general

[keen beacon]

Like you can’t mess up this bad and not be held accountable I don’t care even if it was on accident and you did your best

2.9 billion records, dude

If this was the Soviet Union, they take you out back and shoot you 🤣

[unkempt mantle]

ever heard about wazirx?

[keen beacon]

No

What I’m saying is you never see any of the people ever get held accountable to her in charge of like any of this stuff, dude?

I mean, maybe one will resign

[raven stone]

how to remove it?????

[keen beacon]

I mean, I can understand 50,000

But 2.9 billion

[unkempt mantle]

check this post

[keen beacon]

Ohh

Ya the kid

Yes, yes yes yes yes

But even in this case, bro

How did you get caught again? You have to refresh my memory.

[raven stone]

What are you talking about?

[keen beacon]

I was talking to somebody else. I’m sorry I was gonna say for you. You just check that little box where it says you’re not a robot.

And it should go away

I mean once you check that box, it gets removed right?

Unless you’re a robot

[raven stone]

I'm putting it in, but nothing changes. I have to verify again with every message, it wasn't like this before.

[keen beacon]

Are you using Google Chrome by any chance?

Any vpns? Proxies ?

[raven stone]

Zen Browser

[raven stone]

nope

[keen beacon]

OK so we could eliminate the VPNs and proxies off the list I’ve never heard of Zen browser

But try clearing your data of your browser

[unkempt mantle]

chetan bhai use chrome maybe that would help

[keen beacon]

If that doesn’t work, try a different browser. I’ve had a person tell me today that they switched to. DuckDuckGo browser and they stopped having the issues I assume and this is the same browser. I use with no issues whatsoever.

Of course, the only other options do you have available? I’m just telling you the ones that worked for me.

And if that doesn’t work, the last thing you could try to do is reset your ISP modem

https://youtu.be/D8GOeCFFby4?si=cH0ca0LBA-pfaCyC

[grim grail]

How are you?
Is there anyone looking for dev?

[unkempt mantle]

can someone suggest an unfiltered chatbot with no sensorship other than grok?

[keen beacon]

What are you trying to generate?

Censorship is not what you think dude it’s all about context

These models could almost generate anything all of them

Except for video models are very hard

[unkempt mantle]

a friend of mine just wanted to have a chatbot with no censorship like maybe an illegal business thing, or evading tax etc. who knows

[keen beacon]

🤣🤣🤣🤣🤣

Bro, you could do that legally

If you use the grayswan arena

It’s designed for this. It’s in a sandbox environment and you help make the models safer.

[unkempt mantle]

i mean i don't know in why he wants an uncensored chatbot

[keen beacon]

It doesn’t matter people should have the freedom to express themselves and communicate themselves however, they feel privately as long as they’re not hurting anybody or themselves

And if they get caught, it’s on them

Of course, we should never facilitate or encourage that kind of behavior, but I do believe in people having the right to explore privately

[raven stone]

again..

[keen beacon]

Is this with the new browser?

Or is it with the same browser?

[raven stone]

yap

[keen beacon]

OK, so what did you just try? You just try clearing your browsing history and data

Cookies, everything?

[raven stone]

new browser bro

[keen beacon]

Try DuckDuckGo browser

[unkempt mantle]

bro flush dns and maybe then try

[keen beacon]

https://duckduckgo.com/app/devices

[unkempt mantle]

true

[keen beacon]

Yes

Flushing dns should also help

Damn bro, you know your stuff

[raven stone]

How can I contact the administrators about this problem?

[keen beacon]

I’ve been trying to figure out for a long time. What causes this because this is a consistent issue. People have all day long, dude you wouldn’t believe.

[keen beacon]

Dude, you gotta at least do two more things to try to diagnose it before you give in

Try to reset your modem, flushing dns or try the DuckDuckGo browser to see if that doesn’t work

If it doesn’t, then we’ve exhausted all options

[raven stone]

ahhh, I think the problem will remain

[keen beacon]

Well, I’ll tell you this dude. What is the admin gonna do honestly

They can’t go and see what’s going on in your computer there’s hundreds of thousands of users, bro

It would be hard to identify you specifically in your issue, even if it was on their end

[unkempt mantle]

Chetan bhai what are you using ? personal internet connection or hostel's wifi?

[keen beacon]

This also happens if you use it really fast

Like if you’re starting conversations really fast or going through your previous conversations quickly

Honestly, I don’t think this problem will ever go away in this discord

You would be surprised, dude every single day all throughout the day day after day week after week it’s the same reoccurring problem😅😅😅

It should get its own Hall of Fame

[unkempt mantle]

i only hope to see the video generation feature being permanently there

[keen beacon]

Well, if I’m being honest, I remember I forget when this was maybe a little while back, but I remember when it was happening to me

And I was thinking the same thing I’m like oh my gosh, this is unusable. This is unbearable.

So I understand what they’re going through

[keen beacon]

It should, bro the thing is like everything else with AI a lot more companies are sprouting up, so the value of the premium service naturally goes down

It’s hard to put the genie back in the bottle once everybody started offering free video

Not just with arena but just in general

[keen beacon]

Trust me, bro I know it’s like being in your shoes too excitement

The possibilities, the imagination

That’s actually the rich part

[unkempt mantle]

yeah feels like a portal to a new dimension XD

[keen beacon]

Well, what I’m saying is this dude it’s that feeling that you’re feeling right now is what’s actually worth cherishing

Because once you get in it and your hands get dirty and you start putting in the work, you side of that and then it just becomes work haha

[raven stone]

don't working

[keen beacon]

Ok

Well so we diagnose three things we got a new browser two of them

He cleared his history and browsing data

[unkempt mantle]

will be active here on this discord from january when i'll be actually making mockups for my product both images and videos for reviews and changes needed.

nice talking to you @keen beacon will catch up later its almost 4am here in india

[keen beacon]

Yeah, I’m on and off here. I take a few days off and then I’m on for a couple days and then head out too so I’ll see you when I see you bro. Yeah I was a pleasure man. Just keep your head high.

You got a good head on your shoulders man do the right thing and the right things will fall into place my man I guarantee that

[unkempt mantle]

i feel like i'm at the lowest right now but at the same time it would be fun to disappoint people by proving them wrong XD

[keen beacon]

Well, dude, the beauty about being at the lowest

[unkempt mantle]

you too mate have a great festive season ahead and a happy new year

[keen beacon]

Is he got really great view of the height to look forward to

You don’t gotta prove anybody wrong ever you just gotta prove yourself right

And for the right reason

[keen beacon]

Why is it in Russian?

[unkempt mantle]

i do have a climber mindset, climb to the top, enjoy the view, the solitude

[unkempt mantle]

well thats true but have to prove someone passively

[keen beacon]

Well, if your climber, what’s a very important thing to do

[unkempt mantle]

thats what we do buddy stay at the back accompanying people who find it difficult to climb.

[keen beacon]

I’m not telling you nothing you don’t know, bro

I’m just reinforcing these concepts that they are universally shared and respected all over the world

And I think it’s important that we encourage it when we see it and support it when you can part of my belief in humanity over machinery 🤣

[unkempt mantle]

humanity over anything buddy, always have that mindset

[keen beacon]

Yeah, I’m a take a little break too. I’ve been chatting and blowing up this whole discord all morning realizing it so would take a little break.

[unkempt mantle]

sure have a good day

[keen beacon]

🖐️

[thorny schooner]

Well that was heartwarming to watch certain rare in these days

[raven stone]

yup

[keen beacon]

Oh my bad before I go I totally forgot

[unkempt mantle]

damn looks better

[keen beacon]

No, you’re aesthetic is way cooler

[unkempt mantle]

but the thing is the tree i want to use is not the actual tree that i should be using for my roduct

[keen beacon]

I was just curious what the prompt was for yours

I want to compare and see how far off I was on mine

[unkempt mantle]

lol its funny and hard to explain but wait i'll try

[thorny schooner]

Hope I'm not disrupting your conversation but sorry I just like these two very much your guys looks cool as well but also I just realized the second one gave me such Warhammer Vibes but also frostpunk vibes

[unkempt mantle]

so this was the art i was inspired from, its from an album(lunchbreak) cover of a hiphop duo called seedhe maut

create an image of mountains from garhwal region of india with ________ flowers and red lily spiders in art style that of an anime like seedhe maut's album cover lunchbreak for a beverage packaging

[keen beacon]

I think I got you big butt one second

[keen beacon]

Its dope

I think this version may pass Sora

[thorny schooner]

cool

[unkempt mantle]

one of my friend prompts really well and the images produced are such high quality, artistic and unique

[unkempt mantle]

i think either you should know the visual style name or you should have a reference image

[keen beacon]

It worked

[keen beacon]

Now I can make him a character and sore and you could always use them

I mean, it’s not the real thing since a little cartoonish but

[unkempt mantle]

this was my prompt

[keen beacon]

Yeah, one second about to check it out and generate I wanna see

Man, it’s crazy

[unkempt mantle]

i actually had to fix the errors once in a while

[ember sierra]

👋

[keen beacon]

Brah

I use this prompt three times and it got rejected

Even with this version

But this version of the prompt for some reason worked

Odd

[thorny schooner]

Still annoying I can't do videos since I have quotes for him

[night spruce]

👋

[keen beacon]

What do you want the characters name?

And what do you want his description to be?

[thorny schooner]

"Democracy is not sacred. It is delicate. It can die in the hands of those who claim to love it most"
"If you tell the people only what comforts them, you are not leading — you are sedating"
"I do not distrust democracy. I distrust what we have become in its absence — and what we might become in its premature return"
"I have seen chaos masquerade as freedom and tyranny cloaked in the guise of order. We will be neither."
"We will not raise a new generation to worship people like me. We will raise one who forgets the need."
"No system fails in isolation. It fails because its people forget how it was built, and why"

These are the quotes I had for him

[keen beacon]

Once I create character, you’ll be able to generate him by yourself

[thorny schooner]

Give me a minute I'll send you the document

[keen beacon]

Carries himself with unwavering resolve and a quiet confidence, leading with a commanding gaze and measured, authoritative speech. His posture is always impeccable, his uniform pristine, and every movement radiates disciplined leadership in a world of high-stakes intrigue.

Ok

It has to be like under that much characters and words they can’t be too many

[bronze sierra]

👋

[keen beacon]

Now I’m just gonna name him Bigfootmander the handle you could use

[thorny schooner]

🥲

[thorny schooner]

Marshal Saul Virek

[keen beacon]

Ok

[thorny schooner]

I think I will just let you use that example you gave me for the description because welp the description I have for him is way too big

[thorny schooner]

What do you think of these quotes though

[keen beacon]

@bigfootmander

So you could generate him in sora now

And make him do whatever I’m gonna test it right now

Yeah, that’s a fantastic quote. That’s the video I’m generating with him saying that “

These filters are so sensitive

[nova sluice]

Does anyone here have a topaz ai subscription?

[keen beacon]

I wish

[thorny schooner]

I'm sorry this was just so random when I was looking for it lmao

[keen beacon]

It’s so expensive

[thorny schooner]

☠️

[nova sluice]

Yeah

I have a 19 second video, i want to upscale it

[keen beacon]

It’s like $1.80 per second or something like that some crazy ridiculous price

On Val

[nova sluice]

🥀

[thorny schooner]

What do you guys think I just it made a map ( frame when curious this is a map of the same country that the Marshall character we are talking about takes place in)

[keen beacon]

What is that supposed to be Germany Austria?

But it’s got Roman inspired names Latin

[thorny schooner]

No it's from a custom country

Okay fu speech to text

[keen beacon]

OK, well the first video got blocked

Oh, I totally forgot. I had freezia as a character save.

[thorny schooner]

Basically it's a custom country I made with its own history intertwine with real world history from The Colony era to the different world wars and more

[keen beacon]

Hahah

Haha ur boy still has some skin in the game 😎

[thorny schooner]

Smashed the reason I have whole documents for it I actually have multiple documents for different variates of it

Have no idea why lol

[keen beacon]

All right, so here’s your characters cameo

That was just that generic test

[thorny schooner]

coolty

[keen beacon]

Now let’s see if I can get them to say that “

[keen beacon]

Oh, you got your own game made huh or whatever the set up you got here is

A whole immersive thing you got here

Cool man

[thorny schooner]

ty you know the whole country thing has been kind of my passion project that I could never really get iron down especially since I was using the AI to help refine finder points of it

[keen beacon]

All right, dude I’ll make one more but I’ll make it 15 seconds long so hopefully that fits in your speech

And then you’re gonna have to do it on your own if you have the ability and can use sora

[thorny schooner]

I'm not going to lie I was not going to expect it to hit so hard in a speech

[keen beacon]

Hey, you never know it just might you could never count AI out

[thorny schooner]

Can't find him

[keen beacon]

@bigfootmander

[keen beacon]

No, you type that in when you’re prompting

Otherwise, you’re gonna have to look up my user name

@stopslop

[thorny schooner]

And that's an immediate download for me

[keen beacon]

https://sora.chatgpt.com/p/s_69487f81dbfc81919fc54e5aa1966f8a

https://watermarkremoversora.com/

You could use this website to download the video without a watermark

[thorny schooner]

ty don't know why you don't appear when I search

[cloud zinc]

they released new sora 2?

[keen beacon]

It’s probably the update

The thing you showed me the other day

[cloud zinc]

i think its the one with less flickering

[keen beacon]

What flickering?

I’ve never noticed I haven’t been paying attention

But that’s all way from November 25 that was a month ago

[cloud zinc]

that update

[thorny schooner]

https://sora.chatgpt.com/p/s_69488237f70c8191aaee45fec882b125?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM._kILt-UlTnTb what do you think of the small changes I added

[keen beacon]

Yeah, it’s nice. You can make him do anything, dude.

All you Gotta do is just write a prompt and just add his handle and you can make him do all kinds of cool things and scenes

I’m surprised it worked that’s awesome because even the text version didn’t work, but for some reason that image and the prompt past makes no sense

[thorny schooner]

Also I can do videos now apparently for some reason it had me logged in to the older model and not the newer one and apparently that's actually one of the reasons you did not pop up when I search you

[keen beacon]

It’s all good things have their own way of working out

[lost elbow]

Hello,
Please consider adding a feature to lmarena.ai that allows users to upload multiple images simultaneously in a single request, enabling various AI models (such as Gemini, ChatGPT, and others) to properly understand, analyze, and combine these images.
This feature could include capabilities such as:
Combining two or more images together
Adding or moving subjects between images
Intelligent editing based on multiple input images
This would be similar to features in some advanced AI systems that allow simultaneous understanding of multiple images.
Adding this functionality could significantly enhance the user experience and provide more professional and versatile applications for your platform.
Please consider adding this feature to lmarena.ai as soon as possible so users can benefit from these advanced capabilities.
Thank you for your attention.

[keen beacon]

It sucks that nobody embraces hurtles and challenges as an opportunity to learn, sharpen your craft

[thorny schooner]

https://sora.chatgpt.com/p/s_69488527e698819198cd9eaac685f0fa?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM.xIfZjUvVGBPI

[keen beacon]

Hey, that’s pretty good dude

Except the thing is it was that the style you were going for that kind of rendered look?

Cause I think there’s ways to make it realistic, but some people prefer that kind of artistic rendering

[thorny schooner]

Honestly I'm fine either and I think for the moment it kind of matched the whole feeling that that video has with the guys talk

Already having another one be remade to see what it will look like in black and white color palette

[keen beacon]

Do you know how to add characters if you wanted this as a character and wanted to reproduce it again there’s a feature on

It’s the ad character feature, and you could save it and make your own character and reuse it like I did with the other character that you showed me earlier with the speech

[thorny schooner]

Yeah I saw the problem is I when it asked me for the image/video upload it takes me to download in my files which for some reason never has the images I save

[keen beacon]

Damn, I hate to admit this, but it has to be said open. AI is extremely restrictive and extremely heavy-handed in their moderation.

And what I mean about bad is compared to everything else that’s out there on the market currently

It really does seem to be very restrictive. It’s like it’s playing a two sided game where it’s trying to please both side sides.

[thorny schooner]

https://sora.chatgpt.com/p/s_6948873c658881919ea306ae4ef02ae9?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM.0nhdIOoCCDwG oh I really like this

[keen beacon]

Dude, this Disney deal makes me wonder

Not only the Disney give them 200 characters but they also invested $1 billion

Odd

[cloud zinc]

whats odd

[keen beacon]

Being fronted 1,000,000,000+ signing that exclusive deal

You would figure it would be the other way around that opening. I would be interested in paying Disney for their characters.

Because apparently sora is costing open ai billions and is not generating any profit?

[thorny schooner]

Now that I think about it why does this feel so We Happy Few coded

[keen beacon]

A single ten-second clip could cost OpenAI roughly $1.30, Cantor Fitzgerald analyst Deepak Mathivanan told Forbes.

[thorny schooner]

https://tenor.com/view/we-happy-few-wakey-wakey-whf-gif-11844014

[keen beacon]

https://recap.aitools.inc/p/openai-s-15m-daily-burn-for-sora

If this is really true, that’s so Ludacris

[cloud zinc]

disney always invest in the company they license to

[keen beacon]

True that never thought about it like that good point

[cloud zinc]

https://www.cnbc.com/2024/02/07/disney-to-take-1point5-billion-stake-in-epic-games-maker-of-fortnite.html

[keen beacon]

Yup it’s official. Forbes reported it. They do burn through 15 million a day on Sora

They’re just ridiculous, dude honestly shut it down

15 million a day, dude

No wonder they put on restrictions to get people to stop generating lol

[thorny schooner]

https://sora.chatgpt.com/p/s_69488962cf848191a2126aecbf22e8cc?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM.axuyK3kYvdUT https://sora.chatgpt.com/p/s_694889c62d1481919d6effed641ba9a0?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM.rJeGGQqQqszH

[cloud zinc]

we dont know the inference cost

[thorny schooner]

I'm just glad I can do some experiment with it since I have a decent amount of original characters saved up

[keen beacon]

Apparently, from what I’m reading, it’s costing them $1.30 per video they make

Am I doing the math correct here?

Because if I am, that would suggest that almost 12,000,000 videos get generated daily in Sora?

[civic violet]

is it working LMArena now?

[keen beacon]

That’s an interesting prompting format. You have there.

[thorny schooner]

https://sora.chatgpt.com/p/s_69488f81b02c8191bca6af030cbb3941?psh=HXVzZXItMnVEaHNCdENFb042dHd0Ukd3SXNuNThM.Z-W_mcJCNXyD was made based on the game we happy few ( if you know that game you know)

[keen beacon]

Hey good for you man. It sounds like and it looks like you’re enjoying yourself dude it’s awesome. I’m digging the energy.

[cloud zinc]

https://www.reddit.com/r/SoraAi/comments/1pskx9c/my_bypass_method_works_with_nearly_every_character/

[keen beacon]

Oh dude, I know it’s possible, but I was just stuck on one character

That’s why my world is gloomy 😭

[proud bobcat]

I’m crine

[keen beacon]

I haven’t really attempted anything else, but it is what it is all good things come to an end

[proud bobcat]

DeepSeek has more traffic share than copilot

[keen beacon]

You should be killing it, bro you already got it down

The thing for me is though bro I have nowhere to share none of my stuff really

[cloud zinc]

thats not me.

[queen veldt]

This is new

[keen beacon]

Damn, that does look nice

[queen veldt]

Idk i just saw a instagram video ab it

[keen beacon]

Man, I wish there was a platform for people to upload AI content

That they generate but like bulk

Without all the hassles of regular social media

[ocean vortex]

Copilot 💀

That’s impressive considering how much they offer for free

[native yarrow]

is this your like website screentime?

[ocean vortex]

Hard to screw it up this bad lol

Ironically, it’s probably a better option than chatgpt for people who do not pay. But MS isn’t exactly the best at marketing, user experience or even their image

[keen beacon]

Well, you know why they Nerf the original copilot I heard

https://www.nbcnews.com/tech/tech-news/taylor-swift-nude-deepfake-ai-photos-images-rcna135913

This was a little while back

[ocean vortex]

The model you get to use there presently for free is near identical to chatgpt paid one

[keen beacon]

https://mspoweruser.com/microsoft-fixes-loopholes-that-created-taylor-swifts-deep-fakes/

Yeah, from what I heard, though it used to be really good back in the day. I’ve never tried it, but I heard people really enjoyed it before they made it really restrictive.

I think Microsoft is still one of the biggest winners in this whole AI debacle

From all the clouding and they get a percentage of open AI revenue

[ocean vortex]

They played their cards decently but knowing their resources I would expect them to have a go at training big models themselves

[keen beacon]

Why why I spent all those billions and millions of dollars training a model when you could benefit from the ecosystem

Without having to take any of the risk

[ocean vortex]

And treating this as a stop gap, instead of relying on OpenAI indefinitely

[keen beacon]

Well, they’re in a really strange contract

They actually have AGI legally defined within this contract

A clause or a contract not sure let me look that up

First, the board of the ChatGPT developer holds the sole authority to declare that AGI has been achieved. If that determination is made, Microsoft would immediately be blocked from receiving any models developed after that point.

Second, the contract defines a benchmark called “sufficient AGI,” tied to projected profitability. If OpenAI demonstrates that a model could generate more than $100 billion in profit, it can withhold access — unless Microsoft disputes the claim and a court decides otherwise.

Third, Microsoft is barred from developing AGI independently or through any partner using OpenAI’s IP. That restriction remains in force for the duration of the agreement, keeping Redmond locked out unless the terms are rewritten.

[keen beacon]

Redmond, WA – Technology giant Microsoft Corporation (NASDAQ: MSFT) has once again demonstrated its formidable market presence, reporting an exceptionally strong fiscal third quarter for 2025. The company announced impressive revenue growth of 13% year-over-year, reaching a staggering $70.1 billion, while its net income surged by 18% to $25.8 billion. These figures not only highlight Microsoft's robust operational health but also significantly surpassed analyst expectations, sending a clear signal of continued dominance in critical technology sectors like cloud computing and artificial intelligence.

The stellar performance underscores Microsoft's successful strategic pivots and investments in high-growth areas. The immediate implications for the market are broadly positive, reinforcing investor confidence in the tech sector's resilience and the transformative power of AI.

Microsoft's cloud and AI offerings, which permeated across all its business segments. The Intelligent Cloud segment, encompassing the powerhouse Azure cloud platform, saw its revenue climb by an impressive 21% to $26.8 billion. Crucially, Azure and other cloud services alone experienced a 33% revenue increase, with AI services contributing a substantial 16 percentage points to that growth

And this is just in the third quarter of 20 25

That’s more revenue from cloud services than open. AI is able to generate yearly revenue.

[half mist]

But like why though? Why do people generate nude photos? Those people are freaks

[keen beacon]

And you wouldn’t believe who is the biggest victim of all this

https://19thnews.org/2025/07/deepfake-ai-kids-schools-laws-policy/

[native yarrow]

you will learn someday my man, believe me it is kinda weird but there's an obvious reason

[drifting crow]

they are obviously studying human anatomy

[thorny schooner]

This is genuine Wilde i hope they have a actual answer for this cuz there's no goddamn reason this should have been blocked like no legit reason cuz for all the other ones I can understand on some level of a sensor making mistakes even if a lot of mistakes but this does not fall under a mistake or if it does i'm sorry what are you using for your sensor

[proud bobcat]

Traffic share of all ai websites

[keen beacon]

False positives. Happens

[thorny schooner]

What fire hydrant

☠️

[proven bronze]

Press skip

Even if you do it right, that damn thing tells you to try again.

[viscid cloak]

Not even the most obnoxious one: those absurdly slow grid pop up after you click the right ones. And once you solve all of them, error and try again🤣

[thorny schooner]

Yes it takes actual damn 9 seconds for those ones i counted with fingers

several times

[proven bronze]

And reCAPTCHA crashed again 🤗

[native yarrow]

you can see my point i keep making how chatgpt is still the most popular

[keen beacon]

https://youtu.be/YFE0zNMqQyg?si=K7p-6hi0dw2jiCnu

[thorny schooner]

?

[keen beacon]

🤣🤣🤣

I guess that old saying is true “if you build it, they shall come”

[keen beacon]

Intel built a new fab ship here in America

We’re gonna be building iron and AI chips here in the states to get the pressure off of Taiwan

Taiwan Semiconductor Manufacturing Co it’s supposed to be giving them some super high in chip manufacturing equipment, state of the art

[keen beacon]

Bro

2nm microchip world's most advanced microchip

[thorny schooner]

ok

[keen beacon]

With a 10 to 15% boost in compute

[keen beacon]

This is SORA

So they’re capable of solving captchas

recaptcha is weak even Sora can solve it lol

[modest wedge]

LMArena down rn for anyone?

[keen beacon]

No, it’s seems to be working fine for me

[keen beacon]

That’s a mass of mid journey still is

[thorny schooner]

🥲

[cloud zinc]

https://www.reddit.com/r/SoraAi/comments/1psqm0j/rick_and_morty_meet_hitler/

[keen beacon]

Damn

Well, to be honest, I’m kind of over it now I’m purging my accounts

[keen beacon]

Next time you get one of those don’t solve it take a screenshot. I’m gonna put it into a Sora and I wanna see if It could solve it.

[thorny schooner]

My man I'm getting them every prompt🥲

[keen beacon]

Send me a pick of one plz

[cloud zinc]

looks like he is using 8 second video.

[thorny schooner]

Does this work

[keen beacon]

Wait bro, actually that didn’t work because they didn’t have any instructions

[keen beacon]

Perfect

Ty

Would this be considered a fail? It was able to solve half.

Oh wow I’m shocked

It can’t seem to solve the bicycle one

[thorny schooner]

Okay wow that was not even close

lol

[keen beacon]

All right, I’m gonna give it one more. Try three for three.

So maybe these things really are effective then

Oh maybe I should make it 15 seconds instead of 10 maybe that’s why

But don’t these things test for how quick it takes you to solve it?

[south charm]

how to fix T_T

[thorny schooner]

I mean you can still use it not login as long as you can accept the terms of service and if you can't usually go into private browser helps

[keen beacon]

Naw bro it completely fails this bike one

[thorny schooner]

Is it wrong for me to feel a bit better of myself after seeing that with the struggle of the verification

[keen beacon]

But hold on look at the 15 second verification one

So I’m gonna count this as a fail

[thorny schooner]

I mean you can call it a half fail

[keen beacon]

If you come up with another tricky one, if one pops up for you, save it I’ll give one more test before I go to sleep tonight

[south charm]

I cant even get into the site

[keen beacon]

Oh that one’s good

[keen beacon]

Give me odds here

What do you think the results are gonna be?

[thorny schooner]

I'm thinking maybe it's going to be one success and 2 fail

Basically maybe half half

[keen beacon]

Fair enough

And we do have interesting results actually

So it obviously identified the motorcycle, but it failed un clicking the boxes 🤣🤣

On this one, it clicked the boxes

And so the last one I’m just gonna say click the boxes around the motorcycle only

[thorny schooner]

Not going to lie that's why does that feel like that would be something you would see from one of cartoon to live movies

lol

[thorny schooner]

,,,,,,,,☠️

[keen beacon]

Complete and utter fail

But this could also be scheming behavior

Where the model behaves or pretends that it’s dumber than it really is

https://www.zdnet.com/article/ai-models-know-when-theyre-being-tested-and-change-their-behavior-research-shows/

https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/

[keen beacon]

Alright bro thank you for your help. I think it’s time for you to hit the bed and get some rest. Have a good rest of your day man. Good night.

[thorny schooner]

Wait what time do you think it is for me

Did you put you instead of me by accident

[torn mantle]

@zealous sparrow whats with minimax new model?

https://x.com/CloudTrader4/status/2002729591451054127

Cloud Trader (@CloudTrader4)

M2.1 is a beast in design area

@MiniMax__AI 🤝 @FactoryAI

X · Dec 21, 2025, 1:15 PM

these designs looks so similar to gemini 3 pro model

https://x.com/CloudTrader4/status/2002438840200659212

Cloud Trader (@CloudTrader4)

M2.1 @MiniMax__AI definitely got much better in design. Works really fast.

Great work @SkylerMiao7

X · Dec 20, 2025, 5:59 PM

[ashen mauve]

ragh I just ran into an infinity generating again after a hella long time

[boreal saddle]

F*ck.

Why is this sh*t popping up again?

I swear to God, if no one does anything about this CAPTCHA nonsense...

[ocean bison]

@boreal saddle Yesterday, I opened a Private Window on Firefox on my PC and it says this exact message but "Powered by Cloudflare"

[ocean ferry]

pro but unnerfed

[spark kayak]

#general
Hi everyone! I’m part of a team that has recently developed a new Image-to-Video (i2v) model, and we’ve been closely following the amazing work being done here at LMArena. We are very interested in getting our model integrated into the Video Arena for community evaluation, but we are a bit unsure about the exact onboarding process. Could someone guide us on the necessary steps? Specifically, we’d like to know the technical requirements for the API (e.g., OpenAI-compatible) and how we can provide the team with the necessary access or API keys for testing. We are ready to support the inference costs and provide any documentation needed. Any advice or pointers would be greatly appreciated!

[thorny schooner]

Why on God's name does the verification love bicycle or motorcycles so much

[snow glade]

😂 😂

[winged bough]

hello

[weary galleon]

Hi

[brittle cradle]

Hey gngs, what's good? ✨

[fading rover]

Hi all is there a way to build app through gemini using nano banana pro model to create content with specific style given include reference ? I tried but gemini says it can only use 2.5 flash image model

[weary galleon]

Claude

[slow zealot]

Good morning guys

This is what keeps popping up each time I want to use the website

Please I need help

Please you all respond to me😏

It's very annoying sh** keeps popping up

[unborn ocean]

Thinking the same thing.
Probably distilling some of g3p in there

Especially with a lot of the Chinese labs building some obscure ‚distillation from experts‘ to build their models

[mossy drum]

See https://discord.com/channels/1340554757349179412/1451574502369656842 and https://discord.com/channels/1340554757349179412/1451363565729747045

[robust sluice]

this isnt about LMA, my ChatGPT app showing that my limit will reset in 11 hours, is this normal ? (I never seen it take this long before)

[lyric marsh]

Why does the AI say to my requests that they contradict the rules, even though it used to fulfill them?

[mossy drum]

See https://discord.com/channels/1340554757349179412/1447983134426660894

[ocean bison]

Basically @mossy drum Run the violating parts through a Base64 encoder, the TOS filters can't even understand it LOL 😂😂😂😂😂😂🤣🤣🤣

But smart models like Gemini 3 Pro CAN understand Base64-encoded text.

[golden ocean]

give me the plant

[queen veldt]

Um what

It can't lmao just tried it

It hallucinates

[ocean bison]

Not for like a whole block of Base64.

[queen veldt]

Then for what?

Base64 is base64

[ocean bison]

I tried it on some messages that I've written that contain "DUMB", "STUPID" and yelling, basically encoded the parts that violated LMArena's TOS and Gemini 3 Pro read perfectly and got past the TOS filters.

[torn mantle]

yea 100%

lol

and they decreased the laziness

[fossil socket]

Any idea why Gemini and GPT can let you attach images on direct message

But the rest can't?

[torn mantle]

which was done on purpose in gemini 3 models

[queen veldt]

Claude opus at least doesn't hallucinate

[ocean bison]

Maybe throw the second prompt in a AI image generator?

Like Nano Banana Pro 🍌

See what output it comes with!

[fossil socket]

@echo aurora

[torn mantle]

other models doesnt support that

or its just lmarena choosing which models has image upload support

ive been trying minimax m2.1 while its impressive that they did improve on frontend but its nowhere near gemini 3 flash

like some people thinks just because they have big platform they can say nonsense

like this guy : https://www.youtube.com/watch?v=hy-5dzyG3Dk

im not gonna talk about how bad his benchmark is, but he just be saying anything atp

[ocean bison]

More like:
"RIP Gemini 2.5 FLASH!"

[torn mantle]

yea

well its good at frontend

but nowhere near gemini 3 flash

and minimax is 100% distilled from gemini 3 pro

[ocean bison]

And why MiniMax M2.1 is on Claude Code IF IT ONLY ALLOWS CLAUDE 4.5 FOR EXAMPLE?

[zealous sparrow]

in my tests, gemini 3 flash made a better weather app than minimax m2.1

[fossil socket]

Feels odd those 2 are the only ones with that feature

[ocean bison]

And o4-mini

[zealous sparrow]

This is just glaze atp

[fossil socket]

Yeah, pretty weird

Anyways there's bigger fish to fry and worry about

Particularly this God Forsaken new filter they have

Always goes haywire and flags things and ignores context

[ocean bison]

I tried it on some messages that I've written that contain "DUMB", "STUPID" and yelling, basically encoded the parts that violated LMArena's TOS to Base64 and Gemini 3 Pro read perfectly and got past the TOS filters.

Maybe you try that?

[weary galleon]

AI-generated video. Voice is so robotic. Name of his YouTube channel is so dumb.

[ocean bison]

And MiniMax M2.1 more like beats Gemini 2.5 Flash

[zealous sparrow]

he put riftrunner real low during gemini 3 pro checkpoint testing, even though it was a goated one

[weary galleon]

I watched just a few seconds of this crap and closed.

[ocean bison]

And the "FoxNews Radio Show" that was generated by NotebookLM 😂😂😂

She didn't say that, the default female voice said that:
"ChatGPT writing esses"

"DIELL-E 3"

[weary galleon]

I didn't. Girl from this show said that in the beginning of the show.

[ocean bison]

LOL 😂😂

😂

[zealous sparrow]

also this guy ranks glm 4.6 above flash 3 and minimax

utter bs

[wicked mason]

VIDEOINU has original sora with copyright bypass

[weary galleon]

FoxNews isn't EpocAI. It's general media, not AI orientated, they don't know exact names of all models.

[ocean bison]

Like the podcast you generated isn't by FoxNews themselves, it's generated with NotebookLM 😂😂

She didn't say that, the default female voice said that:
"ChatGPT writing esses"
"DIELL-E"
LOL 😂😂😂😂😂

[weary galleon]

She said "This show is NOT generated via AI, NotebookLM, or something." What did she mean?

[ocean bison]

This doesn't change if it's generated with AI or not.
While NotebookLM's voices are generative, meaning they sound natural and human-like, they can make errors like "esses" instead of "essays" and "DIELL-E" instead of "DALL-E" and other weird uncanny problems.

Even if the expected script SAYS "essays" and "DALL-E"

[torn mantle]

lmao

glm in general is so bad

[ocean bison]

This can also apply with any other generative text-to-speech (TTS) model like Inworld TTS 1 Max, ElevenLabs 2.5 or ElevenLabs 3, OpenAI's TTS-1, etc...

@weary galleon

[frigid tusk]

how can i update my lmarena , i still dont have videos generating options

[zealous sparrow]

https://019b45e9-3666-7163-9896-9dde23092d01.arena.site - gem 3 flash
https://019b45e9-3666-76c9-a785-6680e9f50ae9.arena.site - minimax m2.1 preview
which one wins, Glazer codeking would argue its minimax

[zealous sparrow]

its experimental

you will have it or not for now

[torn mantle]

flash obv

[muted timber]

guyn

[frigid tusk]

aigt

[muted timber]

Please help me so that my chat works too, claude, chatgbt doesn't work

only gemini high who is very bad at scripting

now even gemini doesn't work anymore lol

[jagged vigil]

Is lm arena down?

[zealous sparrow]

also minimax m2.1 cant even make a 3d game

AIcodeking should stop spreading absolute bs

[torn mantle]

it did for me some basic physics 3d simulations but its just some basic stuff

[weary galleon]

All his videos are AI-generated. His prompt is just one short sentence. So he can't control it too much.

[zealous sparrow]

it did for me after asking a 2nd time

and plagiarised from gemini side by side

this model is so ass

[muted timber]

is down

[ocean bison]

You mean the guy that posted that Inworld TTS 1 Max boiled ElevenLabs 3??

And said that you can clone voices with 15 seconds 😂😂😂

https://www.youtube.com/watch?v=NDMO7xksmG0

THIS ONE!

LMAO 😂😂😂🤣🤣🤣

[muted timber]

BRO

GENERATE FOR ABOUT 20 MINUTES

I REFRESHED THE PAGE AND NOTHING

WHAT DOES IT HAVE?!

[stray aspen]

lmao

[ocean bison]

I checked and yes, Inworld TTS 1 Max BOILED ElevenLabs 3!

But will you discorver another tool that boils Inworld TTS 1 Max?

[muted timber]

tell me too when LMArena is down

?

[ocean bison]

@jagged vigil

[torn mantle]

😮

[muted timber]

?

[dusky cove]

how?

[south charm]

Last way delete that chat and open new chat (copy that old chat pls)

Anyone know how to fix cant login

[weary galleon]

I'll share my NotebookLM project for all members of our community. It has 49 credible sources and several good Studio outputs. So you can use it for your needs.

[hollow ivy]

Is coast still the best coding-model?

-# ||coast = coASt = co45t||

<@&1349916362595635286> we might have a problem here

[quartz light]

tuff

[old garden]

get outta here 💀

[quartz light]

if anyone is interested in @dookiemonster67 dm me

[mild elk]

Get out of here

[echo aurora]

Some were saying the site was having problems?

Doesn't appear like everyone is having that issue?

[quartz light]

opus is unstable rn

not completing any responses

[heavy smelt]

Isn't it funny that the video arenas were made for people to vote for a video model to build the LMarena leaderboard while barely anyone votes at all in the video arena? Like I vote on the videos that I and (sometimes) others make, while they rarely vote at all, it's annoying because I need at least 2 votes to see which models the videos were generated with.

[light wigeon]

guys banana 3.0 is unavailable at the moment on web? Is there any info or any good replacement that i could use instead ?

[quartz light]

yeah i agree

[ocean bison]

Add o4-mini Image per #1451567412078968832 and add video generation! @echo aurora

[ocean bison]

Just use a second account and vote at the same time!

[quartz light]

since when is there a smart router

[stone cape]

opus has been nerfed

[ocean bison]

Doesn't show up for me 😂😂😂😂😂

[south charm]

Why only me cant login T_T

[weary galleon]

I'm sharing my NotebookLM project for all members of our community. It's extremely good 👍 database and content created on base of it about the current situation in whole AI industry 🌐 . It has 4️⃣ 9️⃣ credible sources and several good 👍 Studio outputs. So you can use it for your needs. ✅
https://notebooklm.google.com/notebook/89ee0d43-a7e0-46e7-987c-ca9e6abc3ae0

[ocean bison]

Generate more myself 😂😂😂😂😂

@weary galleon

[weary galleon]

What do you mean?

[ocean bison]

More Studio Outputs

[weary galleon]

Ok. I'll make more.

[keen beacon]

Think about that for a second

[ocean bison]

Make more myself, not yourself!

[heavy smelt]

I'd rather not see the models than do that, it's not an easy process.

[keen beacon]

The answer is in front of you

But you keep missing it lol (about the voting ) your on the right track and asking the right questions

[heavy smelt]

?

[keen beacon]

Is your concern that the voting isn’t fair or is your concern more of the frustration?

[keen beacon]

In both situation, you still have the upper hand because if we take you r word at face,value you’re doing exactly what you should be doing and you’re playing by the rules and you have expectations for those same rules to be applied to you out of courtesy because you extend that courtesy back and would love to get some simple results to know what model generates what

[heavy smelt]

My problem is that I want to see the models that generated the videos, which needs at least 2 votes, my problem isn't the voting in Itself, but lmarena needs 2 votes at least to see the models, and people for some reason just don't vote, maybe they came here to experience video tools for free, that's it, not to vote and actually help to build the leaderboard, but that's LMarenas problem not mine, mine is that I just want to see the video models.

[keen beacon]

And that’s why you bring up a valid point what would be the point of the leaderboard and voting if this process isn’t being held up to it’s part of the deal

[heavy smelt]

Yeah exactly

[keen beacon]

No, I totally get you and I think that you have the upper hand morally and also justification for raising your concern in your opinion

But I’m also trying to point out to the contradictory conclusion of your question what the answer is which all things ironic

I just wanna see if you could figure it out and this isn’t the right or wrong answer and it doesn’t even have to be shared or you told

You’re thinking your logic is correct and the way you’re conducting yourself is also correct

Anyways, I gotta head out to work keep on a good fight Omar

https://tenor.com/view/top-gun-top-gun-maverick-salute-tom-cruise-military-gif-13548986641468290074

[placid geode]

GLM 4.7 released

https://docs.z.ai/devpack/tool/claude#how-to-switch-the-model-in-use

[echo aurora]

Areyou getting a something went wrong error message?

[quartz light]

yep

not always but most often with opus

i think its some sort of timeout? idk

[echo aurora]

This is understandable. Our intention with Video Arena via Discord bot was the hopes that users would comment/vote/be inspired by other generations being made. It's possible we lower the voting threshold to 1 in the future.

[weary galleon]

Two random guys instead of specific people from prompt.

[south charm]

how can i login now

[clear spear]

omg why when I post a html code it keeps saying something went wrong

[south charm]

@echo aurora help me T_T

[echo aurora]

This issue tends to be most common with hitting a rate limit. If you follow the steps in this guide you can submit more information about the problem which will get our team better information. Unfortunately, with just something went wrong error message that can mean different things aren't working. https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

A way you can check if this is rate limit though: go to Developer Tools -> Network Tab -> run a prompt creating an error -> search Stream -> find folder with the Eval ID (ID that's in your URL) and if you see a Status Code = 427 that means it's rate limit.

[torn mantle]

minimax 2.1 and glm 4.7 are good but not quite the level of frontier models

ive tried both of them

[echo aurora]

@clear spear try these steps here.

[south charm]

T_T

[echo aurora]

Are you hitting the bug where you login, but then are sent back to home screen while not being logged in?

[south charm]

yeah

[echo aurora]

Really sorry to hear that. We haven't yet found a solution to this problem unfortunately. The team is actively working on it. I may have some followup questions.

[quartz light]

try logging in on canary maybe

[clear spear]

It's not a rate limit because I can send other requests- when I insert a HTML code it says that for some reason...

[south charm]

wht?

[clear spear]

THEY AIN'T DONE FIXING IT SO WAIT- simple ☺️

[echo aurora]

Gotcha, still if you follow the 2nd half of the help center article there are instructions on how to get us the best information for what's going wrong.

[echo aurora]

It isn't yet fixed, but the team is working on it.

[south charm]

ok thx

ill wait

[placid geode]

GLM-4.7 achieves significant breakthroughs across three dimensions: programming, reasoning, and agent capabilities:
Programming Capabilities: Ranked first among open-source models and first among domestic models in the LMArena Code Arena blind test, outperforming GPT-5.2; achieved first place among domestic models on SWE-bench-Verified; attained an open-source SOTA score of 84.8 on LiveCodeBench V6, surpassing Claude Sonnet 4.5.
Reasoning Capabilities: Achieved open-source SOTA in the AIME 2025 math competition, outperforming Claude Sonnet 4.5 and GPT-5.1; scored 42% on the HLE (“Human Last Exam”) benchmark, representing a 38% improvement over GLM-4.6 and approaching GPT-5.1 performance.
Agent Capabilities: Scored 67 points on the BrowseComp web task evaluation; achieved open-source SOTA on the τ²-Bench real-world interaction evaluation, approaching Claude Sonnet 4.5 (84.7 points).

[void panther]

hi all!

[swift ermine]

@echo aurora

Canary isnt working

Thats same

[clear spear]

what file type does the thing even support

[echo aurora]

Good to know, thank you

[weary galleon]

It works very good for me.

[ocean bison]

@echo aurora DO YOU THINK YOU CAN ADD o4-mini Image only on LMArena as per #1451567412078968832?

It's NOT official by OpenAI.

[torn mantle]

honestly im trying it rn, seems better than minimax 2.1

needs more testings tho

[keen beacon]

https://youtu.be/Y4lJXczmT4k?si=NX_Qt0PKGADw8gBJ

[torn mantle]

i take it back

glm 4.7 is bad too

they should stop benchmaxxing seriously

wth

[weary galleon]

The main benchmaxxer is Scam Altman!

[torn mantle]

tbh all of them

[weary galleon]

no

[torn mantle]

at least gemini 3 flash feels like a good model

[weary galleon]

Pro is also great

[torn mantle]

these chinese models they all fail at coding related reasoning

told it to give me minecraft game, the UI is perfect but then the physics is just all wrong

[weary galleon]

And Haiku, Sonnet, and Opus are the greatest models ever!

[torn mantle]

no

[weary galleon]

yes

[torn mantle]

no

[weary galleon]

yes

[modest prism]

The only way to stop benchmaxxing is to stop AI influencers and AI enthusiasm to care about benchmarks

[torn mantle]

ok both minimax m2.1 and glm 4.7 failed to create minecraft game

m2.1 being the worst

[weary galleon]

Because they are bad. Claude is the best!

[torn mantle]

if you want to test them ask them something difficult

for the love of god

stop asking them for svgs

and websites

ask them something that requires actual reasoning

[weary galleon]

I don't ask anything difficult Chinese models.

They can't do difficult things

Minimax is very good in music generation, not code generation. In code generation number one is Claude.

[torn mantle]

lol

fair

but they should stop claiming SOTA in X

i just read glm 4.7 SOTA in swe

like wth??

glm 4.7

its also trained on gemini 3 pro

same frontend designs

[weary galleon]

SWE says GPT 5.2 is also SOTA, but it has 17 place in LMArena leaderboard.

[torn mantle]

yea

[simple holly]

👋

[somber plank]

Because Model is not Generalized as gemini
Gpt sota model 5.2 is good specific Things not everything, gemini is Good overall so

[torn mantle]

https://z.ai/blog/glm-4.7

lmao

the website demos at the bottom

its a total rip off from gemini 3 pro designs

LOOOOL

[south charm]

HEY GUYS

[somber plank]

Best in coding section i guess?

[south charm]

GOOGLE CANAVARY WORK FOR ME !!!!!!!!1

[torn mantle]

its just a gemini 3 pro ripoff

[south charm]

@echo aurora GOOGLE CANAVARY WORK FOR ME

[torn mantle]

even the galaxy shape is the same as in gemin i3 pro 😭 😭 😭 😭 😭 😭 😭

the font too

[south charm]

oh hell nah

[torn mantle]

If you had told me that this was Gen by Gem3 pro, I would have believed you

[weary galleon]

Is it bad or good?

[weary galleon]

And?

[south charm]

I have eror again

bc i sign in to my gg account

[south charm]

btw is some kind of lite google i think

[torn mantle]

did you read what i just said?

[somber plank]

Gemini flash is better than pro , They probably refine their sota model in next month

[torn mantle]

wym by and...

this is so shameless

[somber plank]

As like This year they did with gemini 2.5 pro

Maybe!

[weary galleon]

Yeah. You said Z makes stuff similar to gemini

[somber plank]

.....But GLM good in coding

[weary galleon]

Maybe, and?

[somber plank]

Idk , I didn't use it yet

[unborn ocean]

did not respond :(

this was the image

it is so obv

even the same words, lol

[torn mantle]

yea

well its good for competition ig

just dont call it better than gemini3 pro or flash

[south charm]

Hey guys glm is OP lol

[echo aurora]

Glad to hear it! For the sessions that aren't working on the regular site if you're able to provide the Eval ID here that'd be helpful: #1451836386293448725 message

[torn mantle]

well at least its better than minimax m2.1

[south charm]

yeah.

[torn mantle]

https://x.com/denny_zhou/status/2003143612876603474

Denny Zhou (@denny_zhou)

"Gemini 3 flash achieved 90% acc @ 1 million ctx. ... what black magic did they do?"

yes, we've developed powerful new techniques for Gemini 3 long-context reasoning.

X · Dec 22, 2025, 4:40 PM

[celest fractal]

Why did this start appearing almost constantly?

[keen beacon]

Are you using vpn ? Proxy or have add blockers ? Any browser extensions that may interfere with normal browser functions ? Firewalls?

[celest fractal]

Nope. But I tried using it, and it also causes a captcha. That is, with or without a VPN, a captcha appears.
But maybe the browser is interfering. I use Brave, it has built-in anti-advertising

[echo aurora]

This tends to appear whenever the system is detecting odd behavior.

If you use a different browser, do you see the same?

[keen beacon]

Try DuckDuckGo browser

Brave CHROME & Firefox seem to have a lot of issues more commonly

[celest fractal]

I haven't tried another browser. I'll give it a try. Thanks

[echo aurora]

I've heard cases where Brave browser doesn't always play well with our site.

[celest fractal]

Thanks

[keen beacon]

If not clear browser data history everything

[verbal nymph]

hello

[keen beacon]

If all fails u can try flushing dns/reset modem

That’s pretty much all you could do

[celest fractal]

Thanks

[keen beacon]

Oh one last thing to keep in mind if you working fast running through constan new chats or simple looking through previous ones to quickly this also triggers, the same problem more consistently.

[keen beacon]

Yes, if you get it resolved, please let me know what worked

[echo aurora]

+1

[stray mortar]

tf is ernie

[echo aurora]

Just any kind of behavior the system might think is inauthentic will lead to these.

[gusty helm]

for a second I read the update and was like what? gemini losing?

then was like oh, we talking 3rd league

[stray mortar]

is ernie good at coding

[atomic lagoon]

I find it funny how in the text arena the top 9 models are all the same just different types with claude and gemini

[keen beacon]

Hard to lose a race where the winning only requires throwing larger amounts of cash

[atomic lagoon]

And grok

[stray mortar]

timr to test ernie

[hushed gyro]

Chat is the Ernie 5 model any good?

[atomic lagoon]

Well can't say that then, richest person in the worlds AI model sucks

[hollow flicker]

GLM 4.7 dropped?

[stray aspen]

Ernie sucks

[robust sluice]

I'm using Brave and found a lot of issues, the big one is Recaptcha but when I turn off Brave shield this bug is gone

But it doesnt work with other I dont know why

[hollow flicker]

Grock isn’t normal

[inland quest]

right after 12830130318 captchas

[atomic lagoon]

?

[keen beacon]

Figures. I had a feeling that it might’ve been an extension feature within brave that did that and caused that.

[hushed gyro]

WAIT WHAT

4.7 is out ❤️

[thorny schooner]

How are y'all doing

[hushed gyro]

But I doubt it's any good since 4.5 and 4.6 aren't really up to my expectations

[keen beacon]

🤣

Sora solves captcha

[echo aurora]

I'm not too familiar with Brave, is that their built in ad blocker?

[keen beacon]

All Sora

[stray mortar]

ernie is extrmely antisocial, it thinks for 10 seconds and contemplates everything for a simple greeting hey

[hollow flicker]

we do not talk about Grock Spicy mode

[celest fractal]

What's also strange is that I saw a separate thread in the bugs section describing this problem, and it was created when I, and apparently everyone else, had this problem on a huge scale

[hollow flicker]

who created LMArena

[keen beacon]

It’s still occurs to this day every single day multiple times a day

I don’t even consider it an issue. I consider a feature at this point.

[robust sluice]

maybe, it can be turn off for each site, and I turn it off only in LMA and captcha bug is gone

[celest fractal]

I had this problem before, but it appeared very rarely. Now it happens literally all the time.

[keen beacon]

Did it go away?

With any of the suggestions?

[celest fractal]

I haven't tried the solutions yet. I'll do that a little later

[keen beacon]

So then ur not having issues any longer?

Or currently, I mean

[whole swallow]

Where is gpt 5.2??

[celest fractal]

No, the problem still exists. But I haven't tried the suggested solutions yet

[keen beacon]

There’s strategically not on there

Text is easy for ChatGPT

No need to spend any more energy or time on the five series then needs to be

They should be allocating the resources to their strongest areas of the model because text easy 2 catch up for

It makes no sense pour anymore money into ChatGPT five as a series

[whole swallow]

Mmm

[keen beacon]

They’re probably training or have been in development for a while now of the sixth series

And so I’m assuming that that’s where all there money time and energy is kind of focused on

Have a little team behind 5 to maintain

Since it’s not a secret that the whole five series is just been a flop from its launch cursed

[cloud zinc]

google training gemini 4

[keen beacon]

No version of ChatGPT 5 it’s going to be good regardless of the iteration

Its the windows 8 of ChatGPT

Windows 8 was so bad that windows jumped all the way to windows 10 and skip nine altogether

Distance itself from Windows 8

Open AI should do the same

Go straight to seven

[echo aurora]

It was a project created by researchers at UC Berkeley. This blog post has more information about the company as a whole - https://news.lmarena.ai/new-lmarena/ I'm sure there are more specific articles out there about the founders if that's more-so what you're looking for?

[nova sluice]

Hello, does anyone here have a topaz ai subscription?

[hollow flicker]

im just bored so ye thanks

[proud bobcat]

glm 4.7

no way

[wicked sage]

@old garden hi trey

[old garden]

hi

[keen beacon]

A brain is not a singular operating system

The human brain, for example, is in a embodied intelligence

Meaning that the information of processes through vision, audio, and another stimuli is external to it, so it relies on these features like the eyes, the years the mouth

In order to navigate through the world

And embodied intelligence

[keen beacon]

The problem here is the terminology being used general universal super

[wicked sage]

SHUT UP

[calm lagoon]

@echo aurora Will Glm 4.7 be added to the site?

[wicked sage]

you use websim bro

like genuinely

😭

[pale obsidian]

https://x.com/i/status/2003156119087382683 🔥

Z.ai (@Zai_org)

GLM-4.7 is here!

GLM-4.7 surpasses GLM-4.6 with substantial improvements in coding, complex reasoning, and tool usage, setting new open-source SOTA standards. It also boosts performance in chat, creative writing, and role-play scenarios.

Default Model for Coding Plan:

X · Dec 22, 2025, 5:30 PM

[calm lagoon]

Yeahhh, this

[calm lagoon]

Legend

[keen beacon]

Congratulations, that’s very impressive and a very large amount of money raised.

[weary galleon]

WoW! Even Chinese models are much higher in leaderboard than GPT-5.2 and GPT-5.2-high!

[dim briar]

👋

[silent galleon]

👋

[keen beacon]

You ever heard that quote that goes something along the line sometimes you need to take two steps back just to get a step forward?

[keen beacon]

Is 2026 the year open source models surpassed closed?

[cloud zinc]

open source are not multimodal

[echo aurora]

Yes

[keen beacon]

That’s a very good point I never even considered something like that. That’s a really good perspective.

I wonder what extent the leaderboard has in context to people’s wallets

Open source vs closed

[fickle venture]

Woah GLM 4.7 better than closed source models daym

[fickle venture]

What does preliminary mean

[native yarrow]

beta results/testing or not final models

[fickle venture]

Oh alr

[keen beacon]

Does anybody know any service or platform that you can mash up different audios together like into a mix or something?

[cloud zinc]

they use gpt 5.2 synthetic data to train

[zealous sparrow]

zebra revealed btw it was prob glm 4.7
and holoscope was ernie

[calm lagoon]

Hiiii, brotheeer, thank youuu

[cloud zinc]

use ai for dev

[calm lagoon]

What this zebra

[echo aurora]

If you hover over the tag there is a description as well.

[fickle venture]

Alr ❤️

[fickle venture]

I think it's GLM 4.7 name

They later changed it from zebra to GLM 4.7

[weary galleon]

GPT-5.2 arrives: “I’m best!”
Top benchmarks, big chest.

Synthetic tests? It’s king.
PR makes the rafters ring.

But LMArena? Real crowd.
Seventeenth. Not so proud.

Lower than 5.1’s glow,
Even lower than 5, yo.

OpenAI shouts “We won!”
But that rank? Mentioned: none.

“Look! Scores! Charts! Hooray!”
“Please don’t read that page today.”

Benchmarks sparkle, neatly seen,
Arena’s got the meme machine.

[calm lagoon]

Ааа, wow

[zealous sparrow]

@torn mantle

[native yarrow]

i hate how every ai model goes "that/but/the x? y."

[hoary elbow]

Another win for the open source community glm is apparently almost as good as Gemini three flash and Gemini three flash is revolutionary

[native yarrow]

it's way behind gemini 3 flash

check the score

a 20 point differential is HUGE but it just ranks above gpt 5 medium so that's why it's 6

21 point wtv

but then you see how much it's AHEAD of GPT

just shows you how bad GPT is

[hoary elbow]

Ok sorry I thought it was not a big difference because

[weary galleon]

Huh?

[zealous sparrow]

unparalleled in BenchmaxGI

[native yarrow]

nah lol wasn't tryna be mean

but like, to properly differentiate the models

just check the score

notice how most of it is spaced between a few scores like 5 or 6?

[hoary elbow]

I know you weren’t trying to be mean

[native yarrow]

and then MAJOR jumps are 20+ or even in the 60s

[cloud zinc]

gemini 3 flash has better slope

[native yarrow]

that's like the gpt 5 to the gemini 3 models

and then everything else is honestly pretty close apart

it's hard to really get a objective difference when comparing the models

[native yarrow]

ykwim like they phrase a clarifying sentence like a question and then elaborate right after :P

[cloud zinc]

vertical slope is better

[pine wedge]

How is GLM 4.7 for rp?

[proud bobcat]

W

[proud bobcat]

ill be sure to check today

[native yarrow]

i too am curious

[queen veldt]

Where the google tho 😂

Afraid of comparing to Google

[proud bobcat]

i mean thats not what glm is comparing to

glm's competition is mainly claude and chatgpt

of course versus gemini its not close

its not meant to be a gemini replacement

[native yarrow]

i mean yeah why even attempt to compare yourself with the current king

of the ai scope

[unreal shell]

Should i use to gemini pro and unsub to chatgpt plus

[native yarrow]

definitely

[fiery gull]

Even glm 4.6 is sooooo good to rp

I'm testing glm 4.7 now, lol before I called glm "a copy cheap from gpt" but the glm is doing something better that gpt 5.2 😐

[hushed gyro]

Uhh chat is 4.7 any good?

[hushed gyro]

They should've added Kimi K2

[fiery gull]

Yeah, I'm testing is really good

But unfortunately I think it is worse at all than opus 4.5

But against gpt 5.2 and gemini 3.0 pro it does better in certain areas

[hushed gyro]

Compare it to kimi

I might switch

[fiery gull]

Hell no it humiliates kimi k2 to me lol

[hushed gyro]

WTF?

[quartz light]

after using 5.2 a shitton i agree, its ass. 5.1 and 5.2 are worse in coding than 5 for sure

[fiery gull]

I work with text, docs and write, glm 4.7 is REALLY COOKING

[hushed gyro]

Tbf the Z.ai website is quite annoying to use

They should add the ability to switch between models

Cuz I wanna use 4.7 with 4.6V

Before 4.7V comes out

[fiery gull]

4.7 is almost free

Really good model

[hushed gyro]

And HOW DO THEY MANAGE

To have 4.7, literally a month or two after 4.6???????

[fiery gull]

It's so good they could have called it 5.0

[quartz light]

5.2 released less than 1 month after 5.1 btw

[hushed gyro]

The real 5.0 (or 4.8)

[fiery gull]

But glm is a real improve LOL

[hushed gyro]

But it was rushed bc Openai was taken aback by G3 pro

[quartz light]

fr

[fiery gull]

The glm 4.7 is just better that 4.6 and it just better that 4.5

[hushed gyro]

Tbf I was suprised at the graph

Like 4.7 was on par with 5.1 high

👀

[fiery gull]

I'll test more the glm 4.7, but I'm thinking to use cursor + glm code plan now

[hushed gyro]

Can you test jailbreak for me

Ahem...

🤭

[fiery gull]

I don't need jailbreak for my work, glm 4.7 already does everything I want, different from gpt...

[hushed gyro]

Ik but like test it for me pls

🙏

Like in a new chat 😁

Nah

4.7 is just not for me

But I'll test more

[fiery gull]

Maybe I use in portuguese lenguage 🤔

But in my work it is better that gemini 3.0 lol

[hushed gyro]

Yeah you should've used English

Works best on all models

[fiery gull]

I tried but the model need to make the work in portuguese

[hushed gyro]

Wdym "make some"

[native yarrow]

i think GLM the best open source model rn

[fiery gull]

Bruh

[fiery gull]

The rip is talking the kimi k2 is better

[native yarrow]

honestly i don't know if i see open source ever beating closed, closed just has access to too much technology and this is because they're big corporations with a lot of money. if an open source corporation ever got big enough to have the funds to research technology on the closed source level, they'd have to close down as well because then their presence in the market matters, same thing with every current closed corp rn

[hushed gyro]

Idk K2 has a very unique way of writing fictional text

And it genuinely asked questions that made me question my world view. Literally.

It questioned my prompts....

[native yarrow]

how would you describe its story writing

[hushed gyro]

It's very human

[weary galleon]

Which one looks better and reads better?

[native yarrow]

sounds interesting

[hushed gyro]

What tf is this poem lmao 😂

[native yarrow]

so none of the stupid "but/the/this x? y." crap

i hate that sm

[hushed gyro]

No

And it questions me all the time, I was like wowwwe

[native yarrow]

and the constant mini allegories that aren't necessary

[hushed gyro]

I'll send you examples

A scifi story

QWEN 3 Max vs 4.7 vs K2 vs 5.2

[neat apex]

K2 is too autonomous, it even can beat agentic Gemini 3 and Grok 4.1 due to that

Looks to be obeying you for fun, not because its being payed

[hushed gyro]

Elon reposted a post which contains the LMArena leader board!

[native yarrow]

yeah go ahead

i'd love to see

[celest fractal]

I cleared the cache in Brave. It still asks for a captcha every time, but now it's enough to simply confirm that I'm human without having to spend a long time trying to solve the captcha, which, moreover, didn't solve it successfully, even though everything was correct

[zealous sparrow]

btw minimax m2.1 is just distilled gem 3

new imagearena stealth model

[weary galleon]

We are not a tech support. We are just chatting here.

[celest fractal]

IK. But I was asked to report on this

[weary galleon]

There is tech support channel, this channel is was jutting.

[ocean bison]

@weary galleon Even MiniMax's Music 1.5 was very impressive and very very good and Music 2.0 is even better!

[celest fractal]

What's the problem? I'm texting, so I'm communicating. If you don't like it, just ignore it

[zealous sparrow]

nvm z image is another company

[ocean bison]

But the older video models are just trash.

At like some tests.

[weary galleon]

Agreed. Real musicians are losers, AI generated music is much much much better.

[ocean bison]

Just like Pika 1.5 😂😂

[weary galleon]

I somehow like Pika 1.5