

but gemini 3 is even crazier. I am just tired of them not releasing it

this is not a place to advertise. Dont spam and go away

[hollow imp]

real

[pulsar saffron]

How to ruin the ads marketing in this way:

Who will win?

[pulsar saffron]

😲 😲 😲 😲

YES, YES, YES, YES

[hollow imp]

No no no no

[pulsar saffron]

https://tenor.com/view/funny-dog-pet-lover-brothers-gif-24940061

[hollow imp]

They say 0 and charge 1$

[pulsar saffron]

oh

then no no no no

❌

[rough hornet]

Hi, if I purchase nitro, can I do more than 5 videos in 24hrs?

[sharp umbra]

why i can't send messages

can you help me please

[glossy umbra]

Every model that "thinks" has a weird feat

You can check their thinking yeah it'll say "So I'm implementing X" when X was already implemented

They don't think correctly it's just an illusion

[glossy umbra]

No sir

[keen beacon]

He did again it

https://www.testingcatalog.com/openai-updates-sora-with-credit-system-and-prepares-cameo-economy/

[keen beacon]

Open ai can go to hell

[boreal saddle]

Yeah, recaptcha still pops up over and over again.

And it is never enough to just tap the checkbox, no.

It always demands the infamous "Click all parts of the picture that contain XYZ" test

And the worst part is that it never remembers that i already passed it multiple times.

God, this is so annoying and timeconsuming that it is practically sitebreaking.

[gaunt spade]

wtf is that website

and how much does it cost

[keen beacon]

Sora 2

[gaunt spade]

wait u have the paid plan?

[keen beacon]

A pro plan lol

[gaunt spade]

the 200USD/month?

[keen beacon]

Yes turn off alarm. This is a total rip off.

[gaunt spade]

bruh i wouldnt pay that much to openai

its not worth it, the usage is very limited

atleast make sora 2 unlimited

[keen beacon]

I know, but you gotta pay to play, bro

But I hear ya this is a complete scam

[gaunt spade]

u get 30 free gens a day with sora 2 basic plan

and i got 3 sora 2 accs

so its 90 vids a day

[keen beacon]

Not no more lol

[gaunt spade]

why

[keen beacon]

https://tech.yahoo.com/ai/chatgpt/article/openai-now-sells-extra-sora-223905557.html

Because the AI industry is full of greed

And they grandfathered everybody in and the classic bait switch

Lucky for me I knew this was gonna be the case

Once again it’s gonna catch a bunch of people off guard.

I hope that open AI goes bankrupt and that ChatGPT will be installed in people’s bathrooms for the only purpose of flushing the toilet

So scummy

[warm zodiac]

Well it costs money to produce the video

Like what do you expect

[keen beacon]

I had no expectations. It doesn’t affect me either way. lol

Like I said, I knew this was coming

[warm zodiac]

its funny though because the AI industry is subsidizing so much use right now

[keen beacon]

Well, it’s just kind of ironic that they did it a few days after they went public

[warm zodiac]

like its kind of the opposite of greedy

[keen beacon]

I understand both points of view but there’s gonna outcry it’s already starting

[sullen quest]

Not public yet

[keen beacon]

https://www.reddit.com/r/SoraAi/s/KQzHz2MV8l

https://www.reddit.com/r/SoraAi/s/C4hpDDq737

The rest of the world hasn’t woken up to the news yet

This is only phase one

Well, first you need a pro plan at 200 a month

At .3o cents a sec ?

Even better

I’d rather use mini max

There 768 p better then Sora hq

Sora is a slop factory

It’s even got the same amount of letter in its name

Look guys I’m here isn’t that they’re trying to make money. I’m all for it. I got no problem with it.

Obviously, the technology isn’t cheap to keep this running and going, and if you wanna play, you gotta pay, I understand this

Real talk

Yes, I’m speaking from a pure consumer point of view

The AI industries ruled with many many consumer interest violations

What’s up, buddy?

https://youtu.be/jdqe5UVfbI4?si=qv0O_J4gOu-fOcqe

AI has become a commodity unfortunately

I’m not cooked 🙉

Why are you making it look like I’m spamming bro?

https://tenor.com/view/wack-crazy-serious-gif-14781154

[gaunt spade]

damn sora 2 died so fast

now we wait for opensource chinese free model

[keen beacon]

We gotta wait for the compute to be affordable until then it’s just gonna be convenient over. Wait time.

I doubt there’s gonna be any real affordable high-powered GPUs like that on the market for a while since everything has been kept behind a gate and need a lot of money to get going

[warm zodiac]

You just need to wait for algorithmic improvements. Always been like this with AI and free trials are even older. You can simply not pay it if you don't want to.

[gaunt spade]

u seem like a good cattle

for openai

[warm zodiac]

in what way

like I just don't understand how one can be a treatler for AI videos

[keen beacon]

Bro..l

[gaunt spade]

i have been with openai since 2023 (GPT3) and they were always greedy as hell

[keen beacon]

If you were a consumer, this isn’t good for anybody

I mean, do you like paying more and getting less when you go shopping?

[warm zodiac]

Have they? Why is there a generous free tier? You do realize the consumer surplus has been extremely high and this directly affects their order book

[keen beacon]

Why do you think do you think about that?

Because that is a really good question actually

Do you think they’re doing it out of the generosity of their heart?

[warm zodiac]

No, but I also wouldn't complain if costco stopped giving out as many free samples

[keen beacon]

Yeah, but I’m not on a free plan. I’m on a $200 month plan

With expectations of

[warm zodiac]

Okay then cancel the plan

like you aren't forced to pay

[keen beacon]

https://www.reddit.com/r/SoraAi/s/0wYrxPuy2D

Dude ur missing ur point

The world just hasn’t woken up to the news don’t be on the wrong side of history

[warm zodiac]

hahahahah the wrong side of history is crazy

we are talking about AI videos

[keen beacon]

Phase 1

Wait until credit is coming into ChatGPT

[warm zodiac]

did you think operating it was free?

[keen beacon]

I’m not doing this lol

[tulip tree]

1 prompt to chatgpt costs money

[keen beacon]

I’m a consumer I could care less about what it cost to operate. I’m not a big tech company.

It’s not my problem as a consumer

[warm zodiac]

right its also not their problem either and also not the governments problem when someone charges for a service

no one is going to be outraged for this cause. its a silly thing to be outraged about

you will get your open source model

[keen beacon]

Why do car mechanics have such a bad reputation or car salesman?

[warm zodiac]

they don't where I live

low trust society problems

[keen beacon]

Ok

[warm zodiac]

funny because those two examples you don't even need government intervention, just reviews and secret shoppers can provide info wrt quality

[keen beacon]

I don’t know why he even brought the government in with the government at all

[warm zodiac]

You're going to get an open source model, algorithmic improvements, Veo 4 will be subsidized for a while. Generations will approach the cost of inference. Don't worry about it

[keen beacon]

Well, except all the lobbying open ai did and cried nd complained when deepseek came out

[warm zodiac]

its all going to work out

[keen beacon]

Asking the government to demand it because they couldn’t keep up with the competition

There’s paperwork, bro

https://www.msn.com/en-us/technology/artificial-intelligence/openai-calls-for-allied-nations-to-ban-deepseek-models-security-concern-or-competitive-strategy/ar-AA1BdUlD

https://www.whitehouse.gov/briefings-statements/2025/02/public-comment-invited-on-artificial-intelligence-action-plan/

[warm zodiac]

yeah we don't have to listen to them right

also there is a genuine although weak argument to be made

[keen beacon]

Whatever man, I mean, I respect your point of you. I see where you’re coming from I do.

It really don’t matter to me. I’m just saying all the real ones know what’s up?

[warm zodiac]

I think mostly a ploy to secure compute restrictions to China

[keen beacon]

No, they made a deal in China doesn’t even want our chips

[warm zodiac]

Nah they really do

One of deepseek's training runs literally failed because they were using Huawei chips

[keen beacon]

https://apnews.com/article/nvidia-china-ai-chips-h20-trump-91588c36559bc881b8e010a9ed95cf0a

https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidia-to-resume-h20-sales-in-china-says-u-s-government-has-promised-to-grant-licenses-deliveries-to-start-soon

https://www.ap.org/news-highlights/spotlights/2025/us-will-get-a-15-cut-of-nvidia-and-amd-chip-sales-to-china-under-a-new-unusual-agreement/

[warm zodiac]

Yes those are not on the level of the B200, B20

[keen beacon]

https://www.cnbc.com/2025/08/22/nvidia-halt-h20-chip-production-china-cracks-down.html

[warm zodiac]

The compute gap between China and the US is growing and I think its fine

[keen beacon]

https://www.investopedia.com/nvidia-halts-china-chip-production-as-beijing-warns-over-security-reports-say-11795741

[warm zodiac]

Really is a national security thing

[keen beacon]

Not according to OpenAI

[warm zodiac]

Like if AI is even 50% real it absolutely makes sense to do export controls and hog all the compute

[warm zodiac]

all true idk what point you're trying to make

[keen beacon]

What is the point you’re making dude?

What gap are you talking about?

How is it a national risk? If what you said that there’s a gap between chips why would it be a risk?

I mean, competition is good for everybody lowers the prices for everybody

What are we afraid of?

[keen beacon]

That they could do it just as good for cheaper?

Yeah, they don’t want nobody dipping in their pockets

[warm zodiac]

China has many other advantages

So it makes sense to lean on ours

[keen beacon]

You mean the difference in prices lol

Yeah, that’s a big advantage. That’s why ChatGPT got cheaper actually.

[boreal saddle]

How are people then supposed to make anything better in the world though? Just giving up and leaving isnt really a solution to anything.

[warm zodiac]

It's a direct signal to OpenAI that they need to find a way to make this cheaper or offer more generation credits

[keen beacon]

lol

[boreal saddle]

The "Dont like it? Then leave" argument has never produced any good outcome in history.

[keen beacon]

All right man it is what it is

[warm zodiac]

Competition and price sensitivity are the basis for productivity gains

In markets its absolutely true that this produces good outcomes

especially in a market with extremely low switching costs hahah

[boreal saddle]

Not if the market is a oligopoly or monopoly, that is, and lets be fair, OpenAI is a giant.

[warm zodiac]

its not at all

OpenAI has a genuinely better product

There are dozens of video gen models

OpenAI is charging a premium because they have the best one

[keen beacon]

OK, that’s the difference

[warm zodiac]

Again this is the basis for productivity gains

[keen beacon]

Your wallet

[warm zodiac]

If you make a better product you can get away with charging more

[boreal saddle]

Gehlo s whole complaint was about paying a lot but still being screwed over, so the problem is not with the individual at all.

[warm zodiac]

And in that case if you think its not worth it, don't pay

Like no one is forcing them to pay for a bad product

[keen beacon]

I mean, whatever dude I’m just letting you guys know before the storm

[warm zodiac]

yes AI is going to cost money, I'm sorry, there will be nearly free open models that trail behind by a few months. overall I think this is a perfectly fine state of affairs. the consumer surplus is already huge and will keep growing

[boreal saddle]

And alternative open source AI models are simply not really viable since they just dont have access to the same resources OpenAI and similar Silicon Valley corpos scooped up and are thus of drastically poorer quality, so as a customer, you really dont have much of a viable choice in the matter.

[gaunt spade]

until the AI bubble pops in 2026, then we live normally again

[boreal saddle]

"It costs money" doesnt justify predatory anti consumer practices though.

[warm zodiac]

the good thing about an infrastructure bubble is that the consumer surplus actually grows after the bubble pops

like fiber buildout, railways

it actually made a lot of things possible that wouldn't have been since prices were depressed

[warm zodiac]

what about it is predatory?

like that's shockingly cheap

OpenAI is losing money on you

hard to call that predatory

[final idol]

Hi everyone, is there a way to create an intro like this?

[keen beacon]

Not really they’re really not worth much

[warm zodiac]

if they weren't worth 12 USD to you then you wouldn't buy it

[keen beacon]

People can’t even sell any because nobody wants them

[boreal saddle]

Hell, a lot of Big Tech would have to be broken up by antitrust action right now to actually free the market, but good luck getting that in the current political landscape.

[warm zodiac]

AI is remarkably non-concentrated though

[gaunt spade]

Yeah with sora 2 pro, but it costs lotta money to run

[keen beacon]

How you gonna call a consumer predatory but you can’t call out open AI practices as predatory

[warm zodiac]

Google Deepmind, Anthropic, OpenAI, xAI, ZAI, DeepSeek

Im not calling you predatory

[final idol]

OK thank you

[warm zodiac]

There is really no argument for a breakup

There is high innovation, 0 profit

[boreal saddle]

Yeah, he is basically calling you a leech while going out of his way to defend a company he likely doesnt even hold a stake in.

[keen beacon]

https://settingsinfotech.com/news/ai-news/openai-offers-chatgpt-to-us-government-for-1/

Gee I wounder why

[warm zodiac]

you do realize this benefits you

the government is going to buy AI services anyway

[keen beacon]

https://youtube.com/watch?v=TgYlOmIABjc&si=rY7BnWyBFqHU31jN

Oh, that’s right

[warm zodiac]

in this case the taxpayer isn't paying for it

[boreal saddle]

Remember how Republicans tried to sneak through a bill that banned any regulation of the AI/LLM industry for the next 10 years? Yeah, we really dodged a bullet with that one.

[keen beacon]

Trump already signed it into law

[boreal saddle]

I think they removed the "Regulations banned for 10 years" line from that bill later on.

[boreal saddle]

The modern robber barons, ladies and gentlemen.

[keen beacon]

The magnificent seven

And do you know if they have some of the largest AI companies in the world?

[boreal saddle]

Andrew Carnegie ould be absolutely weeping with envy at Jeff Bezos and Elon Musk s wealth.

[keen beacon]

And you wouldn’t believe where that picture is at

[boreal saddle]

The white house press room?

[keen beacon]

https://apnews.com/article/trump-inauguration-tech-billionaires-zuckerberg-musk-wealth-0896bfc3f50d941d62cebc3074267ecd

[tulip tree]

I would need a hammer to fix that hairline of Elon musk

[boreal saddle]

Oh, the inauguration

Welcome back, gilded age.

[keen beacon]

Just a coincidence that all them have large AI state of the art models

Or provide the services for with cloud

Do you think they have the consumers best interest in mind?

[boreal saddle]

They totally won't use it to further entrench GOP power or help crack down on their opposition.

[keen beacon]

Do you think it’s a coincidence all these layoffs or were seen on the news or also from the tech companies?

https://techcrunch.com/2025/10/24/tech-layoffs-2025-list/

[boreal saddle]

We have been in the second gilded age for a long time now. Trump just made it official.

Though this is more of a Silicon Age.

Or Chrome Age, if you wanna talk cyberpunk

[keen beacon]

Let’s see I’m just curious what are these companies worth with me? Let me look that up real quick.

https://www.fool.com/research/magnificent-seven-sp-500/

What’s that they made up 37% of the S&P 500 in October

Oh gee looks like they made a lot of money. I’m sure they can afford to get ethically sourced data to train their models if anybody could afford it could be them right?

https://www.tomshardware.com/tech-industry/artificial-intelligence/meta-staff-torrented-nearly-82tb-of-pirated-books-for-ai-training-court-records-reveal-copyright-violations

I’m sure this was an accident. They didn’t know what they were doing. They accidentally torrented and powered 82 TB of e books to train llama.

And you wanna know the funniest part that that makes us also comedic?

https://arstechnica.com/tech-policy/2025/02/meta-defends-its-vast-book-torrenting-were-just-a-leech-no-proof-of-seeding/

😂😂😂😂

The same platform that won’t let you upload a five second video on Facebook with copyrighted music on it

Or open a claim that China is gonna endanger intellectual property

This must’ve been an accident

Or this? I mean, they did write that letter to the White House claiming that China is gonna rip off intellectual property and that they’re here to respect it.

I mean, I rest my case

Say bye bye to ur Sora gens

For regular users, they’re gonna be like two videos a day lol

I generated six videos on the pro plan and I’m already out of credits lol

[fiery gull]

only six videos? lol the openai is hungry 🤣

[echo sinew]

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

[keen beacon]

4$ for 10 videos

https://www.reddit.com/r/SoraAi/s/iPvgEZalyo

[echo sinew]

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

[keen beacon]

“ can we get a credit Sam”

Notice how it says, and other advanced features across open AI

That means come to ChatGPT soon and codex

I wonder you’ll wake up that won’t even be a $20 subscription plan

It’s all gonna be credits

[hollow crypt]

Heh for the better

Let people try other cheaper and better ais

[keen beacon]

Here’s how open ai sell it to the public

[tulip tree]

Sora 2 expensive

[fiery gull]

Sora 2$ per video

[tulip tree]

Wtf

[keen beacon]

Four dollars gets you 10 videos

[tulip tree]

I rather buy candy

Or nitro

[fiery gull]

Openai needs to profit from the billions in losses 🙄

[keen beacon]

This is the reality that he thinks he’s living

[fiery gull]

They must have invested so much billions, and a chinese company without nvidea chip to do something like it with 200k dollars

[keen beacon]

Yeah they did They thought money would not dry up.

Who would pay for that to have that ugly watermark there?

Well, I talked to a guy earlier who didn’t seem to have a problem with this so as long as those kind of people keep on supporting such ethical companies

If the AI bubble pops, it was because of greed

I really love the heads up that they gave the users. lol

[misty marten]

hello

[fiery gull]

hi 🙂

[gilded torrent]

Hi guys is there any way to fix infinite generating responce?

[final idol]

Friend I bought sora 2 can you tell me the right prompt?

[brave orbit]

i think xAI since they win on ARC AGI and more

xAI 100%

for me

[hazy kernel]

what is j Berman model

[brave orbit]

yea but gpt 5 pro sucks for me i think yea gpt 5 pro was just not released when grok 4 was i thinked the current leaderboard for arc was groks demo of grok 4

[brave orbit]

but grok 4 makes it cheaper then gpt 5 pro but being ilke in the same almost one

if its cost per dollar xAI wins since gpt 5 pro costs much

[hazy kernel]

Mmm yes

[brave orbit]

my bad typeing didnt help

grok 4 is much cheaper then gpt 5 pro but at the same ilke agi level almost but much cheaper

[hazy kernel]

Gpt 5 pro scores more than grok 4 thinking but cost more $ than grok 4 thinking

[brave orbit]

but grok 4 thinking is almost the same level

[hazy kernel]

Yes I just retype that

[brave orbit]

grok 5 is gonna beat gpt 5 pro i think

[fiery gull]

I think gpt 5 pro < gemini 3 pro < grok 5, but the gemini 3.5 pro will be "agi"

[cedar tide]

When leaderboard update

[fiery gull]

when the dev wants

[gaunt spade]

we need grok 5 now

[hollow crypt]

Next week

[odd kindle]

Guys what happened to seedream

[hollow crypt]

Too expensive

[odd kindle]

damn thank god i used it as much as possible when it was up

[gaunt spade]

which seedream

[odd kindle]

seedream 4-2k

it was the best one

[gaunt spade]

i think its free on their website

if u signup with ur account

[odd kindle]

the website's one is less accurate for some reason

and you need a chinese vpn or something

[gaunt spade]

not really, i got free access with no vpn

[odd kindle]

oh cool

[gaunt spade]

in which way

[odd kindle]

its more cartoonish

[gaunt spade]

idk i havent tested it in a while tbh

[mint snow]

Why is the AI detector so faulty

all of the imgs rejected

[neon briar]

Is reve v1 working?

[halcyon nimbus]

pretty sure huny 3 is better than seed 4-2 anyway

[odd kindle]

for what purpose

[halcyon nimbus]

yeah seedream has better resolution i guess, but huny adheres to prompt better

[odd kindle]

thats true

[halcyon nimbus]

if you want mutiple characters doing different tasks with specific clothing and backgrounds nothing beats huny 3

[odd kindle]

what prompts do you usuaully use

[halcyon nimbus]

im usually glazing japan so like 'sakura background' lol

[odd kindle]

nice

[halcyon nimbus]

tbh i keep prompts simple

im not like '50 mm lense soft glaze' or something, i just throw in what i want to see and pray

my last prompt "two women, one with blue hair the other silver, wearing futuristic yet historical clothing, in tokyo (no kanji) in the future teleoperating mechanical android construction robots which are building a small house in a village (sakura nearby), small village yet futuristic"

huny nails it 99% of the time, its a step up from nano b

[odd kindle]

damn thats beautiful

[halcyon nimbus]

im not even sure how we are going to get better from here, yeah resoultion can get better but most people are on they phones so who cares

it maybe needs like 5% better to be photo realistic, but i dont even want that, i like the slight 3D animated feel to it, i feel there would be no artistic value if it was just photorealistic idek

like, im pretty sure it is only going to get WORSE from here ironically lol

my one complain is scale, had to give up on this idea, the women were just way too big

i give it like 6 months till all image generators are virtually identical and perfect and also free or runnable on a basic pc, our kids will be like 'papa tell me about the image generator wars'

llms i dont really see a ceiling on tho, the llm wars will last centuries

[valid helm]

A cinematic scene of a mysterious man standing on a dark stage, wearing a long black galabeya. His face is mostly hidden in shadow, illuminated only by faint red and golden stage lights. He sings into an old vintage microphone, his voice echoing across the empty theater. Thin smoke swirls around his feet, and torn curtains move slightly as if by a cold breeze. The atmosphere feels unsettling and spiritual, blending mystery, melancholy, and power. Realistic lighting, cinematic tone, deep shadows, dramatic composition.

[halcyon nimbus]

this aint where you type prompts but i got nothing better to do

[valid helm]

Thanks bro Please tell me where to write it.

[halcyon nimbus]

https://lmarena.ai/ direct chat, select image down the bottom, hunyuan image 3 at the top

i dont think you can even generate images on the discord?

only videos

https://tenor.com/view/confused-stare-at-paper-gif-5257839294650729742

[valid helm]

Got it

[halcyon nimbus]

i forgor i can just ask ai to translate, its an essay question about using natural resources or something idk

[wintry tinsel]

I approve of your pfp

[jolly current]

I made a video, where can I find it?

[vagrant barn]

anyone have sora invite code

[echo aurora]

I don't believe you have generated a video. Be sure to review the information in #1397655624103493813 as it'll show you how to use the bot.

[gaunt spade]

go on openai discord server bruh not here

[vagrant barn]

ok

[jaunty ridge]

#general photo The provided text content discusses how to use a platform for evaluating and developing AI models, but it does not provide information on how to create a prompt request replacing coffee with almondos and customizing text as requested.

To create the prompt you're asking for, you'd need to use an AI image generation tool, and phrase your request clearly and specifically. For example, you could try a prompt like: "Generate an image of almondos replacing coffee, with a down border of almondos. Include almondos bread. The text 'Focus Dose M.H.F.' should be replaced with 'All you need is', and place smaller almondos in the middle.

[cerulean elk]

#video-arena-1 Generate the video of a tardigrade

[jaunty ridge]

#general Oh, thank you so much for this I really appreciate it

[neon briar]

Is reve v1 working?

[ocean vessel]

Sora video watermarks now include the username.... 😯

[drifting crow]

Wow it looks just like in my dreams

[rotund crag]

Hola!

how to find, what is violated ?
❌ Generation failed. Your prompt violates our terms of use.

[echo aurora]

Sorry to say but there isn't a way to figure what exactly what the violation was for.

[median trout]

Hello

[frozen bane]

Can someone please tell developer to make video arena and choose the video model a website

[stray aspen]

Is lithium flow back

[echo aurora]

This feedback is on our radar

[whole sundial]

this

[burnt vector]

is there anyway to remove watermark on video

[echo aurora]

Yeah I've been thinking about making some changes. However, I'm not too hopeful this will stop users from putting in video prompt requests into these channels. Reason being those that make that mistake I don't think are paying too close attention to the channel names before posting.

[whole sundial]

you could also add some sort of word filter with a bot that tells people to use the right channel

[echo aurora]

Funny timing as I'm right in the middle of implementing this now.

[obsidian cargo]

Can you make it only apply to new users? Just in case?

[echo aurora]

It may take some adjusting, but right now I'm thinking simply if you type /video in our text channels ( #general #ai-creations #leaderboards etc.) it'll delete the message. This won't catch everything, but yeah need to find that balance of removing what we should but not preventing actual conversations.

[obsidian cargo]

true!

[hollow ivy]

Which model in LMarena has the largest context-window size?

Is it still only Grok4?

[wintry tinsel]

Guys it’s too late

Gemini 3 took too long too arrive

I went insane

[fiery gull]

Minimax m1 has 10mi

But is useless

[hollow ivy]

what about M2?

[loud robin]

anyone got a solution the video generators' issue in generating dialogue?

[echo aurora]

As in not producing sound?

[loud robin]

or is it just an unfortunate reality, that no model can be consistent in making good dialogue?

[loud robin]

no as in changing dialogue, overlaping...

i can only think of one solution for now, which is to produce one sentence by one character for every scene

[drowsy blaze]

hello all

[loud robin]

wassup

[thorny cove]

sora is paid now??

[nocturne rampart]

#ai-creations

[echo aurora]

Hmm yeah sorry to say I don't have a good recommendation for you here

[loud robin]

no worries my guy

[quartz pike]

ngl is it just me or is the ai image gen scene absolute buns rn? we are only image editing barelly image gen anymore

when are we gonna get gpt image 2. or google imagen 5

😭

[thorny cove]

i love imagen because i can give people i dislike shirts that say they are dumb

no i mean

imagedit

[viscid cloak]

@echo aurora wow! Image gen finally got quick Editing feature! Amazing one👍Btw, is the multi editing thing still existed within a single chat tab?

[echo aurora]

This is something we're experimenting with, so it's not fully released (meaning if you refresh you may not have access to the feature anymore).

multi editing thing still existed within a single chat tab?
In the same experiment you should see different behavior.

[viscid cloak]

Gotcha. Anyways glad to see progress and experiments 🫡

[echo aurora]

[ocean vortex]

Sonnet. Minimal is bad