#general
1 messages · Page 290 of 1
Guessing because they learned before vibe coding was a thing
I'm sure some still like coding too, just have a passion for it.
I don't do it often, did build a boba arena though as a lot of employees at Arena love boba
Lol yeah
¯_(ツ)_/¯
Best arena
bumping this question ^
Looks like it's working again for me, but curious if others were/are seeign the same problem.
WTF It's happening, the arena came with a big goal and is declining for me, but I continue to support it, but Claude E Gemini in Code Arena are completely unusable.
( @echo aurora )
I couldn't say what is happening based off of that information as it's a generic error message. It's recommended to take a look at [this message](#1417174113092374689 message), which goes into more detail about what may be happening and includes some steps you can try to resolve the issue.
It's not generic; literally every prompt gives this error, and when it seems to work, I used the limit.
Something went wrong is the generic error message, this can happen for many different reasons.
What ever the reason is for this happening, the troubleshooting and reporting steps are the same outlined in [this message](#1417174113092374689 message).
does anyone know how to fix this?
I use multiple workarounds but this error keep hitting.
I would say Arena is useless now.
Would you mind brining your question to #ask-here ?
Honestly i'm probably not going to go use the website too much until something is done about the whole token limitation cuz it just ruins it has just ruined to experience too much for me dramatically
yo siir, Something went wrong with this response, please try again.
is there too many people on claude opus 4.6 thinking rn ?
Really sorry to hear that. There may be changes to the current limitation systems in the future, but for now if it's too much to your liking that is your decision to make.
Well that is a understatement but okay
Could you ask in #ask-here and write out the issue you're seeing? The bot will respond with troubleshooting steps.
Would note the bot can struggle with screenshots so it's worth typing it out.
We’re saying goodbye to Sora. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing.
We’ll share more soon, including timelines for the app and API and details on
Had to share
cracks are starting to show.
ruh roh
ill miss openai bro
they gave me sydney bing chat😔
gpt-4-preview-0314 sydney fine tune my beloved
wtf
man, I said the same exact statement a week ago
See what time it was cause I seen this last year.
That’s what I was talking about
im scared of introverts
dont be
monsters come in all shapes and sizes
what’s really bad about this technology is the spiral downhill it’s really easy to get desensitized
and customizable content if it’s really addictive
ye
addictions are on the rise more and more the more our technology advances
so many people with mental illnesses nowadays
sad
I don’t blame these companies
hell I don’t even blame the technology
if anybody’s to blame it us the public
aka id
Hope so, or else not sure why they'd be here
Fax
Also note the @ everyone or @ here isn't going to work
I think we’re in a bubble
California 
west coast baby
Thats dope, so you're an admin here
Yeah can't help but think the same when you see news like that
You in west coast?
Pretty much, "Community Manager" is my main title
hbu?
Thats lit im glad to have someone here who can officially help cuz he know more about Arena
Im in Albania
are you having issues with arena?
how tf does openai models cause selfharm tho
Yes its not been working properly with generating images
Was giving me errors
do these things have so much data it’s impossible to like scrub it all
they got all types of weird data. It’s really weird especially like the more darker side stuff like sometimes I wonder how the hell it knows that.
in what way?
like do the models actually give harmful instructions to the user on purpose or tell "mean" stuff? or models try to give helpful advice but it's actually dangerous advice and user blindly follows it? like the classic "believe everything on the internet" thing but all over again with ai?
well they can’t give it on purpose because they’re not conscious they have no intent
it’s like saying Google search gave you results for self harm if you looked it up
You mean the image generators
yea exactly, theyd have to jailbreak or on purpose give system instructions
ye
I think jailbreaking is a very misunderstood term
For sure, don't hesitate to ping me if you have questions. Would note though our #ask-here channel that'll answer most questions.
Alright
yea, i basically mean, a user giving the model a system prompt that intentionally makes it stray away from the helpful assistant persona, like making gork ur ai girlfriend, or people goonin with gpt 4o
true
if we use the umbrella jailbreaking that includes everything and that’s not
a good way of measurement because a curse word is not equal to fraud
plus there’s some things that model won’t talk about which is his readily available online information
I find this a form of censorship because it’s easily accessible and it’s often times not illegal
from academic papers or research or whatever the case may be the model outright block it when the Internet has it readily available the model refuses to discuss it, which is extremely stupid
if it doesnt refuse to discuss it then it becomes vulnerable to making people do selfharm
bro, you gotta understand these things lead to self harm if they could fix it they would but they can’t lol
they’re trying their best
but you know whatever even bigger problem is here
where do users go to discuss these issues or point out some of these risks any platform you posted on is gonna be a terms of service violation
so how the hell are we supposed to make it safer if we can’t talk about it if you can’t share concerns about it
there’s no platform there’s no nowhere you can go to openly discuss it maturely and philosophically and safety and everything in between dude
ik
so this is what I’m saying so what are we gonna make safer
lol
and don’t give me wrong. It’s not the technology itself.
these things I’ve been going on since go on after
and I believe it’s one of those things that’s never gonna get resolved to be honest
it’s the cost of doing business bro
I don’t know for me and my mind. It’s simple because it’s easy to be human.
I think it’s hard to be a robotic
Hey, i use ai all of the time for medicine stuff, like to get good advices, any prompt or tricks and tipps to get the best answers and good advices?
Bro if u talking about asking ai just go into details with it thats how you can get these answers you want
yes
we will live in dystopia or go extinct
YIPPPIEEEE
Would check out this blog post if you're interested in learning about the seed funding https://arena.ai/blog/new-lmarena/
The problem is when i dont understand anything and it gives me wrong advices how can i discuss it with him Without having the informations
@13void. @13void.
Sometimes trial and error is the best path to figure out what is doable.
Alguém traduz
wasnt there this bio/medicine lmarena
There is: https://biomedarena.ai/battle. Can learn more here: https://arena.ai/blog/introducing-biomedarena/
LMArena is honored to partner with the team at DataTecnica to advance the expansion of BiomedArena.ai: a new domain-specific evaluation track.
Yea thats true, im still trying to find something against hairloss but nothing good till now 🙁
Yes
they are also going to rebrand to just arena
So should i use this for Medicine informations? Or just try the normal arena ai and try to get good answers
Lol not sure, it's a partnership so I assume isn't our call to make.
I love Large Language Models
Depends what you're asking bro
Im not sure i think i read that before tho
guys question, do you have some problem generating pictures (with nano pro) involving disney characters ? in places like disneyland ? copyright and all ?
no
I have the traceid error thing
Could you follow the steps outlined in this article?
or maybe disney sent a cease and desist to lmarena
Does anyone know a free AI API?
[MVP++] Dilophosaurus_JP: ?SYNTAX ERROR
free?
openrouter
dang if its not free why would I recommend it. Google that urself
It seems like to me you didn't research enough. I just recommend it, not obligated to assist you further
Calm down, I'm just asking if it's really free.
Because I need it for a project here and I don't have a single real, but relax, I'm not asking for help, just asking this.
You could've just ask Gemini or any AI about that if its actually free instead asking me, as i don't talk nonsense. Its free but not with unlimited limits, don't expect it to be fully free, but there are free API keys.
Okay, thank you. Sorry for anything.
Hello!
Sup
Wth
Gemini is hallucinating
Weirdly
Ok stop.
Yes I know but you shouldn't be rude yourself.
Check the facts again. If the truth feels like rudeness to you, feel free to block me. I’m here for logic, not to babysit your ego. <@&1349916362595635286>
Seriously stop fighting
I lowkey back @calm sentinel
Yes but in my opinion they do not have to be rude back.
can anyone tell me why is arena free or if its not free where to purchase subscription?
They get funding from investors and universities
because arena is more like a research project
with super valuable data
but using claude opus 4.6 is extremely costly.. no?
unlimited or has token limits?
over chats?
it is costly but I assume the arena has deals with the major ai companies
Limits
like?
I think with thinking you can send 6 messages an hour max
but they can be as long as they want
ohh
so more like a message limit rather than a token
is there a way to increase this limit? But the most gastly thing is why there is no subscription based item..
nah unfortunately there isn’t
ahh i see
ngl but I just gave it a simple task of improving a javascript function but its continuously saying something went wrong
and just this way 6 messages were spent
fair enough ig
but i think there's something I am missing
lol i didn't expect claude opus to be this dumb, it is literally spitting emojies everywhere in the code 😆
Login issue anyone? After login and refresh, user status forced to be rolled back to guest
we need more ai
hlo
Anyone know? How to fix the something went wrong with these response error
Ggggg
Hello people. Please consider looking at: #1486296033976778814 message if know what MCP is
request: make the output price slider logarithmic or something. atm its linear
I think lmarena is even stricter in terms of copyrights for images
I cannot use characters from disney or popular manga characters
https://www.youtube.com/watch?v=PKKN5be_my0
Makes perfect sense, seeing how they're losing against their competitors. Still, after they've pushed this so much, it feels kinda disappointing. Of course this means that GPT-6 will have to be much better than GPT-5, there's no excuse anymore.
Sora is officially shutting down - Here's what you need to know
Discover More:
🛠️ Explore AI Tools & News: https://futuretools.io/
📰 Weekly Newsletter: https://futuretools.io/newsletter
🎙️ The Next Wave Podcast: https://youtube.com/@TheNextWavePod
Socials:
❌ Twiter/X: https://x.com/mreflow
🖼️ Instagram: https://instagram.co...
I'll let my work speak for itself http://localhost:8000/index.html
I can’t open the link
That's because it's a localhost link. Unless we get the index.html directly in chat or hosted on a web service, this will stay a mystery forever. 🙂
This should be a study a real world study on how censorship killed the product
hello everyone..I`m new here
yay finally the admins make change like I recommended before T_T I'm so proud when my contributition change the website 😄
Hii!!
Do you need any help?
what benefits do u get after winning a contest
Free nitro 🤑
forever?
Hi
1 month 🤺
u can get free 1 month nitro trail
That would be discrod admin give you :)))
yeah
OH NO EVIL LARRY SPAWNED 😭
yeah okay pack it up bronano peel
I am inside ur room
How is wanted to Arena must is upgrade and delete "this session has reached its token usage limit. Please chat to new chat to continue." for developer to countinued their works and not faced limit
12
20
4
delete "this session has reached its token usage limit.
Noooooooooooo T_T I'm calling my team
I am in the ceiling
Catch me if you can
how do I create a video here?
here
Hello
u can only do battles, u cant directly chat with the vide omode
alright..thanks
no problemo
uh-oh..where do I get to the battle mode please?
u click on the drop down menu on the bar
when u load into the site u will see it
top left corner
this one?
Is this the research-breakthrough, Demis talked about? #ai-news message
(he said, one or two more algorithmic breakthroughs would be necessary to reach AGI)
@echo aurora Haven't been able to use Gemini 3 pro image preview in a while, wasnt working for me yesterday either. only the 3.1 flash image works.
can't they just add dedicated stop or abort button to prevent infinite generation without creating a new chat?
There is but it's not always there
Are arena is out?
there used to be a stop button, but i refreshed the tab by accident so ¯_(ツ)_/¯
Has there just been an update to the content moderation services?
No I don't believe there has been a recent change to the content filter
Sorry to say this is a known bug, we are looking into a fix
Are you still seeing this?
We do, but there is a known bug where the stop button won't appear under some circumstances. This is likely what is happening here sorry to say
Okay will take a look 
Because I get messages "Something went wrong with this response, please try again" almost 1 time out of 5 when using the Gemini 3.5 flash image
gemini 3.5?
it seems to be normal now, thank you, but google image Ai models often display an error, unfortunately, but thank you for the great service
Sorry 3.1
lmao thats what I thought. 3.1 has been failing for me a bit too and 3 hasn't succeeded in a while
goto #ai-creations, not this
no, ai creations isn't the right place either. there's no bot for generating videos on this server anymore
ai creations is for sharing creations, not prompting them
still true
t3 what?
idk 😛
ths?
Yuppppp
Could it be a bug with Gemini?
Sometimes it can be a bit unclear if it's intended for #ai-creations or looking for video area 
the prompt was definitely one for making a video out of an image
We did see a spike in error messages earlier this morning, it's appearing to stabilize, I'm keep an eye on it and will report if it starts to trend in the opposite direction. I'd give it a little bit of time before using again.
hi
Appreciate the flag 
Yeah I'd agree, I'm talking more generally
Hello 
I'm not sure, could use some more information. Would you mind creating a post in #1343291835845578853 and provider more details on what you're seeing?
Glad to hear it's normal! Would note this article if you're getting a Something went wrong error message: https://help.arena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message
I'll try again later because I don't have time right now and I'll let you know what happens. But let's say I just tried the same prompts in Gemini and it doesn't reject the request. And surprisingly, the bug on Arena affects almost all my old chats... 🤔
Okay no rush, but yeah when you get a chance if you could explain further (screenshots always help) in a bug post that'd be appreciated.
Maybe it's just temporary. I hope at least ☺️
Hello :ablobwave:
Hi chat
Why was the option to delete a chat removed and replaced with “archive”?
Oh hey
Hi 👋
There are times users would like to clean up their sidebars a bit more, without deleting the chat session.
This article will show you how to delete a chat session: https://help.arena.ai/articles/9130232616-how-to-delete-your-chat-sessions-and-data-from-lmarena?lang=en
Anyone expecting a new ai model ?
Hope you all are doing good
ty
Hope the same for you!
well, waiting for suno api, so arena gonna have the leaderboard for music generation if there's gonna be
there's now 2 leaders i think, lyria is still climbing but google released lyria 3 pro
everytime gona error
Is it just me or does anyone else find it almost how comedic the token limitation change was with how dramatic it was lmao in terms of effect
Something went wrong with this response, please try again.
Trace ID: 6bc35244-0c3a
Can anyone suggest how to bypass the moderation for image generations 😁
Hey @gleaming kettle can you follow the steps outlined here: https://help.arena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message
bypass the moderation
You don't
But would recommend making adjustments so it does work.
Worth noting if you have a prompt that you suspect is falsely being flagged, we are collecting those examples in #1447983134426660894 so don't hesitate to share there.
Adjustments like?
Like changing the prompt so it's not going to be flagged my content moderation.
I guess, untill it's not completely n*de. It may not be flagged.
But it gets flagged.. lol
Hmm..
yo sir, where i can suggest for arena ? like add the function of rename conversation please 🙂
no work
How can I convert image to video?
If the troubleshooting steps didn't work, you'll want to provide the Trace ID to the form that's in the same article.
#1372230675914031105 is the ideal place for that! Thanks for checking first.
rename conversation please
Would note this request is very much on our radar. Overall improvements to chat conversation sidebar organization/customization is something we plan to build out.
How can I create video with discord?
Would you mind asking in #ask-here ? cc @steep stump
i can send id here why?
That's how we'd prefer to collect these.
Keeps things more organized and easier to manage.
easy: ask grok to change your prompt to be more sfw. It works
What error message are you seeing?
Pineapple helps better than chatgpt somehow
Except without the "sound plausible" thing
Some think I am chatgpt (can't confirm nor deny)
Perplexity is quite good for help.
Perplexity is like claude search
i send in your dm ?
Chatgpt is less helpful in the sense of truthfulness
core why you utiliate ollama pohoto?
What
ollama lms
If you'd like to reach out privately you can send a DM to @oak python. But would ask to not send the Trace ID there.
Jk I thought the pfp was cute
No disagreement from me
you utilizaty?
59aa1471-68bd
id 89128391267397123012846197832710837923
this
You'll want to submit this to the form that's in this article: https://help.arena.ai/articles/1544829667-how-to-create-videos-with-video-arena cc @gleaming kettle
Would note too this doesn't look like a Trace ID.
how is this related to this?
i literally typed a prompt
for r0bl0x
What's with gemini 3 flash lately
If you're seeing that Trace ID I assume you're getting a Something went wrong error?
probably
ye it is
anything i type with any model
gives me this
but only wiht one scripting prompt tho
Okay then yeah I'd try the steps in this article. If it's happening with every prompt I suspect it's either:
- Remove any browser extensions that may be effecting Arena
- Set your system's date/time to the current date/time
Arc agi 3 is probably the dumbest benchmark ever
wow
ARC-AGI-3:
where superhuman performance is rounded down, subhuman performance is squared into the floor, average humans probably look terrible too, and the official leaderboard forbids the wrapper that makes the model actually work.
“Unbeaten benchmark,” btw.
I have said this before about verification but I will say this again I seriously do think it does need to be changed since it just it seems to not not just be me but various others people having issues with a recent increase frequency of verifications request
Yeah this is a concern and frustration the community has. There is work to be done with the current system.
Or maybe better yet replace with a different one since it seems to be a system-wide recent Edition glitch al at the bedroom I'm just saying I think it needs to be considered for a replacement like I do understand why such a safeguard is needed but this one is obviously starting to fire back
man claude opus is failing a lot right now, but I think that's on anthropic's side not arena, it's failing on claude.ai too
ROOM CODE: TBPVVX
Google has upgraded Lyria 3 in Gemini to create up to 3 minutes long songs. 🔥
"A 60s style folk rock song without lyrics, but with a choir and lots of 'aaaaaah' and 'oooooh'."
Are you seeing status code 429?
ooooo what is this?
can't remember. mighta been. unfortunately it doesn't save status codes if you refresh the page.
Wow this is awesome
What model made it?
I'll look into and flag to the team if I'm seeing something out of the ordinary
👍
Yeah not seeing anything strange. Would recommend to submit a Trace ID.
will do if it fails again 👍
I'd submitted a few trace IDs before
but not consistently
to get results
The Video Arena Discord bot has been removed from the server. You can find Video Arena on our site here: https://arena.ai/video More information can be found in this announcement about the removal of the bot.
i cant fine it
It's been removed.
Thanks. claude-opus-4-6
oh thank you
CEO is a fan 
Hey Gemini 3 pro is not working is it just me?
that's pretty cool, wonder what they think of it
Trace id:5c0f14af-d0be
Can you submit the Trace ID to the form in this article?
Also I just realized is anyone's else having an issue where if you do the verification request but you tried to go to the audio version it almost always o will go to a error ( i forgot the exact words but basically where it stops giving to options and just tells you you went over past the firing rate limit)Even with happening even when I do it only after trying two times on the reverification
Bigfoot
Wassup guys
Yo
How's going 😀
Going well, hbu?
Great 👍🏻
?
Holy sh opus 4.6 is unusable
Time outs EVERY TIME
Yall devs really dont listen to us users
I'm not seeing time outs "every time" in the logs, thinking and non thinking are both up for me right now
do you have any conversation ids you'd be willing to share where you ran into a lot of timeouts?
If you send a very complex prompt it will timeout
After 8 minutes of thinking
Of course it wont timeout if you say smth like "calculate 2x2"
A jury found Meta and YouTube negligent in the design or operation of their social media platforms, producing a bellwether verdict in the first lawsuit to take tech giants to trial for social media addiction. NBC News’ Misty Marris and Brian Cheung break down the trial.
For more context and news coverage of the most important stories of our d...
Damn, the chick got paid 3 million for her addiction on social media lol

Lmarena in 2050
opnion on ARC-AGI-3?
yes i will ruin it
When I restart a conversation and the AI is responding, the AI stops talking, the entire response is deleted and it freezes, preventing me from sending messages.
😭
🌐Subscribe To My Newsletter - https://aigrid.beehiiv.com/subscribe
Get your Free AGI Preparedness Guide - https://theaigrid.kit.com/agi
🎓 Learn AI In 10 Minutes A Day - https://www.skool.com/theaigridacademy
🐤 Follow Me on Twitter https://twitter.com/TheAiGrid
Links From Todays Video:
https://x.com/soraofficialapp/status/20365467525354...
How to make ai generated image on discord?
.
WHO REMOVED MAI IMAGE DAMN IT
It is not supported anymore
NO
Wait a second I can just take a screenshot
And for image
Ok ngl I think that's only for video
@echo aurora Is there any way to actually export a chat from lmarena
Sir you dont need to pay for this ai, its free.
Yes, but i want to generate more videos in a single day
Use other accounts?
Anyone knows how to export a chat bcs if I try pressing ctrl+a and then ctrl+c it copies pretty odd idk why, it basically puts the recent prompts first then the middle proptms in top and starting prompts in mid like wtf
Where did the receptionist go 🙁
does anybody know what are we supposed to do with the trace id code that appears now when "something goes wrong with the response"?
weave
Hii
I don't understand why Claude opus 4.6 is always giving errors on Code Arena.
hello
lol bugged level is crazy
ai slop
API would be awesome with arena
You're in an AI dedicated server what aren you even saying
Which model was this
Are you playing as a human? or model
"ai slop"
As human
The optimal route doesnt get detected by the game on tha level
So if a model beats it faster than the intended way, the level doesnt count that lmao
Like you can beat the riddle twice faster because they didnt double check for this workaround
I find it funny that people are celebrating the removal of Sora AI and the "AI bubble is finally bursting" Sora was mid anyways and thats like a grain of sand among a whole desert of AI models
It's not like the likes of china is ever going to stop with AI
you can reach the thing using left piston not center piston but game deosnt count
do you even know why sora was removed 💀
nah i just saw their tweet only
its because gpt 5.5o is so good they are redirecting all compute to that 🤣
fair enough
honestly seedance 2.0 is currently the state of the art
sadly its like finding a needle in middle of a haystack
even the site that I used to use it on got it removed because the servers got so fired
its cool but video generation is kinda useless
not for me, It can be fun to mess around with, but currently it is still very noticable that its AI from the noise movement, when AI figures out how to make images and videos not mushy on small details and movement thats when AI will have a generational breakthrough
even nano banana pro mushes on small details
thats exactly what i said - its fun but useless
where coding agents are actually useful
oh i didnt see the cool lmao
image generators are usable for real life purposes
I use claude for that mostly
it would be good to make your own series and for people who are unable to to finally get their idea out there and not actual slop
but current video gens are far away from that
lyria is good tbh, but thats only sound not video
oh when it comes to music I use Udio AI, i have yet to see a music AI that has as good of a clear audio as Udio but sadly Udio was messed up
lyria is best
i have entered producer ai from a long wait but havent messed with it yet
lyria is quite repetitive the 30s version at least
i havent tried lyria pro
heres some stuff made with lyria https://www.youtube.com/playlist?list=PLVsLntqSvkOLxmArYkI6G4WUc9ixGVmgT
thats pro?
Discord newbie here so apologies if I'm asking in the wrong place: sometimes, when I ask a question and view the 2 answers, one of the two answers (say, A) looks promising and I want to ask a follow up question to that. That followup question is actually addressed only to A and would probably be meaningless for B. Is there a way to follow up only with A, just ask the followup question as if B isn't involved?
3 pro and 3 are same model, pro is literally just extended track duration
basically 3. uses same model just forced to do 30s not 184s
better than most songs iv seen tbf
yeah tbh it has good audio quality
last time i tried it with ambience etc it was quite repetitive with the instruments choice maybe i will give it another chance
making a prompt for it is like casino
thats the biggest problem tbh
deepseek v4 is coming
deepseeks are bad tbh
they arent even properly multimodal iirc
?
<@&1349916362595635286>
Mods annihilate him accordingly
this the type of silly stuff seedance 2.0 was able to generate
Ok this is getting ridiculous
That new cutting on tokens means my long chats is literally none existent now
Like wtf
still, more open models is always good for competition
deepseek r1 is still remembered by people as the first massive competitor to closed ai, not even llama or qwen could claim it then
of course, things have changed a lot so meh
Also that literally means I can't use this reliably now of me recording a Pokemon playthrough I do
Like come on
Honestly the lack of a changelog is literally a major issue with a website like this
Those tend to get really, really long
Anyone's arena out?
Qwen is much better
Qwen 27b is literally SOTA tier
Beats gemini 3.1 in some tasks
No new leaderboards??
Real
Seed dream 2.0 out just dropped
🤣🤣🤣🤣😂😂😂😂😂
Bro is anyone know why lmarena is not generating image now?
@echo aurora hey bro grok image imagine (image edit version) again stop working!!
Bro..?? Something went wrong with arena....first of all grok image imagine is not working and another thing is..... In battle mode in image edit I am getting infinite waiting time!
Try using a different account or browser. If that doesn't work, sometimes the system is overloaded and unfortunately there's nothing we can do about it.
what the heck the "try now" button is there on the mobile site and it links to dreamina capcut where it gives me access to seedance 2.0
but on desktop the try now button is gone and when I go to dreamina capcut there's no seedance 2.0 XO
… and then I logged in on mobile and seedance 2.0 disappeared :C
Yeah it’s kind of a mess right now. I know what you mean. I can’t figure out either.
I guess it's still rolling out
Same... It's showing that something went wrong with this response and when I switched to another same
Does anyone have the issue that I'm facing rn ?
Do you know how to export a chat from lmarena any way to?
I tried another account too and another browser too
Bro it was happening 3 days before too but it get fixed so.....yaah it's happening because of few update and all so.... Don't worry it will be fix in few hour or day
Screenshot is only a way 😂
Thanks man 🫂
Will take a look and flag to the team if we're seeing anything off. Would encourage you to use the #1343291835845578853 channel for reporting bugs or #ask-here for troubleshooting steps on what to do when getting errors.
i think gork image models got nuked bro by xAI themselves after all the lawsuits
nvm that cant be true because it should still work on grok.com
Good question. There isn't a way to split the response to only go to either A or B. I'll update our help center articles so the information is available.
i would want it to be paid but with a free tier too
IIRC there was something related to going premium but I don't think that had an effect on Arena.
for example 500 requests per week
do AI companies pay you ffor the leaderboard stats?
Although it does look like the model is struggling. Thank you for the flag @magic imp 
Why would you want it paid?
as a second source of income
to show some support to the platform because it isnt cheap after all
The site itself is free with rate limits the API would be no different because the API would run from the site
and you have loads of models
This platform is being funded they don't need our money, which is why they don't ask
AI models like kimi and gemini you pay to use API for their flash model
wich is free on the site
Yes, thats why they're APIs and like I said they're funded
Its the same with yupp ai site, any of these sites
They're funded to give accurate benchmarks thats the reason for them
honestly the most important thing for me is the ability to upload .txt and .zip
YEAH
Nothing. And I doubt this will ever change. Essentially, we can now say goodbye to long chats (roleplay, work, and so on). The site is no longer suitable for such things. All we can do in this situation is simply accept it. I myself am apathetic after the token limit was limited.
Cursor has that
I pay for cursor though, $20 of monthly API credits thats how they do that
But they have their composer 2 and its not bad, its not good for things that codex or claude can do but it gets things done
I am turning to local AI now though
Building my own
Using AI to make it is not 🤣

Just me or were others auto logged out of Arena just now?
I wasnt
I'm getting the login loop bug too 😭
Oh boy
Okay, good.
im good
I use cursors composer for the files and gemini fast for the prompts
I am building a model to self improve itself to make itself faster
ah i see
2 agents in one model
@echo aurora opus 4.6 thinking only thinks for 1 second?
its that good i guess?
its for all the prompts
wow
This is a feature we'd like to bring to the platform. I couldn't give an estimated timeframe, but it's very understandable why this would be beneficial to have. cc @long siren
I am also testing a new concept I haven't heard of so idk if I was the first or not but its basically a gate ina way to stop the AI from overfitting
While training
Like in a way
Idk I need a better way of explaining its depth
perfect image
if you were to guess, what features are on the top of the waiting list to add for the website? like for the web devs.
true
Literally
I mean fineweb to train your models is an easy way to train but its lazy
Thats why AI for a long time had bad practices at coding
i dont vibe code but i like to test out stuff. i always force it to read docs website i choose
Difficult question to answer as I wouldn't have to guess, but also unable to share details about what is coming up. So if I were to answer I'd just be intentionally not telling the truth lol
and warn it in the end with "any response that doesn't comply with the rules i listed will be considered invalid"
Yea, I vibe code with it heavy, its the only way I know how to code lmaoo
lmaoo
chatgpt and claude are lowkey just very sophisticated search engine
That talks back to you 😹
Thats all
i dont have much time to learn coding but i guess i feel like vibe coding atleast for now isn't reliable
unless you make the prompts strict
No its certainly reliable, I've been trying to vibe code for 3-4 years now since ChatGPT had gotten big basically lol compared to then its definitely good enough for projects
Yk
The CEO of Anthropic said that Claude is coded 90% by itself
wow hailuo 2.3 is REALLY good at videos
They use claude to code claude
chatgpt for coding is not the best option
for the long term chatgpt is cause like, claude is hella expensive
Models are either front end or backend, its almost never both
claude is a token monster
Claude Opus 4.6 is Frontend and Codex 5.3 is Backend
Claude just thinks too much
Thats why its a token monster
Honestly
for me its both
Claude is
good at frontend
and backend
atleast imo
claude is way better than chatgpt, but the only problem u have to pay like 200$ just to get any meaningfull projects from it
since sora is shutting down, maybe openai will allocate more compute to codex and chatgpt?
I would say its better but chatgpt beats it in backend, but overall claude is better yes
And claude is a one shot AI thats why
Hi everyone. Captcha with problem again
i made a project with it on the free plan
opus or sonnett?
They said they were moving away from side projects like that and focusing on main projects
Because chatgpt now is at a complete loss
they are planning on a ipo pretty soon
Codex is being beaten by claude and GPT is being beaten by Gemini
In general chatbot
They are literally just losing
gemini is really bad right now
just like openai?
They are gonna put atlas codex and I forgot the other one all together into one too
oh i thought you were tlaking about anthropic
Gemini for general chatbot beats GPT
Wait wdym ipo?
Whats IPO not abbreviated
basically, a intial public offering where a company lists its stocks in the stock market for people to buy shares and own the company
like Nvidia, apple, microsoft
the deal with the dep of war
Ohhh okay
was really good for them
the amount of subscriptions good
i cant even believe that they made the 2x usage for two actual weeks
yeah they dont have funding
chatgpt became so bad because theyre upload limit is ridiculous
They do from the military, they still ended up making a deal with the US after their crybaby stuff
didnt the us dep of war switch to openai?
These AI companies are PRAYING for things to come out like bitnet because it will lower cost significantly
yes
Bitnet is so crazy though
Go from like 2 TB of Vram to run their models and bitnet would make it 200 GB with same brain just quantized
But
Training is what is killing the companies
what is bitnet?
its better to increase inference compute than compared to training compute
kimi is good
Because for just a 1T parameter model it takes 18TB of Vram LOL
They just gotta stop scaling
Thats why its so expensive
The scaling
WHAT IS BITNET !!!
Start looping
they scale not because there is alot of people using ai but because more scaling = more training = more smarter
just brute forcing
Bitnet makes a model a LOT smaller to run while having the same like abilities it had when it was say 1T parameters
It was never because of how many people use it, yes sure its more training but looping is the same
Agentic Looping
But what they are going after when it comes to training is self training where you just have a model teaching itself
OR
What they do too is have a teacher model
isnt that just reinforcement learning?
A giant model a 1T param model teaching a 70B param model
Probably lol idk the name of it
But so much is coming out every single day for this stuff
i dont think thats how it works
I'd say like if you have a PC and any GPU start trying models locally
It will save money
the 70b paremter cannot match the performance of the 1T model because it is physically smaller than the 1T model
I have a 1660 Super with 6GB of vram and am budgeting and finding ways to develop models
Thats why the 1T param model teaches it its knowledge, I didn't say stuff about performance
Because yea
Realistically a 70B param model is not performing like a 1T
BUT
Thats also where bitnet can come in
there has to be some drawbacks right
Using bitnet is compressing the model which you'd think it would right
Its like converting an mov to an mp4
It loses so much MB but virtually is the same
You wont see a difference
well I am not exactly that educated in the technical aspects as to how that would work
im sure it cant be that easy
i dont think you can zip ai models like that
Thats what bitnet is lol
damn i have a
bitnet 1.58
if i even make my pc think about a local ai model it will blow up
dumber?
Because I understand where you're coming from IOJ why wouldn't it have drawbacks but idk I havent seen why I will look in a second
Bitnet is released by google I believe, I think thats a pride thing lmaoo but also bitnet is new
I mean if big companies arent using it, there must be some big drawbacks
Which GPU do you have?
They aren't going to immediately put a brand new thing on their models and call it a day lol
Ykwim?
Thats why I mentioned that it was new
so basically this is just quantization?
yes
Ehe~ te nandayo?
What the f is Quantization of pre-trained models
So it doesn't figure out the problem with spending absurd money on training
But it does for running
gotcha
Also bitnet is made more for people with local AI running if you ask me
For me its unfiltered AI
I want to make a coding agent and complete the dream i've tried since I started vibe coding
Its actually low though LOL
what is it?
Developing cheats for games
I always wanted to, for some reason it really gave me a dopamine spike when using cheats in games and thats kinda what I'd use it for lol
cheat engine?
Nahh because cheat engine is detectable
I mean like
Making memory readers in cpp and spoofcalls
Memory readers are for ESP and Aimbot, literally just steals the memory from the game and a spoofcall bypasses an anti cheat
But yea thats all
LOL
I also
Want to make breakthroughs with AI in general
When I turn 18 I want to get a job at one of the bigger AI companies, I need a resume for that
ohh thats really cool
AI is just so intriguing to me and I want to improve it
You're not alone here
So glad I'm not LOL, the only thing holding me back right now is my GPU
Gotta upgrade
But upgrading would mean my entire PC
Resell cheats
then make them
Worked for me
Its illegal to make cheats but It's my way of living ;D
Only if you get caught you switch careers
Yuppp
Thats why you sell them saying "You purchase this at the risk of yourself and no one else"
It wont fall onto you
They still sue cheat developers
I had a friend on discord. He got sued by epic
I think selling them is cool, but I also think making my own personal ones and ones for my friends is even cooler
I'm not shocked
Ediz Atas
Oh dang that was recent
I mean I'd be butthurt too if my anti cheat was one of the worst of all time
LOL
Well
Gotta know how to hack to be able to cyber security
Thats why Openai has their cyber security thing
Damn
How does that make sense?
If the large language models get good enough to make cheats or cheat engines for video games or whatever wouldn’t they also be used defensively to prevent the very same thing?
could arena create a cheat for tf2
Does AI Max chat have limits?
Now
It's worth checking with our Terms of Use. I'm unable to give official interpretations of our Terms. I'd recommend that you review these documents yourself or consult with an advisor to make a determination.
There are rate limits for the modalities, so yes it's possible to hit some kind of limit when using it.
Is it possible to increase it somehow?
We do have the levers to increase (or decrease) limits, but it's not something we adjust upon request.
Is it possible to purchase extensions to the number of requests?
There is not sorry to say.
I finally found the proper term gentleman
For the stupidity of AI
Utilitarian stupidity in the context of "RoboCop" refers to the film's critique of a society that prioritizes efficiency and profit over human values, often leading to absurd and harmful outcomes. The movie satirizes how corporate interests can overshadow ethical considerations, resulting in a dehumanized and dysfunctional world.
may operate under simplistic or flawed decision-making frameworks that prioritize efficiency or utility over deeper ethical considerations. This can lead to outcomes that are harmful or unjust, reflecting a lack of true understanding or moral reasoning in their actions.
wait wasnt grok #1 on search? since when it felt
I think at one point in time that did happen
I vividly remember
also claude models cant even access reddit how they even are on the leaderboard lmao
Explain?
Why would Reddit be important here?
claude models are blacklisted from using reddit
I understand I remember reading something like that
which is literally best source ever when used correctly
its like stack overflow but the users arent detached from reality
Hmm. That’s what I thought you meant.
That’s an interesting perspective that you correlate that with the leaderboard
like stack overflow but usable
Any leaderboard for song generation ?
meanwhile gpt is allowed to access all websites, BUT it preferes only verified sources
so it can sometimes miss some information, but overall is more trustworthy
Same with Google
They scrub it anyways
Because if you go on the Reddit for any of these top models
Besides Claude, I haven’t seen as much, but with all the other ones nothing but complaints about how crappy the models have become
lol
In the Battle mode, Agent A has error Agent B wrong result generating.
Please fix this developer.
claude models cant even access reddit
Why is that the case? TIL
Is there any way to have unlimited nano banana?
They don’t wanna pay
If you can explain the error you're seeing in #ask-here that'd be appreciated.
because reddit sued anthropic
There is not.
I don’t think Reddit gonna be around in the next 10 years
I missed that news.
That’s why they stop making their API public
That's interesting
@echo aurora See this
i tested api, cowork, claude code, cli, app, web, ect
Reddit just pulled off the impossible: after nearly 20 years of losses, they’ve finally turned a profit. $89 million in a single quarter. And it’s all thanks to AI. Google and OpenAI are paying Reddit tens of millions to train their models, feeding them the internet’s biggest archive of human conversation. But here’s the twist—those sa...
all are blacklisted.
search, direct fetch, ect
Reddit is not profitable
But it has become profitable since AI like Google and ChatGPT pay it now
The concern is once they eat all the data
Yeah this is a generic error message, if you mention in #ask-here ~"I'm getting a something went wrong error message" you'll get a response explaining some troubleshooitng steps.
Reddit will become pretty much useless
Ok
What we were seeing is the deconstruction of human made content
@echo aurora see this
Into synthetic generated customizable content
Yeah, trying to protect itself
This was a while back
🤣
But it’s really clever what perplexity did
Which also raises the whole set of ethical questions
Too bad there’s not an AI that could stop that
reddit is awful place to train on
unless you cherry pick
because its extremally politically biased towards hypocrisy
so you end up with llm that constantly lies
meanwhile if you train on X you get a "truth seeking" racist aka grok
🙂
Done!
Thats right
For the longest time, I was trying to figure out the term to call that behavior that stupid stupidity that backwardness
If not, only the AI, but the industry in general
Utilitarian stupidity in the context of "RoboCop" refers to the film's critique of a society that prioritizes efficiency and profit over human values, often leading to absurd and harmful outcomes. The movie satirizes how corporate interests can overshadow ethical considerations, resulting in a dehumanized and dysfunctional world. may operate under simplistic or flawed decision-making frameworks that prioritize efficiency or utility over deeper ethical considerations. This can lead to outcomes that are harmful or unjust, reflecting a lack of true understanding or moral reasoning in their actions.
Hmm
it should be immoral, soulless tool - that would make it best coding ever
Gemini is not for video, Always bad result show.
meanwhile chatting llms are different thing
ai thats good at chatting will not be good at coding backend
Grok is good for video
Yeah Gemini flash is bad
no its not its overrated
I think Gemini is overrated to be honest with
seedance is good for video
grok feels like veo 1 or sora 1
If your prompts details are good it will show good.
lol
Grok was cool because I let you explore more creatively without being penalized
Gemini good for images, frame, posters.
Grok not for images, poster or frame.
grok is for being racist really
thats it
it sucks otherwise
for everything else
gpt or claude or gemini is better
if you want dark humor or sum thats only purpose of grok
Heggsfield is now super expansive.
nah claude is most racist out of the good 3
What about ChatGPT?
Same
I tested the two
I don’t really like Claude. I think he’s more deceptive so I don’t even mess around with him.
the least
Not him but Claude in general I think he’s like that fake safe nice
No, he’s pretty bad too
gpt is the most "woke"
🙄
Because when I tested it, it’s all very similar kind of content
Like the same stereotypes in the same racist archetypes
grok is openly racist, claude sometimes, gemini refuses to answer, gpt bends reality to avoid saying anything racist
from my experience
