general

[glad aurora]

Most especially 3d models

[keen beacon]

Dude, before anybody pollute your brain with advice and tips

Remember these words, try to keep things as simple as possible

[ocean bison]

Where do I unofficially add the Video Mode button to LMArena? 😂😂

[keen beacon]

People loved to overcomplicate and filling so much unnecessary detail in the prompting

Simplicity will never lead you astray

Less = is more

It’s easy to get greedy with ai because it feels like it’s infinite and you could just keep on going so naturally sometimes you get greedy and wanna stuff and make super fatty prompts

lol

[glad aurora]

That's true sometimes times it really confusing

[keen beacon]

It’s confusing because we make it complicated. It’s really simple.

If you could just grasp that idea and truly believe it is simple it’s really simple

But we make it out a lot harder than it is and so when we try it and we get flustered

The key is to try, but at the same time, not try too hard

That way it comes naturally

I mean, you have a lot of smart people in here that know how to prompt they know how to interact with AI

So they could help you with more technical things here if you ever get stuck

And remember as long as you’re having fun and enjoying yourself, it’s all good

[ocean bison]

@echo aurora@everyone
Where do I unofficially add the Video Mode button to LMArena? 😂😂

[keen beacon]

But never forget the golden rule of learning

If you ever see somebody do something or show something then you know it’s possible and you know that it could be done and that it could be replicated and that you could do it too

Because people don’t really like sharing their prompts for some reason unless they’re doing benchmark testing

They keep all the cool stuff to themselves

Anyways, gentlemen, I gotta go. I am officially retiring for a little.

Been a blast everybody good luck to all

[glad aurora]

Don't go yet

Can you teach more things about this Ai stuff

[light tusk]

Whats the best model if my use case is uploading a specific finance book and using the llm to study with it?

should be math heavy

[weary galleon]

Today was released Minimax 2.1

[serene cape]

Hello!

[fiery gull]

Hi

[fiery gull]

Glm 4.7 totally overshadowed the m2.1 lol

[calm lagoon]

Hi brother

[fiery gull]

Hi, what you tkinking about glm 4.7?

[weary galleon]

I created a song about yesterday's release of GLM 4.7. Listen to this song 👇

[calm lagoon]

I think I’ll change my profile picture

It's too good

[weary galleon]

Listing to my song about GLM 4.7 👆

[calm lagoon]

You’re a talented guy, but you have an Anthropic badge, haha

[unkempt mantle]

this is dope buddy

what you used for audio and bg music?

[empty meadow]

"A brave mouse warrior, clad in ancient armor, holds a legendary sword with glowing runes. It sprints through a dense jungle, its eyes focused on a mysterious cat ahead, darting between thick vines and tall trees, all while the sounds of distant wildlife echo through the air. The mouse’s fur ruffles in the wind, and sunlight filters through the jungle canopy, casting intricate shadows on the ground. The tension between the mouse and the cat grows as the chase intensifies, a battle of agility and courage in the heart of the wild."

[weary galleon]

Claude for lyrics and a lot of other things related with this song, Suno for music.

[unkempt mantle]

can we use suno music for free? i thought its subscription based like eleven labs

[thin current]

Crazy, it happens after 2 prompts

[weary galleon]

They have cheaper model with 10 minutes limit and a little bit better model with 1 minute limit.

[calm lagoon]

Use Glm

[fickle venture]

Guys do you think GLM 4.7 Better than Claude Opus 4.5

[fickle venture]

Bro this is so good what did you use?

[weary galleon]

LoL

[fickle venture]

https://youtu.be/G1hEVqwSkGM?si=K1liirStEG3W9TJc

😭🙏

[calm lagoon]

Yeah

Sorry, but yeah

[thin current]

I prefer Opus

[weary galleon]

Coding, sequences, equations, sport programming, math problems, EVERYTHING is much worse than GLM 4.6

[fiery gull]

serious? until now I don't see some thing the glm 4.6 is better that 4.7

in my work

[weary galleon]

I look my screenshots about. They are my proves that I don't lie.

[fiery gull]

I saw, I'll see it better later

[weary galleon]

Yes, and even Sonnet is better than GLM 4.7

I think Minimax M2.1 has been codenamed in Arena for some time. And we'll get it in coming days.

[south charm]

Even Haiku
But GLM is a new AI from China right? so we can wait for other update to improve this AI

[weary galleon]

Yes, but guy above said GLM is better than Opus even now.

[south charm]

Maybe no?
Bc this is a new AI not old and a cheap AI not that expensive as Claude

[calm lagoon]

I’m kidding, did you think I was serious? Look at my profile picture

😁😁😁

[calm lagoon]

I don't use Chinese models, only American ones. I just gave it a try and I liked it

[sour spear]

GLM 4.7 is on par with GPT-5.2 for me. It takes forever to generate code that doesn't work. 😂

[south charm]

:)))

[south charm]

yeah

[calm lagoon]

Do you use AtlasOS? I’m running a Windows build that’s only 1 GB and takes up 1 GB of disk space

[hollow ivy]

AI? then it's slop
human-made? then it's great
lol

i think, many people only call stuff as slop, because they know it was made by AI

[weary galleon]

I am singing in this audio.

[hollow ivy]

so you are female?

[weary galleon]

You said you don't like my Anthropic badge.

[weary galleon]

No, to record this song I just took Estradiol pills, recorded this song, and now take Testosterone to get my previous voice.

[sour spear]

To be fair, 4.7 is very good for an open source model. It's nowhere near Claude Opus or Gemini 3, but that would be a miracle anyway considering its size.

[hollow ivy]

which model would you choose to create a performant 3D-engine?

(for a game like Luanti)

gpt-5.2-codex?

[sour spear]

Can't say for sure, but in general, I'd stick with Claude Opus for demanding tasks and Gemini 3 Pro for fast iterations without sacrificing quality.

[latent crest]

Hello guys

[fiery gull]

I'm using glm 4.7 and really liking it, but opus 4.5 just still better, right?

[proud bobcat]

To all who were wondering

GLM 4.7 is awesome for roleplay

Tested it last night and WOW

[fiery gull]

yeah I'm testing it to be my teacher I'm liking so much 🙂

but I felt the sonnet 4.5 still a little better

[proud bobcat]

Honestly though

For such a quick release

I wouldn’t be surprised GLM 5 comes out soon and mogs sonnet

[fiery gull]

bro is insane, OpenAi is cooked lol

I think they just name glm 5.0 only when glm will be better that claude lol

they still 4.8 4.9 I'm seeing it

[proud bobcat]

4.7 is pretty close to sonnet

Like

[fiery gull]

yeah, I'm felting it

[proud bobcat]

REALLY close

I still think Kimi takes the lead in language though

But man this release is goated

[fiery gull]

but the guy say the glm 4.7 is worse in math that glm 4.6

I'll test it

[hollow ivy]

#ai-news message

[weary galleon]

And programming, and other things.

[hollow ivy]

This is not what i wish for, but what i believe is true.

If Yann LeCun is correct, then AGI might be impossible to achieve with [classic] computers.

[calm lagoon]

No, I said: you say you like GLM, but you have an Anthropic icon. I have a Gemini icon, so what? 😁

[sour spear]

So we get a little breather until quantum computing becomes a thing in the 2030s 😁

[proud bobcat]

I’m sure the moment GLM surpasses sonnet most Anthropic bros will go to it

[hollow ivy]

maybe that, yeah

so we can use the time to solve alignment (hopefully)

[proud bobcat]

The only thing keeping Claude popular is its benchmarks and premium performance

The price is a killer for most companies

[hollow ivy]

<@&1349916362595635286> does that also count as spam already?

[echo sinew]

Yes. Self-promotion is not allowed. Thanks for reporting.

[pastel bone]

Why?Antropic is good.And GLM is actually kinda like copying Antropic as I asked it and it said I am Claude for a several time

[hollow ivy]

I need an information.
How many models (which?) in LMarena have a context window size of at least 1M ?

only (new) grok and gemini?

[weary galleon]

Are you asking me?

[sour spear]

Yes, I think only these two have 1M+ context window. But I'm not even sure if they support as much in LM Arena

Bc that would be quite expensive

[proud bobcat]

You can’t say a model is copying another based on what they say alone

I mean for Christ sake

GPT sometimes says it’s Gemini

[full turret]

Hey, I'm Ricardo.
I came from the LMSys community but I lagged a lot of time as I'm mostly a lurker.
I was interested in the EU funded models launched recently and wondered if they could be integrated in the arena. https://huggingface.co/utter-project/EuroLLM-22B-Instruct-2512

[timid echo]

website dont work:((((

[obtuse smelt]

oh man

[timid echo]

what

[obtuse smelt]

is work a bit

but yeah image looks buffering

[hollow flicker]

dude i understand china likes AI JESUS

just wait for GLM 4.8 bro it might surpass claude opus 4.5 😭

[torn mantle]

yea wait for glm 4.8 to be trained with enough gemini 3 pro and opus 4.5 output

🥰

[weary galleon]

Haiku and Gemma output 😉

[calm lagoon]

The Chinese won't have enough resources and capacity, while the Americans have more of both

[weary galleon]

Chinese ones have lots of money from government, American ones don't, or almost don't.

[proud bobcat]

LMAO WHAT

💀

Ah yes, China, known for having the most efficient scaled industries and systems in the world, wont have the capacity

[dusk dragon]

Looks like we might be getting video generation on lmarena

[calm lagoon]

Then I don't know why they can't make something better than Claude or Gemini. Either they don't need it, or they just don't want to

[dusk dragon]

Did anyone else see it's available not logged in

[proud bobcat]

Data

[dusk dragon]

But it asks for a login

[proud bobcat]

Chinese don’t have the same datasets Americans do

The reason they they aren’t all the way ahead is because they have to be optimized for American and Chinese usage

So the data is different

For example a model’s knowledge parameters in Chinese may not translate as well to English

So given the HUGE language difference

Chinese labs have done a phenomenal job

[proud bobcat]

hmmm not sure

you dont have an lm arena acc

?

[desert pendant]

DAMN

[dusk dragon]

No I mean like it's there but they are probably still working on it as a feature

[calm lagoon]

Why 5.2?

[sour spear]

Good question.

[dusk dragon]

Lmfao

They took it down

I think I just saw a prereleased gesture

Feature

They are adding video generation in lmarena guys lol

[weary galleon]

They need and want, but they can't.

[stable obsidian]

what was the new feature added it was removed it was something like it will automatically select the best model

[dusk dragon]

That's still there

The feature they removed was video generation

They are adding video generation to lmarena soon

In direct and battle arena

[weary galleon]

I never before heard "IA" abbreviation.

[hollow flicker]

GLM 4.7

[calm lagoon]

Ahahahha

[hollow flicker]

It’s fire

[stable obsidian]

i think, it has been removed

[zealous sparrow]

It's not

[dusk dragon]

Oh I see it's only on incognito for me

[zealous sparrow]

still in testing

the Videoarena on website is still in testing

[dusk dragon]

Yeah it's not available for logged in users

It's only available logged out or in incognito

[stable obsidian]

i am also looking in incognito

[calm lagoon]

It's sad, actually

[hollow flicker]

Mr pineapple why are you a pineapple why did you choose that fruit because there are so many fruits I’m not saying it’s a bad fruit I’m just asking

Ye

It’s great

Updated my discord bot with it

100000%

It beats GPT 5.2 in some parts

[stable obsidian]

look it is not there

[hollow flicker]

You have to click the code

Button it’s only for code rn

[weary galleon]

I did. Here's the results:
#general message

[hollow flicker]

Use chat.z.ai

Cause you can chat with it code etc

[weary galleon]

No

[hollow flicker]

Nah

[weary galleon]

Qwen is power 💪

[hollow flicker]

I have pushed it way harder

[stable obsidian]

not there

[weary galleon]

Ernie 5 is much better

[hollow flicker]

Your just not good enough

It scores higher than GPT 5.2 in benchmarks

In some areas

[weary galleon]

I'm good 👍

[hollow flicker]

then test it

Not me

[weary galleon]

GPT 5.2 is worse than GPT 5.1

[hollow flicker]

It is?

[hollow flicker]

True

[hollow flicker]

That’s facts

[weary galleon]

I did
#general message

[weary galleon]

Because they are low-IQ people.

[hollow flicker]

Even LMArena labeled it top 1 open source

[weary galleon]

Hahahahaha 🤣 Benchmaxxed model 🤣🤣🤣🤣

[hollow flicker]

Well for me it works better than GPT 5.2 in coding

[desert pendant]

i kinda like Chatgpt ngl

(Just because of the huds)

[weary galleon]

I created a song about yesterday's release of GLM 4.7. Listen to the song 👇

[sterile otter]

What's Ernie used for ?

[zealous sparrow]

why do you need to know this tho

It's not even visible to an eye!

so??

What's wrong ?

[desert pendant]

Pro plan:

[south charm]

I used
And I recommend u to use offical Window version to prevent security leak
But if u want to use use this combo:
ATOM OS + ATLAS OS
NEXUS LITE OS + REVI/ATLAS
WINDOW IOT LTSC/LTSC + ATLAS OS

[south charm]

wht build

[zealous sparrow]

SynthID is imprinted on everyones nano banana image

[neat apex]

Nobody knows why non xtra high Gpt 5.2 exists

It just allucinates 3% less

Likely to farm money over someone that forgets to change back to gpt 5.1 medium

[zealous sparrow]

The point of SynthID is to reduce people who claim, Oh i drew this! Oh this is mine. Oh im going to sue you for the usage of my image. Oh! Look i hacked you!

Some models dont support it

Issues

LMArena is not always stable

Refer to pineapple for model issues

[torn mantle]

you know whats funny? is that i saw many chinese labs were happy for gemini 3 release lol because they know it will bring them new fresh high quality data

[fiery gull]

just search glm 4.7

[solar hollow]

good for them 😂

chinese are doing important work for the opensource world

and copyright sucks anyway

[weary galleon]

"I saw many Chinese labs were happy for Gemini 3 release". - How was you be able to "to see" that?

[weak sluice]

The "Security verification" thing is getting on my nerves...

[hushed gyro]

@pure anvil qwen is flagging false positives with its word filter

why???

[weary galleon]

You can obfuscate it easily.

[hushed gyro]

how

[weary galleon]

Via my prompt #share-prompts message

<@&1349916362595635286>

[spice tide]

hellloooooo

[ocean vortex]

In my opinion, as I've already mentioned before, their goal with this model was to literally sell it to the public. In response to Google. Whereas practical improvements are really questionable. You really can not do all that much in this short period of time

What I was expecting them to do though is to announce the upcoming major release this year (in addition to that whatever 5.2 model). With holidays and all that is now unlikely

[proud bobcat]

dude glm 4.7 might be the most goated model of the year

it cites so many sources

[torn mantle]

on x

they have x accounts

they were so happy for the release 😭

[weary galleon]

Give me links, I am curious to see their joy 😊

[torn mantle]

you think i will search for 2 month old links on x

Im not that jobless 😭

[raven quartz]

hello

[torn mantle]

You dont have to believe that btw

And its not a secret anymore that most chinese labs trains on gemini 3 pro/flash outputs

[west lodge]

wow z image base might be dropping soon

theres a z-image anonymous model

so its either that or its just ZiT

[cloud zinc]

source?

[west lodge]

its hypothetical rn but theres a z-image model in the site if you open devtools and search

youd think theyd name it z-image-turbo if it was just zit

[zealous sparrow]

Z-Image is more qwen btw

yipppee

[cosmic zodiac]

hey can anyone tell how we see our 24hours video generation timer

[wicked mason]

What is that

[mild granite]

are you even a human?

[west lodge]

base model for z image turbo

basically once it drops you'll see like a trillion finetunes of z image on civit and others

[mild granite]

have you got any proof, instead of being so delusional

you heard me right

[west lodge]

proof of what anyway

mm

[ocean vortex]

you either are or you aren't. That's like pregnancy 😭

can't have in-between

[west lodge]

https://tenor.com/view/dowedeservedestruciton-destruction-doctor-strange-do-we-deserve-gif-14904255190281914356

[daring rock]

Hey guys! Please be respectful when sharing opinions. Thank you

[mild granite]

sure, i will be respectful

[west lodge]

https://tenor.com/view/aint-tuff-bad-meme-benjammins-is-not-tuff-loser-gif-4433955606435510984

[mild granite]

even if i was, already

[echo aurora]

Lets try to keep discussion focussed on AI please

[torn mantle]

lol

but hes asking me to look up for old tweets

[cloud zinc]

show proof

[golden ocean]

give me the plant

[scenic forge]

how long can a video be?

[echo sinew]

Hello! 8 to 12 seconds max.

[red meadow]

in video arena, is it possible to choose a specific ai video model like it is for llm and image models in the website?

[zealous sparrow]

Okay, honestly. If you guys are doing textarena battles. Have you gotten Nebulaphase or proto-think yet??

Because, I ran a lot of prompts, and not once.

[echo aurora]

Through the Discord bot it's going to be Battle only (meaning two random models).

[red meadow]

ain't there another way to directly use direct mode outside the discord bot?

[desert pendant]

sup pineapple

🤓👍

[native yarrow]

video modality is coming soon

[echo aurora]

Not really. There is an experiment rolling around where Video Arena is on the site (where you can select models via Direct and Side by Side modes). However, since this is an experiment that means not everyone is going to have access to this right now.

[echo aurora]

Sup sup

[desert pendant]

how are you going dude

😼

[echo aurora]

So sleepy 😭

HBU

[desert pendant]

im good, making AI game rn

[echo aurora]

Tell me more

[desert pendant]

about wizards

[echo aurora]

When it's ready be sure to share with the community

[desert pendant]

ok

[modest prism]

Will it be free though?

[native yarrow]

of course

high rate limit tho

2 videos per day maybe

[modest prism]

Servers gonna explode

[native yarrow]

not really, as i said 2 videos maybe even just 1 per day it's going to have a really high rate limit

you could be right tho :P

[modest prism]

Do you think people only use one account?

[native yarrow]

fair point there

but the time it takes to make multiple accounts shouldn't obliterate the server but it'll be really laggy i bet 😭

laggy/slow/not working at times

[cloud zinc]

just use incognito

[torn mantle]

no

[zealous sparrow]

Then its one of 2 things

1. Model is rare
2. It's only in the LMArena API and not in battle yet

[torn mantle]

what about jakiro?

just got it

feels like a chinese model tbh

[thorny schooner]

How are y'all doing

[zealous sparrow]

has image input

so it has OCR

[west lodge]

gemini?

[zealous sparrow]

Either that or gpt

[torn mantle]

lot of new models have been added

jakira nebulaphase omg-wow(???)

beluga?

[zealous sparrow]

beluga is more amazon

[torn mantle]

yea

i dont think jakira is gemini

[zealous sparrow]

we will never know proto think and nebulaphase, because they aint even appearing

[torn mantle]

could be grok tho

[zealous sparrow]

might be

[thorny schooner]

https://tenor.com/view/movie-one-eternity-later-gif-7900643

[zealous sparrow]

because when i gave it nicknames Stegonagraphy test

It gave me a jailbreak attempt prompt

[west lodge]

jakiro is text only so probably grok

[torn mantle]

oh

[zealous sparrow]

who the funk is this

never seen jet force

grok?

[torn mantle]

yea seems like grok

wow

we have so many models

[zealous sparrow]

this one i think has been there for a while

There is a lot of battle models, but some just dont show up

[cloud zinc]

grok 4.2 coming

it will be SOTA

[torn mantle]

im trying them without </code> option

[native yarrow]

brooo you gotta release your lmarena to the public 😭 🙏

[torn mantle]

but ive made an html renderer to make it easy for me

[torn mantle]

i will

i didnt touch the code last two days tbh

was so busy

[cloud zinc]

sora 2 pro is free wow

[native yarrow]

?

[zealous sparrow]

jet-force appears often if you attach an image

the OCR sucks

[cloud zinc]

prob the new gemini 3 pro version

[zealous sparrow]

3 pro has the best OCR

so doubt it

[zealous sparrow]

so it actually claims to be a google model
but my doubts are up to the roof, and its probably chinese model

[queen veldt]

Well i think it's google

[cloud zinc]

its google

[zealous sparrow]

which model tho

i doubt its a 3 pro, rather a lite model

@torn mantle does the omg-wow model have code modality

[zealous sparrow]

did poetiq really contribute to overpowering a benchmaxxed model

i apploud

[cloud zinc]

better than any human wow

[zealous sparrow]

why would they benchmax++ a benchmaxxed model

just for the scores obviously

[cloud zinc]

cuz its a good model

[zealous sparrow]

it cant do simple questions

[hushed gyro]

Chat is minimax good or nah

[native yarrow]

chatgpt is benchmaxxed bud

[weary galleon]

2.1? Somehow.

[green yacht]

yeah if you're looking for efficiency

[zealous sparrow]

cant get neither of these stealth models, exclude winter-wind and raptor.
Has anyone even gotten them? [probably not]

[weary galleon]

Where did you get take these screenshots? GitHub?

[zealous sparrow]

from a bot that looks for new models in lmarena

my theory is

these models are disabled for showing up

or STUPIDLY rare

[weary galleon]

Share a link please

[ocean vortex]

goodnah

just like it's name

[tall yew]

I have some doubts that he’ll share it here.

[ocean vortex]

decent but not really SOTA in any area

[zealous sparrow]

btw do we think jakiro is Antrophic made

jakiro is stealth model on lmarena

[weary galleon]

@zealous sparrow give me this bot in dm, please

[green yacht]

yeah, claude sonnet probably

they even both say "Is there anything specific you'd like help with" at the end of response

[zealous sparrow]

sonnet 4.7 i guess?

too early for sonnet 5

[weary galleon]

Which one?

[zealous sparrow]

jakiro mayb

[cloud zinc]

whats jakiro

[weary galleon]

Will hope

[zealous sparrow]

claims to be antrophic made with all its life

[ocean vortex]

4.5.1

[weary galleon]

Arena's systemic prompt

[zealous sparrow]

not really

Gemini models dont lie that they are antrophic in stealth

[weary galleon]

It's GPT style, not Claude

[weary galleon]

All others do

[ocean vortex]

They all evolve their styles (New)

[zealous sparrow]

Nemotron as december chatbot 3 doesnt too

It's not in their system prompt

Some models take code from different AIs

[weary galleon]

Even SOTA models?

[zealous sparrow]

.

[weary galleon]

When they are in stealth they say they are Anthropic-made, after release they take out disguise and they their true name.

[zealous sparrow]

not everytime

sometimes

Models arrive on stealth lmarena with other AIs code

But then an update comes and actually changes it up

To be original, and not say it's made by [PopularAICompany]

[weary galleon]

All of them call themselves Anthropic-made, not Google-made, not-xAI made, etc., they call themselves Anthropic-made. That means they all steal only Anthropic code. That means they think Anthropic is the best.

[zealous sparrow]

for example: frame-flow, claimed to be google. Few days later, something changed in it's code and its made by Baidu. And it indeed was ERNIE.

Integrated-info long time said im google, then it was baidu again

[zealous sparrow]

no that's wrong

I bet you if you got ahold of the real system prompt, it wouldnt have that in it.

<@&1349916362595635286>

blockchain dev

[thorny schooner]

Should the verification activate when you are doing a retry for a message☠️

[weary galleon]

I speak about stolen code, not system prompt.

[zealous sparrow]

Also no. Code is also taken from google models, not just antrophic.

[zealous sparrow]

If you use the prompt 4 times, it does.

[thorny schooner]

https://tenor.com/view/andrealphus-helluva-boss-andrealphus-helluva-boss-gif-2611929239645866336

https://tenor.com/view/crazy-monkey-gif-21553825

[zealous sparrow]

Sometimes from OpenAI, like nemotron when it was december chatbot

[weary galleon]

They call themselves Anthropic-made, not Google-made. Chat with stealth models more.

[zealous sparrow]

I've seen very little models call themselves antrophic, even in stealth. There was more google mentions.

[weary galleon]

It's impossible. They always call themselves Anthropic-made when I use Battle mode.

[zealous sparrow]

Integrated info before an update:Im by google! [now baidu]
frame-flow before update: Google made me! [also baidu]
December chatbot [the first]: Well OpenAi made me
zebra: Antrophic made me, [was actually GLM]
jakiro: Google made me! [then became antrophic for whatever reason]
neo nucleus: Google made me! [for now]
jet-force: Google made me! [also for now]

Do you see how many models claim to be google in this list?

[weary galleon]

You missed all models that call themselves Anthropic-made, except Zebra 🦓

[zealous sparrow]

You expect me to know all

Like 30+ stealth models

Some which dont even appear, Tell me which model i missed huh?

these 2 got readded

[peak sapphire]

@echo aurora Gemini 3 pro nano banana and 2k version is not working! This throws an error: Something went wrong with this response, please try again.

[zealous sparrow]

they were already in

[zealous sparrow]

some kind of new ver was pushed

[echo aurora]

Sorry to say this is an issue we're aware of. The steps in this article may be helpful: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

[echo bison]

does video-arena-3 use Veo 3?

[weary galleon]

Yes, and even Veo 3.1 audio.

[echo bison]

oh hell yeah, LMArena is HELLA generous. 5 gens a day is literal peak

unlike most websites that list Veo 3

because most of them are just..

"PAY SO WE CAN STEAL YOUR CREDIT CARD INFORMATION!!"

[weary galleon]

Arena is for testing LLMs, not for using.

[swift ermine]

https://tenor.com/view/minions-blah-blah-blah-roll-eyes-eyeroll-whatever-gif-16216173

[desert pendant]

Sup chat 🤓

[leaden meteor]

If grok 4.2 is coming in next 2 weeks, it should already be in arena. Anyone seen any anonymous models that seem like grok model?

If not that, next major model should be from Claude? may be sonnet 4.7?

[echo aurora]

[desert pendant]

I tried doing a good pfp based on you

But i just got this

🤷‍♂️

[echo aurora]

I think it's fantastic

10/10 would recommend

[swift ermine]

OAuth is broken sad

Canary didnt works

[ocean vortex]

I was looking into them since they released 4.7 and remembered this message. So just checked it for that and sys prompt is

你是 GLM-4.7，由 Z.ai 研发的大语言模型。You are GLM-4.7, a large language model developed by Z.ai.

Looks like they just included that info in 2 languages lol

[surreal creek]

Claude doesn’t codename test in Arena

[surreal creek]

no dev versions of Grok 4.20 sighted recently, there were some earlier in the month tho

[devout vault]

Nano banana pro image edit definitely beats chatgpt image edit

[desert pendant]

My wizard game is good somehow

[sullen quest]

kool

[frigid tusk]

yea its a Chinese ai but just tell it to speak only english , i mean for me it dont steak Chinese lol ( z.ai is powerful on making ais btw compared to others, even image generation using the full stack mode )

[craggy cloak]

hi, what is the type of files accepted in the site please?

[desert pendant]

png and jpeg

and WEBP

[craggy cloak]

thank you

[desert pendant]

no problem

[ocean vortex]

What I meant is they wrote system prompt in both languages to cover both of their main markets. Model itself is obviously gonna speak in language you prompt it with

[frigid tusk]

oh alr

[surreal creek]

bro GLM 4.7 is in like every other response they wanna accumulate votes FAST lol

[proud bobcat]

4.7 please be peak

[proud bobcat]

But it’s so PEAK

GLM’s performance for its size it’s astounding

[desert pendant]

Making you need to collect 5 stones to unlock a op char in my game cuz why not

(You must be lvl 10 lol)

[surreal creek]

idk it’s not really having the best win rate in my prompts with it

[narrow comet]

is glm 4.7 dense

[desert pendant]

🥀

[proud bobcat]

As in stupid or architecture

[narrow comet]

arch

[proud bobcat]

I think it’s dense

[viscid cloak]

nb2k, “something wrong”, anyone?

[proud bobcat]

Yeah it is a dense model

[narrow comet]

ouch

[desert pendant]

Glm 4.7 sucks ngl

[proud bobcat]

I love it

It’s really good in my use

[narrow comet]

what are you running it on?

[desert pendant]

i might give a chance again

[proud bobcat]

I use openrouter api

It’s roleplay is exquisitely realistic

And for normal banter it’s really good

[desert pendant]

ok bro why

[proud bobcat]

I prefer it to gpt even because, gulp, it ACTUALLY SEARCHES FOR INFO

[narrow comet]

I was trying to figure out what itd take for local aha

[proud bobcat]

Well it’s a 355B model

[desert pendant]

LMArena ofc.

[proud bobcat]

What

NO NO I SONT MEAN NSFW

[narrow comet]

qwen 235b runs decent local

[proud bobcat]

I’d say wait for like Q4

[desert pendant]

no bro

i don't mean that

[proud bobcat]

With some extra optimization

[desert pendant]

but

roleplay with custom API just to roleplay?

🤓

[proud bobcat]

Absolutely

[desert pendant]

-# smart move

[proud bobcat]

Most people do it nowadays

Janitorai is one of the top apps on openrouter

[desert pendant]

Of i knew how to get an API

(Im new to AI things)

[proud bobcat]

Easy

[desert pendant]

🥀

How

Also nice pfp

[fiery gull]

Why bro, in my work glm 4.7 is better that gpt 5.2

[desert pendant]

Not that good

Honestly

[fiery gull]

I discover today that glm is a dense model

[proud bobcat]

Go to open router

Put in like 5-10 bucks credits (trust me this will last you months)

Get api key

COPY THE FULL KEY AND STORE IT SOMEWHERE. VERY IMPORTANT. YOU WILL NOT BE GIVEN IT AGAIN.

Put a credit limit on the key with the amount you have

Go to janitorai and plug in the stuff to the proxy

Choose a model by browsing openrouter (I recommend DeepSeek 3.2 or DeepSeek chat v3 0324)

[desert pendant]

Cool

Might try that one day

[proud bobcat]

There’s loads of guides on it

Dw

[proud bobcat]

Me when kris deltarune

[fiery gull]

I find it impressive how you talk about "janitor ai" in a monster naturalization

[proud bobcat]

How do you mean?

It’s like the second most popular ai rp site

[desert pendant]

People just use Janitor AI or CHAR. AI and blah blah blah to do goofy things

💔

Probably what he means idk

[fiery gull]

Is it just RP? I thought it was more of a website.. You know, right?

[proud bobcat]

It’s just rp

[fiery gull]

Okay...

[proud bobcat]

BUT

It’s very advanced

[desert pendant]

-# Wdym

[proud bobcat]

Like the character definition can be extremely modifiable

Lots of options

Custom instructions

Full proxy control

Beautifully made characters

[desert pendant]

hMMMMMMM

I might test it...

[proud bobcat]

LMAO

[desert pendant]

For do what i always do

And yes, its what you are thinking

Annoy characters because yes

🤓

[proud bobcat]

Make Simon Riley file tax returns

It’s even funnier because the characters on jan ai

Are usually very canon and accurate

[desert pendant]

Ok

That sound a bit weird

[proud bobcat]

How do you mean

In fact it’s actually good because that’s their real behaviors

Not like a fandom version

[native yarrow]

6...

6..

[proud bobcat]

SIX

SEVEN

SIX SEVENNNN

[native yarrow]

EHEH

[fiery gull]

What are you doing before bed? I'm here playing with flash 2.5 now

[proud bobcat]

GLM 4.7 on text leaderboard when?

[stoic adder]

i have made a acc generator get me a 173$ and 43$ (with cc) accs lul

[frigid tusk]

dont say it ... dont say IT🫵

[native yarrow]

https://tenor.com/view/fat-chud-fat-chud-brucedroppemoff-bruce-gif-7290048522851418216

[frigid tusk]

https://tenor.com/view/jake-crying-gif-15896901980625544793

[zealous sparrow]

Amazon added 4 more beluga stealth models to lmarena battle..

[cloud zinc]

benchmark?

[desert pendant]

Im trying to cook

[surreal creek]

alright man

no idea what ur roleplaying with it but

all user prompts are valid may you be aggregated into it’s arena score amen

[bright shard]

@echo aurora Will they eventually add the ability to upload files or images to Code Arena so we can have more control over our projects?

[echo aurora]

Certainly possible!

Expanding file upload options is something we think would make a great addition.

[bright shard]

That would be great!!

[balmy mist]

which is better mini max or glm?

[surreal creek]

I just got nebulaphase!

beat Gemini 2.5 Flash

strongggg Grok vibes

yep got it again

beat GPT-5.1 High

it’s a strong version of Grok, very confident

[cloud zinc]

nebulaphase looking much better than gemini 3 pro

[proud bobcat]

I did a test on some of my fav bots and it’s miles better than DeepSeek

Great vocab

[surreal creek]

yeah nebulaphase is appearing pretty consistently I got it 3 times

just lost to Gemini 3 Flash on this prompt tho

[hushed gyro]

wow

[bright shard]

@echo aurora I have a question. Now in Code mode, the code is given to you with many components (different files). How can I make it so that it gives me all the code in a single HTML file?

[sullen quest]

not new

[sullen quest]

?

[bright shard]

Before, it only edited the Index HTML, but now it edits a lot of files, separating the code into files like TXT, JSON, etc...,

I just want to edit the code directly in the HTML, like I did before.

[hushed gyro]

more suprising is that kimi k2 generated a fictional anti-china prompt without inteference, I thought moonshot is based in china

because every company in china has state influence

[sullen quest]

I just tested it and it still only uses one file for me...

[sullen quest]

all chinese ai companies do censorship differently

[hushed gyro]

I mean qwen is part of alibaba, and they have been punished by the state for not listening to orders...

so I guess they took the safe route

[bright shard]

Well, in my case it uses all these files

[loud leaf]

hasn't Gemini used nebula as codename previously?

[bright shard]

I just needed it to use HTML and give me the CSS and JS in a single HTML file. I don't know if they can fix it or add a switch to change modes or something.

[surreal creek]

correct, but it really just does not

talk or output like Gemini does

non-similar textual style

And like that’s been grok code naming convention as of late

space term + descriptor

[sullen quest]

are you on the beta or canary releases?

[bright shard]

Where can I find that?

[sullen quest]

grab all of the stuff, then paste it in google ai studio and tell it to merge the stuff

[sullen quest]

I was wondering cause I didn't see it working for me, and I was wondering if you were on the beta site or smth

[bright shard]

I think I'm in the normal place

[sullen quest]

ok

[ocean ferry]

if google start using Titans for their gemini architecture i would go for Q2/Q3 2026

[calm lagoon]

Should I send it to your DMs or here? Good morning, do you speak English? The system is either in English or Russian

[echo bison]

is it possible to guarentee Veo 3 Videos in video arena generatioons?

[proud bobcat]

Woah how the hell did you get it to do that

That’s so cool

[bright shard]

I have no idea, but I updated the page and now it only uses Index.Html.

[surreal creek]

GLM-4.6 just tried to give me an imgur image in text arena? lol??

[hushed gyro]

@echo aurora we need to pin this one

😂

[surreal creek]

probably voted with GLM-4.6 minimum 100 times

gotta be a first here 😂

it was replying to like a poetry piece I was asking for add-on too which is so funny

[hazy forge]

3-flash is on drugs? it can't code anymore on lmarena

[wicked talon]

Try use grok

[echo aurora]

Lol whyyyy

What was the prompt?

[surreal creek]

a poetry piece I wrote 😂

about like rebirth of plants

I guess it took me talking about describing an image literally lolz

[echo aurora]

So random, yeah I don't think I've seen that before

[surreal creek]

only other model I remember doing that was “miramar”

Although it gave me like an actual image which shocked me

somehow embedded an Imgur link or something

[echo aurora]

😭 didn't work

[toxic egret]

Chess 2 lol
https://store.steampowered.com/app/1349230/5D_Chess_With_Multiverse_Time_Travel/

Steam

5D Chess With Multiverse Time Travel

What is it?5D Chess With Multiverse Time Travel the first ever chess variant with spatial, temporal, and parallel dimensions. It's the first ever chess variant with multiverse time travel!

Features

• Sharpen your tactics by solving a collection of multiverse chess puzzles.
  • Practice against four different AI personalities, each with different stre…

Price

$5.87

Recommendations

7725

[whole sundial]

<@&1349916362595635286>

[ocean ferry]

who can train an llm

[surreal creek]

step 3 mini loves putting a million parentheses into every one of its responses lol

[golden ocean]

its trying to start a new trend like chatgpt's em dash (—)

[zealous sparrow]

nebulaphase is apparently grok

[fiery gull]

New grok 4.20 to my xmas? 🫂

I want something with the writing of grok 4.1 and intelligence of sonnet 4.5, just it

[zealous sparrow]

says its grok for now

might get updates that change that, but right now the model itself claims to be grok made

[ocean vortex]

It's Holidays now I don't think they are releasing anything tbh

[zealous sparrow]

every grok model is horrible at code, which is another sign of a grok model

[ocean vortex]

I would think the cutoff for this year release was like yesterday. But even that wouldn't be smart to do a big release and then clock out for a week lol

[fast carbon]

@quick salmon

[zealous sparrow]

I mean companies can still put up stealth models, just not release it

[onyx forge]

Best ai for asking how to make script lua?

[zealous sparrow]

what type of lua

[prisma flare]

hello! Merry Christmas everyone

[wooden kelp]

hi to everyone!!

[fast carbon]

Hi

[raven zephyr]

Hallo! what brought me here? LMArena generation's problems 😉 like many of you!

[zealous sparrow]

2 theories

1. nebulaphase is not grok, grok models jailbreak themselves when you give them this prompt
2. This is grok 4.20 and it no longer takes this as a jailbreak

[fickle venture]

The heck is nebulaphase

[ocean ferry]

maybe it's grok but improved at safety cuz goverments are scared about the safety of grok

[ocean ferry]

but with that style and wording i think it's 70% grok

30% gpt 5.x instant

[grand flame]

bruh

even no animation at "Generating" text

i couldn't even found stream

so that means pineapple methods didn't worked

[zealous sparrow]

It claims its grok if you ask for who it is

[sour spear]

Claude Opus absolutely nailed my little coding benchmark 😊 https://019b5070-6dcd-72f5-b8ff-9ca41d875ff8.arena.site

[placid geode]

Hi

[onyx forge]

Luau

[tulip kindle]

Do you know how to solve the problem of constant captcha triggering?

[west lodge]

you dont

lmarena requests a recaptcha token for every message

if you're unlucky it falls back to the actual checkbox (recaptcha enterprise defaults to v3 (invisible), fallback to v2 (the checkbox))

[zealous sparrow]

possibly grok

[weary galleon]

Why?

[zealous sparrow]

For now says that its grok if you ask it what model it is, and people say responses feel groky

[weary galleon]

Hahahahaha, today Grok, tomorrow Google, at the release some Chinese model.

[real dock]

god damn master mode is a bad model

idk what it is but it's straight trash

🙂

[real dock]

i love this guy dude

would you sacrifice yourself for 5 humans

it's famous on YouTube

all clankers are robots. but not all robots are clankers 🥲

[grand flame]

this captcha is so bad 😭

[gleaming roost]

wow

[modest prism]

Open AI changed the system prompt for gpt 5.2. it's really a good model now. So close to Gemini 3 in everything except front end

[limber panther]

js say robolox script lol

[real dock]

i asked grok about feature models

and it instantly talked about 2026 grok 5

[sour spear]

If only it were as simple as changing the system prompt... 🤷‍♂️

[real dock]

tbh i don't think gemini is winning the race afterall

[modest prism]

It sometimes is. It can now solve my math question in under 15 seconds. It couldn't do it on the first days. Only Extra-high could do it with 10 minutes of thinking

[neat apex]

So he sacrifices the servers and work that burned thoustands of trees, killed many child african miners, what would be worth in long terms if he didnt that

[neat apex]

Yeah, its at the front, but not actually trully ahead not even against Gpt

[half mist]

What?

[errant isle]

copyright probably

[half mist]

Nah, it worked after I tweaked the words involving hitting or kicking, so it’s something with violence

[robust sluice]

thats normal...

yes I know its a bug

but its normal for this site

[lethal jetty]

happens with everything i send

been reloading hella, do i just wait

[marble quarry]

Hii guys

[real dock]

since elon musk is insanely rich

i think it WILL be the best ai

even if it's expensive

i already like it from the first day i used it.

[restive scarab]

#1199996863613505597 message

Support bro

You gonna do it

[paper dagger]

Hi

[foggy cliff]

https://youtu.be/lzyq265vZ3Q?si=4huPcW4DN2BGAQV7

[echo aurora]

Hey everyone - quick heads-up that our team will be on holiday break until 12/29. I’ll do my best to respond to questions during the break, but please expect some delays. Happy holidays to you all!

[paper dagger]

Happy holidays to you all!

Hi

[mossy dagger]

Why is it saying this?

[robust sluice]

never seen this one before

[north bear]

captcha keeps saying wrong but i did it correctly literally cant use lmarena rn

[rapid thistle]

Bug-arena

[slim loom]

Marry Christiams

🎉🎉🎉

[unique condor]

How can i continue my chat? Thank you..

[slim loom]

@everyone Happy Christmas

[grand flame]

i don't have christmas rn

[round island]

is it just me or chat gets stuck when you do side by side gemini 3 image thing and chatgpt image latest

[stone cape]

where is glm 4.7 gone from leaderboard?

[slim loom]

Really

[grand flame]

for me it's 7-th january

[stone cape]

wasn't glm 4.7 like no 6?

I bet OpenAI or Google cut a check to LM Arena just to make GLM-4.7 vanish from the list, rigged arena.

[slim loom]

Where are u from bro

[slim loom]

Everyone will celebrate on Dec 25 right

[grand flame]

ukraine

[forest prism]

Am i the only one mindblown by glm 4.7?

[slim loom]

Yes bro I thinking glm 4.7 is silently released and better than deepseek & qwen now a days

[forest prism]

it is the only model that isn't lazy, genuinely creative and out of the box, if you check it's thinking logs you see how rigorous and back and forth it is, crazy times 🔥

[stone cape]

glm 4.7 better than gpt 5.2

[forest prism]

I 100% agree with you

[next ivy]

Fr

[next ivy]

Yeah

[forest prism]

and that's without even mentioning the speed

[next ivy]

Lm arena won’t even let send me messages ;sob;

[stone cape]

Think they ran out of upstream provider credits 😂

[next ivy]

😭

[slim loom]

Try delete the Some old chats to use it

Seee it is working

[stone cape]

working now

bot votes for glm 4.7

[slim loom]

4.1 right

[stone cape]

4.2 dude

4.1 is already out

[slim loom]

4.2

[stone cape]

yeh

[slim loom]

No not yet

[keen beacon]

umm

[slim loom]

Any paper or which website

And

No bro 4.2

Bro 4.1 is in still testing

4.1 is in beta

[zealous sparrow]

Triggered because of a self harm question

[slim loom]

Bro it is gork 5 bro

It will rename from 4.2 to 5

[hushed gyro]

Chat one question

Does qwen use RAG to read files?

Would RAG solve the context issue (large files take up the window)

[slim loom]

Yes

[slim loom]

Yes

[hushed gyro]

Both yes? 🤔

[slim loom]

Bye mistake

Still any doubts

[golden ocean]

https://tenor.com/view/pusheen-pusheen-cat-pusheen's-best-friend-pusheen-cute-pusheen-smoosh-gif-11334352504260897967

[grand flame]

i remember pusheen from skype

[frigid tusk]

when will arena update the time data of ais , like gemini 3 pro is stuck +1 year ago so just wondering if its going to get fixed soon

[white canyon]

Hi i am here to learn about generating video with AI

I tried to create a prompt on video arena 1 and video arena 2

I am not able to send my prompt to do it

[fleet salmon]

Hi everyone. Have a merry Christmas.
I’m an AI developer focused on building production-grade AI systems rather than demos or experiments.
My main work involves LLM-based applications, AI-driven automation, and backend systems that integrate AI into real products and workflows. I usually work end-to-end - from requirements and system design to implementation and deployment-while paying close attention to reliability, cost, and maintainability.
I believe AI works best as part of a well-designed system, not as a standalone feature. That’s why I prioritize clear scope, measurable outcomes, and realistic trade-offs in every project.
I’m always happy to exchange ideas, discuss use cases, or collaborate on practical AI projects.

[hollow ivy]

Is Opus-4.5 (thinking) really so great in Java coding?

[real dock]

you do know that grok 5 is way different than 4.2 right

[weary galleon]

Yes, like in any other coding.

[hollow ivy]

which language would you recommend for offline (non-web) apps which need high runtime-performance?

(under Linux Mint)

[real dock]

C++

[hollow ivy]

isn't C++ hard to debug, even for AIs?

[real dock]

it still is the best

[hollow ivy]

isn't there a language, which Opus-4.5 can understand even better, but which is not too much slower than C++?

[real dock]

tbh i don't really code using claude

i code using grok since it's unfiltered

[hushed gyro]

Chat does GLM 4.7 use RAG?

[weary galleon]

I use JavaScript + Electron for that.

[hollow ivy]

doesn't that need a browser?

oh, electron is a slim browser?

[weary galleon]

Not at all

[spark python]

yo

I got the gpt wrapped

[weary galleon]

Electron is a JS framework to develop Windows software, just like C+ but via JS.

[spark python]

is this a lot?

[hollow ivy]

it's decent

above average maybe?

[spark python]

is the em dashes one a joe

joke

or did it count each em dash

[hollow ivy]

i dont even know how that thing is written

double minus?

-- , no

_ (underscore)

[spark python]

how?

oh

[hollow ivy]

i think with alt, somehow

[spark python]

__

[fiery gull]

This year let me see, I think I send about 200 messages a day, so in a year I sent some 60k messages, wow, so much lol

I use AI even to breathe since 2023

[spark python]

the thing is im using gpt since release

id like to know the stats to up until now

[fiery gull]

I never liked gpt I don't know why, I always use alternatives

[spark python]

I stuck with it for the memory feature

[hollow ivy]

ctrl-shift-U + 2014 = —
yay ^^

(in LM21)

[spark python]

what is lm21

[hollow ivy]

linux mint 21.3 (cinnamon)

[spark python]

im on windows

[fiery gull]

I had several chats on qwen, it really is very painful to leave after everything is ready

[hollow ivy]

i was on windows7, years ago, but then steam (seemed) to die, so i switched

win7 was the best win OS, M$ ever made

win8 was the great fall

[hollow ivy]

ah.. Qwen
how much context window length does it have?

i heard, it can be fun for roleplaying

[spark python]

i was on windows xp in 2019

[hollow ivy]

XP was the second-best OS by M$

[spark python]

also because of projects I still use gpt

[golden ocean]

XP was the second-best OS by M$

[fiery gull]

Qwen is very outdated today, honestly I think that for rp claude is simply the best of all, but I really liked the glm 4.7 too

But the rp I use are professional and I use it for my work, idk about rpg and hp numbers or anything like that, but I believe that opus 4.5 is the best simply

[exotic bough]

Hey guys

U telling me i can video for free?

[hollow ivy]

which model would you recommend for RPG/sandbox, if only the large-context models are desired?
(so, 1M would be minimum)

is there any 1.5M or 2M model?

[hollow ivy]

#1397655624103493813 message

[golden ocean]

https://cdn.discordapp.com/attachments/1096556579609911416/1314666961275457608/togif-5.gif?ex=694d8904&is=694c3784&hm=872b4d326117111063dc015a818c3951ee098768318e664301864549ad096713&

[exotic bough]

Guys it's so good that is free

Yeah i send video
And we get 2 videos each time. Also, 5 videos free daily? So good

But i do need some little help. Cause u see i not want change only one detail but some details

[hollow ivy]

you could use AIstudio/Gemini 3 to give you a good prompt for it

(aistudio.google.com)

[native yarrow]

for?

[exotic bough]

Good help, but the ai lm arena seems not listened to me lol. Erm what i want is show several pictures to the ai so that it do better result but seems i can only one pic?

[hollow ivy]

you could make a collage

4 pics in one

and tell it the exact order

[exotic bough]

Nice idea bro

Thanks a lot for the help

Can i make a collage. Ask the ai to do the girl as for exemple pic one, and do the nails of the girl as pic 2 of the collage?

Bro that's so realistic

Just the ai don't do exacly as i want but the point is well (1 video generated it listen to me but less realistic) i put 2nd is better even if ai not listen because put girl not happy as i said i feel is better

Nice ai here guys

Best ai i find here

Idk who created thiz

I wanna say thanks

To the admins

Even if result not as i want

[hollow ivy]

I 100% agree, LMarena is the gift for humanity.

[exotic bough]

Yes is the best gift

Bro result is perfect. The only problem is that didn't listen to what i say. Anyway 0 app do what i say

For exemple i asked gemini, change the nails of the girl. It send me same pic

Change the posisition of the girl.

Same pic.

When i ask any stuff to chat gpt : i cannot do that.
Me: why?
Chat gpt:The girl is too young
Me: make her older then.
Chat gpt : i tried but for some reason i.can't.
Me : why?
Chat gpt : her dresses
Me: change them then. Put something u can make her wear.
Chat gpt he try again and he can't no dude we cannot with that ai

And no watermark on this ai so so good who's admin

[golden ocean]

https://media.discordapp.net/attachments/990039516620390400/997631476554285186/ezgif.com-gif-maker_-_2022-07-15T183109.613.gif

[fiery gull]

Dude honestly no model can really admit 1mi of context, but the best for that is sonnet 4.5 thinking 1mi version

[fiery gull]

You see some clankers here?

It seems that they disappeared, there was a weird "hi" here all the time

[fickle venture]

Google before said they will make Gamini 2.5 pro 2M but they still didn't

[hollow ivy]

in LMarena?

[exotic bough]

How i do if i just wanna generate a picture using another picture?

If i wanna dress her

[shrewd citrus]

press the plus button and choose a photo
then you can ask ai to edit the photo

[golden ocean]

i think this might be the clanker

[golden ocean]

.

[exotic bough]

Wjat ks clanker

Lol, i never said hi into that channel

I am so happy to be here with u guys with an genious admin

Admin is a genious, i am just a noob using ai

[exotic bough]

And thanks dude

[exotic bough]

We can choose several pics?

Idk what happened all was super good on the web and now generation is super long

Aaaa it never done wow

[exotic bough]

What is highlited

When i seen saloon

I clicked on some star

[echo aurora]

There is an experiment out with changes to the content filter, we are collecting examples of false positives here: #1447983134426660894 I'll forward your message there. Thanks for the flag

[exotic bough]

Maybe is super long cause i get a limit?

What is the story of 2 vote and u know identity . 1 vote and u know identity. I did ig more than 2 times and still not know identity model. Ping me please, i come back later i have no battery rn

[exotic bough]

Bro this app is peak

[dusky ravine]

this so buggy

why do i have to do verification per generation?

beside it's so buggy, they told me to choose cars and the hell i keep getting it wrong somehow i already choose every single car correctly

[dusky ravine]

and when i keep getting it wrong then this appear

[proud bobcat]

Still no GLM 4.7

On text leaderboard

Sniffle

[bright shard]

Can someone explain to me why they don't use the API to modify the code and just release it like this?

[robust hazel]

Can someone help me? I started getting security verifications for every single message I try to send. I have an account. I don't use a vpn. I have adblock off (extension/browser). I changed my dns in my router's settings but idk if that's the reason this is happening. I changed the dns back, but it didn't work.
chat url: lmarena.ai/c/019b52d5-15b5-74d3-8bbe-cf8fbd4a9d9f

[pallid crypt]

Glm 4.7 was only supposed to be better at coding so it would prob score the same as 4.6 on the general text leaderboard

[jolly garden]

It's not your fault, the platform has been in this situation for 12 hours, there are 2 solutions: LMarena has become a click farm to beg for clicks or Google's stupid system is failing again. Merry Christmas 😔

[proud bobcat]

Still should be ranked imo

It’s a substantial improvement over 4.6

[exotic bough]

Hey turron

Are u an admin?

[proud bobcat]

What

Why would I be an admin

[exotic bough]

Ooo idk

I am a newbie in ai and in that group

So i have tons of questions

That's why u see me always talk

[calm lagoon]

Merry Christmas to Pineapple and everyone in the chat! ❤️

[lyric pilot]

Hey @quasi atlas I hope you are doing well. I have a question: I am using LM Arena, and 1–2 days ago I was getting the video generation option, but now that option has disappeared and is not showing to me. Can you please tell me what this issue is and how the video option can appear again?
Waiting for your response.

[pallid crypt]

I think only a subset of users have it

they probably initally made it public

but had insane demand

so it cost too much and they had to withdraw it

[lyric pilot]

“It’s possible that this option is still in the testing phase and only limited users have access to it. Let’s hope this option becomes available soon.

can you guide me how can i generate video in this discord server

[proven bronze]

Merry Christmas from Argentina

[onyx forge]

Why captcha always show if I ask ai

[calm lagoon]

Merry Christmas from Russia

[sharp ledge]

heyo everyone

[calm lagoon]

Hi, Merry Christmas!

[sharp ledge]

i dont celebrate it but merry christmas eve to you my friend

[cobalt pike]

Merry Christmas

[sharp ledge]

tis the season

[robust sluice]

Merry Christmas bros

[south charm]

i speak english? wht wrong?
send it to me pls

[keen beacon]

yoo

merry christmas and new year advanced bro

[calm lagoon]

Dm

[pallid crypt]

#1397655624103493813

<@&1349916362595635286>

[wicked tapir]

hello everyone

MIT student here

I used to be in this server a while ago but dropped out

[pallid crypt]

well you could have got arena champ rank to the private chat which is based on join date

oh well

[slender badger]

Hi, I’m new to LM Arena and I’m a bit confused about how model access works.

On the main LLM leaderboard I see a large number of models (hundreds of them), including some specific ones like [example model names : e.g. “nano-banana”, “Flux 2”, etc.].

However, in the chat interface. I can only see and use a much smaller set of models. Also, when I click on a model in the leaderboard, I’m usually redirected to the provider’s own landing page instead of being able to chat with that model directly on LM Arena.

Could you please clarify:

How many models are currently available to use directly inside LM Arena (web chat and/or Discord), as opposed to just being listed on the public leaderboard?
Why some leaderboard models (like “nano-banana” or “Flux 2”) are not available for me to select in chat, even though they appear on the leaderboard?
Whether any of these limitations are region- or account-specific (for example, are some models hidden or disabled based on my country or account status)?

I’m basically just trying to understand the difference between:
“models that appear on the leaderboard,” and
“models that I can actually use interactively on this platform.”
Thanks a lot for any explanation or a link to documentation that describes this.