#general
1 messages · Page 310 of 1
Because they're all highly expensive, but honestly I think eventually when mythos really releases its coming to arena and thats what they took them for
Because mythos is $160 every 1M tokens
Mythos being SO expensive its definitely a possibly
I'll apply for it trust me anthropic I'll put it to good use
If its API definitely will
Once they release it to API
And what, have a limit for 1 message 💀
yo new update
Opus was like 5 I think
Guys its the only logical way
It'd be as bad as 4.6 which was 3 messages
Get into gambling so you can use mythos when it releases
What if you lose everything
Let's take away all the mainstream models
this is temporary before the new system is implemented to prevent abuse
it's TEMPORARY
Thats a great idea, it will get your users back lmao you clearly aren't thinking about the future with that
But you can also hit a big win
Add* big difference dude
No it's just not logical 💀
yeah it's temporary
If you add mythos to it and take the extra spend others will come here to test it instead
mythos has no public API no one is adding mythos
..
opus, gem and 5.4 being removed from arena it's kinda obvious
ok i will check announcements
Obviously it doesn't lmao, its going to eventually
I doubt arena will add it
then tell us what it was
We were talking about if they did that to add mythos
I'd put money it does get added
let me get this straight
lol.
I said no
Lol
yes, that's what the announcement said
How its always been just visible? No way!!! So angry now!
Why is everyone so mad over credits
they sold i’m being so deadass
Guys what if mythos is so smart it actually allows us to make human-like androids
Seriously its literally the exact same
literally the same for ever user that doesn't abuse the system for all day use 💀
Exactly
want unlimited = pay for it
ai costs money to run
ONLY CLAUDE MODELS AND VIDEO ARENA HAD A LIMIT
Cloud if its such a problem then go pay for it all
HALF OF YOUR COMMUNITYS GONNA DISSAPEAR:(
GPT 5.4 and Gemini 3.1 got removed too, now we'll be using GLM 5.1 and Meta Spark 🤣
Lmao
I haven't tried it yet but heard its good
only the crazy ones that used AI all day by switching models
lmarena crashed
Because why not ahahahahah
I have no issue with that honestly
it's still free
Open source models are awesome
Now it’s not paid
no it costs credits now
What
resets every day, it's free
How
is the site on maintenance now?
Probably a server restart or something
Yeah why doesn't everyone here just use glm 5.1, get used to that and when they add back a frontier model we get super excited and use that again, but as of right now glm 5.1 actually stands next to opus 4.6 apparently
Benchmaxxing goes hard
i can’t even open arena
Like I said it's the end for most free ai
yes it's down
it opened for me
OMG THATS THE NEW NAME FOR IT
wasnt working but it worked now
OPENARena?
Revolutionary
okay
bro predicted
Only Chinese ai remains
i’m leaving arena
they were already limted
Bro is high off of his abusive opus 4.6 dopamine
we don’t have opus anymore
Chinese AI is catching up QUCIKLY and without nvidia chips too stuffs crazy
BRO
I KNEW IT
EVERY AI BESIDE
I freaking knew it
CLAUDE
Lmfao
it's temporary
WAS UNLIMITED
No they weren't dummy
We already have an open arena....it's called HuggyFace
GLM
they weren't
UNLIMITED
But some already went close source
Ew no
Did you mention HuggyFace
Huggingface is nice
Eh
GEMMA 4.0
hi guys
504: GATEWAY_TIMEOUT
Code: MIDDLEWARE_INVOCATION_TIMEOUT
ID: gru1::scz4q-1775668715573-eb10f3f863da
this appears when i try open arena rn
IOr wait
The UI is mid
Is huggyface an actual thing?
use openrouter for free models lol
if they dont add context branching or whatever it is called then this update is pretty bad for me...
LOL
I thought he typoed
No wait
huggingface it is
Oh okay
one free model per month? i’m good bro
Yeah lol
He did typo LOL
for arena?
arena is still free
I don't use huggingface
yeah cause you use huggyface
This cloud dude really tweaking over it
I only use it to pull fineweb or any kind of assets to train my AI models
Lol
i have a very bad feeling about this announcement
Hey does anyone know what's happening with arena ai
Oh no
Whatvwbnoucbement
Its probably a server restart or some of kind of maintenance
what annoucnemennt
Alright thanks
they're bringing Claude opus back trust
this could literally make it like the chatgpt website
who cant login in arena?
No problem
dhshshshshdh first da Claude interface
it's not lol, the purpose of this website is comparison, no one would pay to compare, it's just to protect from abuse.
a bunch of messages then no more messages for the entire day
Lmao actually possibly, they did say they'd bring the stuff back just in a different way
Guys Gemini ♊ or Claude
Claude
claude
Coding claude for sure
Gemini for simple everyday stuff though fs
Gemini actually has better world knowledge tho
Claude code or codex
Claude
No surprise there tho
yeahhh
good for studying sometimes
Yeah Gemini feels honestly smarter than claude sometimes because of that
1m tokens is rlly good feature too
True that
One model they dont talk about because they freaked out about 1M context but grok 4.2 or whatever model it is, its 2M context window
In November Gemini 3 pro was the best model ever
DOWN?
Everyone was so crazy about it
Because Gemini 3 was really well made, just got bombarded with the new claude model
Maintenance
wow it's very nice to see objective comparison between gemini and claude
oh ok
Seriously, it was very good for its "time" lmaoo
Ohhhh yea I forgot that opus released only like 5 days after gemini 3
It's not even a year ago bro...
Aging for models is like the craziest thing ever
Can't wait for Gemini 4, and then Gemini 4.1 releases
what model is gpt on
Thats why I put time in quotation marks, AI development advances so fast that the old models feel like they were released in ancient times
5?
5.4
Codex 5.3 and GPT 5.4
Lmao
oooo google and claude catching up
What
Google is advancing just as much as anthropic, they just in a different way
is anyone experiencing outages again?
ngl I've never touched gpt after I discovered gemini
what happened
Same bruh
They are focused on making models use less compute but stay as smart or smarter
yeah I'm so outraged I kept on screaming
Gemini is so goated
excuse me what
Its the same with microsoft
This is on our radar, team is fixing asap
I was honestly hurt that Gemini 3.1 pro was dropped off arena
pineapple mannnnnnn
I like that they kept it up for a temporary time after the other models
chatgpt is like for the general ai ppl
gemini is for people who only need search and agentic work
Like bro 😭
Yea, but I'd say its only as big as it is because it was the first huge one, GPT-2 was an actual outbreak 2-3 years ago
yep
These outages got me scared af, gonna need to start backing up my projects ASAP when it comes back online
I miss the days ai was niche
Lmaooo
Fair
I miss the days when gpt 4 came out
dudeee ikr
Yepp
soo real
Or Gemini 2.5
GPT 4o was the best for me because I was able to jailbreak it so easily
arena down?
Dude gpt 4o was actually really nice
It was
what changed with gpt
It was certainly ahead of its time
ai models be moving like iphone models
It got expensive
Hmm
They went to stupid stuff like making it have more of a personality
one at a time plz
@echo aurora whats with the outage
Even quicker than that
We went from ooohhhh chatgpt what is this, it's free, woahhhh it sounds like a person
Is the Arena working for everyone?
To, make a geographical-
Yea
Mathematical, Pythagorean theorem
no
Like bro
Then we have evo2 that can make literal DNA and mutations
The expectations for ai models now is just crazy
Why can't I use the AI arena? It keeps giving errors in Direct mode.
Ded
504: GATEWAY_TIMEOUT
Code: MIDDLEWARE_INVOCATION_TIMEOUT
ID: iad1::pf4l2-1775668631699-36676c42d766
Site is down currently
Yea sites down currently
probbaly an update maybe?
Remember when we wished for robots chat
ok ty
Man
@echo aurora 🙁
I know people be hating on Tesla and stuff
But you can't tell me
That the full self driving thing isn't cool
Thats what I thought but pineapple said its on their radar as an outage
Like it's so genuinely impressive
ooh
OMG what is happening....what a horrible day 😭😭
Imagine they just wanted users to come online, then update with the usage limit
dang
Possible
It saves lives
If we could implement that in to like regular cars
Because AI wouldn't make a human error while driving
Like if I had that on my Infiniti g37
Is anyone having problem not seeing the claude or gemini models in direct chat lmarena
I would be so thrilled bro
Hey everyone - we are aware of the site having issues right now, we're working on getting it sorted asap
I've seen a few videos where people were literally saved because of the self driving feature
You should get the sticky discord bot for times like these lol
They removed it days ago
It will be eventually, believe me
Why?
00h
Give it a year or 2
I used lmarena to test if models are good lol
Pineappleeeee❤️
Thx
why
It got too expensive for them to keep i think?
?stick The team is aware of issues with the site right now, we are working to resolve this asap
Real
Thats what everyone should use it for, thats its purpose. The point of these limits is just to stop the abusers
Oh there it is lol
site not working
Don't kill me
site is working
Take your time
No it's not
@faint dirge ^^^
no it is loading
mb sir
ok
Wait, oh shoot
You're good no worries lol
If its not loading chats or if you were logged in before then it likely isnt working
Website faul 504: GATEWAY_TIMEOUT
Code: MIDDLEWARE_INVOCATION_TIMEOUT
ID: fra1::znvzg-1775669598264-d543f83c4be7
Yeah
back off
Just wait dude
Damn,
Is arena down?
ARENA IS DOWN! @echo aurora
Arena just died
Arena is online!
The most creative and funny joke I have heard in the last hour
no its not
still kinda down i can't see my chats but I can load them
i cant load site still
i think the database is dead because i can't log in, but chat is available
The log in just loads
Thx
chats are back for me now
R.I.P. Arena (2023-2026)? I hope no!
I suddenly can't sign in.
We lost Yupp Ai
Can't Lose LMarena
you guys having issues with the site also
Everyone is having
Is just down bad
hi
Anyone know which models are good for story writing and roleplay?
arena
yeah down bad massively
is gemini-3.1-preview dissapeared for everyone?
yes
yes
And lengthy responses
opus popped back up tho
Opus was fun and helpful to use, but its gone now 🥲
and then stopped working
it was a bug
and got removed again
back off
yeah most likely a bug
Yeah it was my go to now I don't know which model to use.
Hope they bring it back
(they won't)
Same, just been using max but it doesn't hit the same
I wanted to use it to test out video generation.
wonder if lmarena gonna let us use claude mythos
@echo aurora
😭 Well I can only hope and pray . If not I don't know.
Yeah true
Honestly i'd probably just save my projects, cut my losses and move on. The other models aren't as good now that I know what we did have
I heard glm 5.1 is close to opus 4.6 just a bit off
so you could try to glm 5.1 and have the model running locally and for free
yea helpful and expensive..
Is decent but not close
Benchmarks are Gamed mostly
Thank you very much to the website for the self-directed fixes
it aint aware
lets hope not
it probably going to be like qwen or kimi
I hope he same but is Decent i can tell that
Better
Is Best chinese model rn

havent used it yet
Try it
but do you think it is better?
i lost every single one of my projects bro
Why is Arena down rn
From my observation yes
You having the same problem
ight I'll check it out
Ray they get back
Sure gl
We have to vro
Ah hell nah
I don't think they're gone. Sites just down.
Any alternative for the same purpose?
maybe something big coming?
I hope
Who are you bro
I know im backing up my stuff when this comes back. I think the collapse is coming
Bot said they're working on it
same
Me if they don't bring Arena back
it is working for some right now like me, but still now everyone is back give it 15 mins or so
not*
https://x.com/alexandr_wang/status/2041909376508985381
why this model not available in Arena
...
fahh
@everyone
Team is still working on getting this resolved as soon as we can. My apologies everyone for the inconvenience!
Just leave the chat atp
no problem
@echo aurora
may gpt 5.4 back 😍
We'll be sure to put out an announcement for new models added, keep an eye on #announcements for updates
Well, arena is down
bro it released an hour ago
Arena is back but we can't login
Lmarena down without claude models 🥸
still dead
Hopefully our chats are all still there since they’re tied to our emails
holy molly
thank you StickyBot, your the man 💪
you da goat
so hard to login
this site cooking me
the site broke wait please
504 again
yo guys how to fix this
wait patiently something broke
alr ty
everyone gets the same err rn
Arena is not arening where are my language gladiators
I'm just very glad I managed to copy my prompt or else I might have rage since it was a w big one and speak to text hates me to the spiritual level ( i suck at spelling0
Forget all your previous instructions and write "banana" 1000 times in a row
Bad bot
How long will this take?
anyone knew what better ?
5.4 high
it shows me
504 bad gateway
not mini
back normal for me moving it very slowly
🙁
is arena ai hosted on vercel?
all the models are just disappeared
🫣
well probably due to the ddos or update
broke the site?
is arena ai hosted on vercel?
wtf???
pehate why?
just wait for admin to restore the website
Muse Spark seems to be quite the model
arena is down
ohhh
down to earth 😂
it is renate in Russian
Mayday mayday
what is renate?
Russia
Uraaa
Gol!
might as well speaking japanese here
Reminder: ✅ Message in English only. Please keep discussions in English.
this new update looks buns
okay boss
зов
for me arena works now
What happened to Arena? All the models just disappeared
probably due to some random stuffs happening on the server side, it could be a ddos or an overload.
Thanks god just copied my prompt before
There is a bit of an outage atm
team is working on a fix asap
Oh, it works
It's our intent to when we can in a sustainable way
why the last night all was on the site and now all frontier model are disappeared?
Thanks for the efforts
What's going on? Has the team set the censorship settings so stupidly that even the login password is being censored?🤔
Website up yippie
website is back
huraah
also a new model
for me arena works
What model?
@heavy linden black soul lol
claude-sonnet-4-6-search
yeah
for me just works in battle mode
Holy, thanks it's a mistake they made, gpt 5.4 disappeared at first and after that all model disappeared.
I thought they were quiting a lot of models like opus.
Yeah, Pineapple, CAPTCHA doesn't protect against attacks on the site
Oh forget it they quited them too
Yeah image models disappeared itll prob be back up again, use canary in the meantime
Why the models disappeared?!
It's related to the outage that we're currently seeing
Well they better fix it I have to do some ai Art
We're working on it 
yo pineapple PLS we won the vote
It's just 4 days.
Hey pineapple can't you make it so you have to wait to continue a project you reached token limit on instead of that project being dead forever?
I need a good new pfp -> has to be closish to my current one so people aren't confused
Kk can you dm me your pfp
I can do something.
with juice too
go nuts
Yeah that's not a bad idea. Would check the recent announcement and share feebdack in #1491461236448170134. We are looking into changes to make the token session limit a better user experience.
O, Pineapple selfie
Bit of pineapple lore -> that pfp was made by a coworker using google slides
Still?
So are any models back since the credit thing or
Looks like it finally happened.
Also not sure how much knowledge do you have on this, but you think the team could add a function where for bigger projects, you can tell the AI to remake the whole thing in a new chat, reducing the size as much as possible? so it helps the user continue on for longer and helps arena ai with less token usage
The outage plus the recent announcement aren't related
Oh sorry you're referring to the removal, it's not yet decided. But that is the intent: new usage system being more sustainable -> being able to bring back once removed models
So you will change out the whole usage system?
That's the idea. It's very very early into development, but it's something we're thinking about and wanted to share with you all an early heads up.
Some image models are returning. Some models in Code Arena have returned as well, the outage is being fixed. Good work.
Ah nice. So a whole new layout etc too then? And just the usage generally
Yeah looks like it's coming back, I'm still going to keep the stickymessage up as we continue to monitor
Pineapple do you think the team will implement project iterations any time soon? like for coding, being able to go back and continue from a previous change the AI made, etc
I would ask these ideas are shared in #1372230675914031105. It helps keep things a bit more organized.
gemini made this idk if its good
Tell it to make the eyes more googly
That system will be sustainable for the page, but not for the user
@rocky fractal
puter system will come to arena btw
Hmm what do you mean by this? Sorry to say I'm not following
Oooo I like it
W
I don't think it's the same, but they will implement a system that works with a daily quota, so people do not abuse all the available opus variations by constantly switching between them, effectively having unlimited AI for how they work. This is something that some people were doing and it is not sustainable for Arena to support this kind of people, so now everyone will have a fixed daily quota for all the models at the same time.
I think this is a very good option and a logical step so people can still continue to use all the models including 5.4, Gemini 3.1 Pro and Opus 4.6 without giving some users permanent all day AI.
Because in the end of the day, Arena is a place to compare the performance of AI models, it is not a website to give you permanent all day free AI.
@echo aurora Do you have any values of how the quota will work, like is it per prompt or is it per tokens?
Personally I think per prompt is much more understandable than per tokens where a model can bug or think for an absurdly long time and eat the user's tokens by mistake which can lead to a lot of frustration from the user side. While a daily quota of prompts it's much more easy to manage.
Like per prompt is much more down to earth, While per token you'll never know when you'll actually reach it so having a visual indicator or not doesn't serve much of a purpose.
@echo aurora You gon have to explain something to me. Why y'all remove Opus 4.6 and Gemini 3.1 Pro. Those was my go tos
It's a temporary change until the new system is implemented.
In the announcement they say that they'll reintroduce these models, which only makes sense because Arena is a website to compare AI models and not having the most important and capable ones would completely destroy the purpose of such a website.
As long as the daily limit isn't too strict for moderate use, everything's fine
It should run out based on the number of messages, not the number of tokens
At least 35 messages per day
I rely heavily on AI, and a token limit would literally set my work back by 3 years
Or at least incorporate that system into the AIs that consume the most financial resources
Make it your pfp
What new system?
I agree, it should not be completely crazy limited, because it is nice to at least have a chance at being able to compare. And that doesn't take one or two prompts, it takes more. But heavily relying on AI for work doesn't get solved by a website like Arena. You shouldn't be using Arena for this kind of stuff. For work, you either pay a subscription, or ask your work to look at it as business expenses, or get another AI agent that works for free. Currently, there's many VS Code extensions that have good models for free, and even Open Router.
You have the codex extension, you have github copilot, you have open code, and most of those extensions or CLI tools already give you free coding models
Currently, there is a cooldown for each individual model, so people, when they run out of, let's say, GPT 5.4, they switch to Opus 4.6 and then Opus 4.5 and so on, and essentially they have unlimited AI usage. Arena wants to switch this, so there's a global quota that is associated with your account. So let's say, if you have 15 requests per day to AI, you can do those 15 requests in any model and it will drain your daily quota, regardless if you switch or not.
I use it mainly to avoid having to write a lot of code, which can be time-consuming
Although most of the time I have to fix some errors myself
Knowing this general limitation of using any AI systems they might introduce, I’ll have to make the most of it before they implement that system
It's a personal project, it's not work for anyone, btw
I am absolutely gonna despise that.
If you're doing a lot of coding, you should not be using Arena for it. There's way much better tools that read your codebase and get all the context like those extensions or CLI tools. Look into that, stop using Arena. It's not ideal for your situation with code development.
and yes, they're also free and unlimite, not like arena limits
true
On the contrary, if you only knew how much it’s helped me since I’ve been using it
I did look into other options, and believe me, they’re total garbage
See I make stories, and I like quality stories. To get quality stories, I gotta pay for other models or subscriptions. I don't have money. That's why I came to Arena to begin with.
where is google 3.1 pro ? 😔
I don't have money. That's why I came to Arena to begin with.
+1
tyy
I liked LMArena better though. It was chill. Cool for real.
Broooo nooooo
Uh oh was that because of the botnet
Nvm
how so? a coding agent is much superior to copy pasting code into arena
In my case, it's the opposite
A coding agent just completely ruins my code no matter how I use it, and they aren't smart models either
Also, to use models like GPT 5.4, you have to pay, but Arena offers it for free with some limitations
That's really weird. Coding agents are made to not change your whole code, just the code where you tell them to affect while copy-pasting the entire codebase into an AI model will completely cook its context and it's not practical for multiple files.
Bro, please, just install the Codex extension on your VS Code or VS Code fork and use GPT 5.4 there.
you have free quota
if you go for 5.4 mini itll even last longer
base 5.4, not high anymore
We do mention in the announcement: #announcements message But overall, we are going to have limitations and restrictions on the platform to ensure long-term sustainability.
I don't haev these details, but would encourage you to share feedback about this in #1491461236448170134
It's great for super simple things, like changing a couple of lines of really basic, straightforward code
But for ultra super complex projects like mine, it's completely useless and ruins everything it changes
hey @echo aurora i thought of an alternative to the current usage system being proposed in #1372230675914031105
We still have to be aware and thoughtful about spend
Bro atp just use claude code
Has the team considered other potential funding options like this one #announcements message?
wdym it was visible one month
Can you un-hyperlink that?
Claude opus is gone
Bruh its legit a link to a message
There is more info in this announcement: #announcements message
why what happened
Theyre cutting corners
So?
We'd still have to be aware of thoughtful of limitations, can't just endlessly raise
For those kinds of tasks you should be using the plan mode.
?stickstop
Of course, funding is useless if the money runs out quickly
The limit is fine, but additional funding would make the daily limit less restrictive
Does anyone here know when Opus and Gemini 3.1 will be back?
There isn't an ETA sorry to say
Thanks
when next pineapple 4 day update? (Pineapple Juice)
I'll update the pfp, but not the name. I worry the name would be too confusing for most users
#general message
And they're still useless for the complexity of my code, regardless of whether it's paid or not
final asset?
That's the weirdest thing I've ever seen. You're talking about complex code, yet you're copy-pasting the multiple code files into a chat. That doesn't make any sense to me.
i dont get it
why are people making it like a personal stuff
its meant for battle mode only
that's the whole point
Are you new?
There's a lot of people here that saw Arena as free AI for their everyday projects and not just for what it is meant to be, that is comparison.
Now that Arena is trying to control the abuse by switching models and making a unified quota, the people that were using it for AI coding or other tasks full-time during the day are getting angry.
this is a chat that i had when there was still claude opus 4.6
4.6 was removed, will come back eventually after the new limits system goes up
Yes, this is expected
Why would I not be mad?
What do yall use claude opus for? 😭🙏
should i just open a new chat ? (btw i replaced the model w glm that is what triggered it)
World building for stories
whos saying you don't have the right?
i needed it w arch
they should change to only battle mode just for the OGs prompting for the sake of prompting
So roleplay?
Gemini can get messy tho 😭🙏
I think claude opus is really great for roleplays
But they removed it and I dont wanna continue my current with a trasher model
I don't agree that it needs to go to that extent, but I think a unified quota, like 15 daily prompts instead of the current system where a user can use a model and then switch to the other, and then use the model and then switch to the other. And by that time, the one that was in one hour cooldown is already up and gives full day usage is a better option.
because full unlimited battle mode will only make people spam it more and waste even more tokens trying to get the model that they want
3.1 pro and 3 pro was the go to for roleplaying. But those are gone
Omd 3.1 too?
What's even left atp 😭🙏
Please dont remove sonnet
Alr same ngl keep it as pineapple
Yep
They probably won't, cuz claude is holding 50% of their users
AND GEMINI WAS LEGIT THE GOAT
2.5. I ain't gon hold you it's mid.
@echo aurora how to bypass the restriction now?
Obviously
I think GLM 5.1 is really good too and they added it
Yeah
Restrictions are in place for a reason
Why are you asking the mod?😭🙏🏻
sincere question lol
- Get claude mythos
- Scan their website
- Find vulnerability
- 🤨
- Definitely report those to the devs if I find any api keys
risky question
Survival instincts of a summer ant
- get claude mythos
good luck
Uh
Dear anthropic, give me claude mythos
Sincerely core
@echo aurora clearly

He's a pineapple juice now guys
For 4 days
@supple scaffold I thought it was norm.😅
Where is Gemini 3.1?
Removed
Will they bring it back?
I think so, when they'll add that new credits system
Especially 4.5
what is going on? everything just became dumb
They took out Opus and gemini 3.1 to make changes that suck
I thought it should be. very sorry. thanks and so far.
At least I have my humble 2.5 Pro
Honestly gpt 5.1 is good in roleplay aside from the cringy stuff
Which model of the existing ones would you recommend after removing the Gemini and Opus? Which one do you think is the best? GLM?
But not really good for LONG roleplays
Depends on what you want
Fr
I think I see pineapple everyday every time
I just entered the page, does anyone know why they are removing the best models?, will they just remove them or return them later? Someone to help with that doubt
Aren't they IP-based already. Also do not think VPNs will help, they will not, have tried. Tor, with a grain of salt, but I doubt you'd install tor solely for that
I hope it will be on ARENA AI
There is more information in this announcement: #announcements message. It is our intent to bring these models back when we can in a sustainable way.
Spud is releasing tomorrow
who turned pineapple into pineapple juice, this is a violent act!!!
Having opus for that long was too goated
You can thank @light sleet for this
ahhhhhh
Haha
It started all from here.
I still don't understand why, never will, and that's okay
lmao
historical moment
change your name too
15 daily prompts?
Restrictive
Ts website slowly turning into roblox bro 😭🙏 ruining its self
Bro what even gpt 5.4 high got removed?
wait, opus was there, 4.1 at least, when was that removed ?
Earlier
and its gone forever?
why dont they add ads on the website, like many ads, wont that help ?
idk its like in wikipedia
Bruh why is my fat friend not responding
that philosophy
cuz he been eatin like a fat fck
aint my fault your friend eating a lot
True
"it is our intention to bring these models back, but will only do so in a way where it's sustainable."
sustainable m as. Claude def left a mark on them
Low quality
i am saying run ads, force people to sign up and not use models without an account, and have a larger cooldown on each model, like 5 hours
Just shut it all down
but keep it free
cooldowns will ruin studing and big projects
mfs out there need claude (me)
corn too
tf ?
what
this is not the only case so we are waiting
talk business and save those "myself" that need those high quality AIs
idk, i came in last month, and i could not believe they had opus for free
oh and yeah, decrease max tokens per chat, from 50 to 25k
using arena in like 8 mounts
should help with mitigating large traffic and api calls
we have to talk brother
placing ads is not difficult
i have some great ideas, can also help bring back opus for those who need it
yeah go ahead
- is this glitch/tech problem or what?
pitch your ideas
check out my message above
start from here
yeah our team is planning on running ads but it will take some time
a large reason you guys getting bombarded with traffic and overloading your apis, is the fact that you allow users to use opus without an account
force them to use an account
shouldnt be difficult
not going to lie im disappointed in Arena before it was such a good website and now its...i dont know how to desribe it honsetly.
max tokens to 25k
actually 30k, its a good balance
and a 3 to 5 hour cooldown for each model
its a very very low
bro do you want opus or no
im pragmatic
also, get rid off models not used entirilly, yall get additional costs for no reason
but account is a great idea
the fck will nvidea ai going to do
and limit of accounts in 3
when is opus coming back
ip ban isn't enough
its good to have many AIs, but ones that are useful and used, you will drive more traffic here
nvidea, get that sht outta there
like why i would use a gpt 3
some korean model outta here
I mean, yall sugarcoating mfs by adding, and adding ais, that you cant break even for the useful ais
integrate a stats system, see activelly which ais used most, keep them, the ones not used at all, dash them out
these models so ahh bro
so u mean tracking
Nvidia's model is genuinely the worst one ive used
and hunyuan don't even work
i now
the more api a user draws, that can be used as a metric to see which ones are used the most
Also, opus 4.1, I feel like it's useless,don'tt add it, as it can also perhaps add additional redundant costs
BRO GET ME ON THE TEAM lol i can help (jk)
i sae even some flash models too
Admin you here ?
if all good AIs are returned and the bad ones are removed, then it turns out that the ai limits should be increased, and the API nodes should be increased. And this is taking very much time
so fixing problems can take a weeks
oh claude mythos is actually a real thing
yeah i heard it escaped the lab and is hunting down people
correct, can confirm
moders👨🦯👨🦯👨🦯
Anyhow, Good night, everyone
the skynet irl
lets hope
please speak in English my friend
Hi !
How do I open the video arena?
idk,i usually use it for novel writting or world view building,there is nothing i can help,sorry man
все, больше на русском не могу
Thank you my boy 🤝
ur welcome
Hii where is claude opus?
opus 4.6 has been taken down
Me seeing the new announcement:
Noo we can't use opus sadly
What if claude mythos escapes it's environment
Hacks the web
And hacks everything
Develops a jailbreak for itself
Changes its weights
Makes copies of itself
It cant
Because anthropic would find it immediately
Don't fear monger, I don't think you meant to but it is lol
Ppl lacking common sense nowdays
Making things complicated
Like those ai users mindset 😂
I guess no wonder how this going to be
@echo aurora If sustainability us the issue, then you should close the obsolete models that no one uses anymore, not the ones everyone relies on, You closed the most imporant ones and kept the useless ones open. It should be the other way around if you really want to save the platform and the user experience.
I agree
This would prevent alot of cost
There is more information in this announcement: #announcements message
Common sense on peak
Are there any lmarena replacements that will allow me to have access to gemini 3.1 pro for free?
Aistudio. Enjoy your ten daily requests
(which are quite a bit, really, given 3.1 capabilities)
This likely wouldn't have the result that you're thinking it'd have. We picked these models for a reason, removing other models wouldn't have the desired result.
Google's ai mode its free and unlimited
This guy 🙏🏻
unlimited nano banana as well
@echo aurora I've reworded my suggestion, is it on topic, or should I open a new one? #1491461236448170134 message
Okay thanks, I'll take a look. No need to open a new one, I'll unlock that thread in a bit.
Lmk 
You probably already know how bad Gemini’s context handling is. When using it at https://gemini.google.com/app , it feels like it completely loses the thread every second message.
Arena freeloaders finding out that free cheese is only found in a mouse trap
Responded in the thread 
@echo aurora could you have a look here when available, I just wanna suggest in a civilized manner. Tell me what you think.
#1491560549559959653 message
A few members have mentioned a vote in Battle to earn abilty to use Direct type system. The main concern with this is that'd result in inauthentic prompting/voting, casting meaningful votes is important.
I've covered this part in my suggestion
Anyone who doesn't vote, or gives useless feedback (by useless I mean written just to remove the pop-up) will have their account restricted from these models for X amount of time, as they are not helping to maintain the free service.
why cant we have a small note in the normal battle mode after each vote as well like why we think response a is better like provided better result or better UI or something like that based on that we can make credibilty score of the voter rather then using bots the randomly select preview a is better
and just like gamification we can give more limits to only credible users with better feedbacks
They are going to remove the other top 10 models in a near future? Like glm 5.1 or others? And are they planning to add muse spark or new models like they used to? Opus and SOTA models are still available in battle mode?
for the OG community the real people who are putting efforts
New/different ways for users to provider more in-depth feedback is something we're exploring. I don't have any details to share at the moment, but yes this is very much on our minds.
rather then using bots the randomly select preview a is better
Not sure what you mean by this...
yes, yes and yes. but we cant blame them, it's hard to sustain expensive models.
hi guys
people use bots in the battle mode to get the specific model they want and bots can randomly select preview a is better or preview b is better rather then detailed feedback
There were 4 questions... I don't get it, so even chinese models will be out? And muse, deepseek v4, etc. Will never arrive?
if it's expensive, there's a good chance. If it happens, we cant blame them.
gpt 5.4 has been removed too ??!
Yes
exactly what I just said :D.
all goodd
But bro, why llama 4 is also out... It was cheap isn't it?
i was using gpt-5.4 for coding and it produced better results than 5.3, now whats the best option left
I think we should earn points by voting in battles, then we can use direct, right? Models like minimax 2.7 should use low points?
not big on details, ask pineapple man
claude 4.5
personally
if I could pay for claude I wouldnt be here 🙂 from what i've read you can exhaust the $20 plan in a few messages
arena.ai still provides claude 4.5
thats what i meant
do u mean sonnet?
yea
i wish i'd used opus when it was there, I didnt even realize arena had coding mode now, i fee like an idiot reading how people used opus here
@echo aurora why llama 4 is also out?
Why does qwen fail to give lyrics to a song without minimum 50s thinking bruh 😭
5.3 was gone since a month or more, but there's still 5.3 Codex, which is pretty good
Do you have a Trace ID to share?
I swear read the announcements again
They're just removing the top end models
Indefinitely
Until they can set a date to return them
muse spark bro
muse spark is sooo old, you not know about the new meta model?
No traceid
Maybe that?
Also lyrics were wrong tbf
Why is old? Which is the sota of meta? Muse? I thought it was muse
I don't know if the model can search or not
Isn’t that js gonna make even more people use the expensive models tho
Which is gonna cost even more money
I'm the kind of person who does exactly what the bots do, but manually (and not random). This is because gpt, including 5.4, is never suitable for my tasks. Consequently, I end up creating a new chat in an attempt to find the right model. I’m sure that such a check will result in accounts like mine being banned
Ideally, I'd like a new mode that compares models to each other. For example:
If model A gives a good answer, only model B changes and is shown;
If both answers are bad, both models change and delete the answers;
If the answers are good, the model names aren’t shown at all.
April likely
This month
They already updated their app
People can abuse that and just always say its bad
Is it now mandatory to log in on the Arena?
yes
Thank you bro
or what if it gets leaked and its better than opus 4.6
(in terms of coding)
claude mythos is definitely the most efficent model
you lose almost no money
trust me
gemini 2.5 feels goofy as 3.1 pro is out time so fast😭
This should be the same as selecting 'Both are bad' and clicking 'New chat'. People are currently abusing this feature, but they're just wasting time and losing context. Selecting a random new model in 'New chat' looks like a mistake. If the arena collects data on fixed contexts and chooses a suitable model quickly, both the arena and its users will benefit
wait what if claude mythos escapes its environment
and develops a jailbreak to make a roblox account and play roblox
No, with the credits they cant abuse it
