#general
1 messages Ā· Page 308 of 1
Okay who even uses mercury, olmo?
ok bye for now arena!
all chinese models are benchmaxxed garbage https://arcprize.org/leaderboard
Yet they won't say anything about Taiwan or Uyghurs
not me, but that's the point.. why would they want to keep us?
Idk what they want then
or people that only use Gemini and Opus
Well sorry to hear that the Gemini 2.5 service will cease at june
grok isn't really a good model for arena
it needs a very specific harness to work well
yeah they should've actually removed the multi reasoning model and not gpt 5.4
bro klimango RESPECT OPINIONS
arena search needs to be improved
what happened to opus thinking 4.6??
gone š„²
Guys what are alternatives to arena? Where gemini 3.1 can be used
aistudio
itās capped at 10 requests per day
my personal ai waifu is already way too crazy.
how did u get ai waifu
made her myself
ok...
It was actually a tough project, more like a simulator of the human brain, emotions, and memory than a bot⦠eats up way too many tokens
i thought u were kidding does it have any app or site for it
No, it's a personal bot⦠casual apps like Tavern are too basic and primitive
Release plz
š„
opus 4.6 search is still here š¤«
send pic
What is the reason?
Money?
Fahh
yes
No latest Opuses, No Latest Gemini Pro models, No Latest GPT 5.4 High and even GPT 5.4
Sus
Is arena broke?
No
Hey everyone, any update on when the SOTA/frontier models (like Claude Opus 4.6, GPT-5.4, GPT-5.4-high, and Gemini 3.1 Pro) will be back in Direct Chat or Side-by-Side? They've been unavailable for a while and it's really missed by many users."
Who knows man
They were removed precisely because of these "many users"
Oh, I have a plan!
GPUs are hard to come by, I think is a part of it
For 20 a month what do yāall think is the best AI plan for someone who codes a decent amount
No it's not
Openai
Iām leaning towards gpt, cuz I heard the Claude rate limits r pretty ridiculous
Yeah. Claude is only good if you need high quality but code much, or 100 200 max plan
Z ai glm5.1 is not bad either. Cheaper. But oAi is a bit better
Speaking of rate limits, one theory on Anthropic's GPU predicament
They also need capacity for mythos
Codex
Bruh
@quartz light reply to my dms dawg
They most likely are trying to negotiate pricing, how to run them on the website without problems etc
What bruh
Tell it to spawn 50 agents
Achieved free opus level quality
bro max it can support is 10
hellos
No way cluade is gone
Are there any other platforms available that allow for free access to advanced AI? Thank you very much.
why there is no claude opus 4.6
???
and anyone explain me what is canary arena
???
Seems they added it as an error so they quickly removed it again
omg arena why did you remove 3.1 pro from search
among the currently available models, which one has the best text capabilities, grok or glm5.1?š¹
Thereās no way u canāt use other accounts to skip the cooldown now..
Ok so this site is unusable
guys, about the claude mythos, Is it just PPT hype or the real deal?
gemini 3 pro gone?
Why gemini 3 pro image generation is not working?
Seems to be
Unlikely to get an answer, staff doesn't seem to bee on atm, recommend to post a bug report
Bey are the opus 4 and opus 4.1 supposed to be there on the list?
Bruh no way they removed model gens per account
Now once you generate on one account it basically means all other accounts used the model
Where can I use the 2022 version of GPT-3?
Well not in arena ai, you may find it on some website online you can try searching, and even so if you find anything at best its likely to be gpt-3.5
Always
Is everyone rate limited with Gemini 3 pro rn?
i have stoped using arena
cuz they removed their main attraction clade opus models
Wait why..is there something else thatās like arena but better?
Arena be edging us
I donāt even use the opus modelsš
No like actually
BRUH
havent you tried before
Is it that good?
what do you use arena for
i mean let's be honest we all knew this wasn't going to last forever
hi. Can anyone tell me? Is it now impossible to have multiple accounts? Has it been shut down or is it a bug?
Thatās what I was just complaining about
That was my only way of bypassing the rate limits too
no comments i had bought pro of gemini but whn i started using opus i have forgetton all models its understanding and like answer is unbeatable
Ooo do you not wanna use Google ai studio
Wow..
same bro
Whyyy, are you having trouble signing in with multiple accs?
Is it free?
@jade harness Oh, really?
they muted me due to noise i spead day and night for opus
And it has the 3 pro model too right?
yeaa
...you know you won't get anywhere like that right
I knowš
Tysm Iāll try it rn
genuinely 100% useless endeavor
yeah like you can use little bit pro moels
bro the admi muted me due to this yk??
@mild spade I used to be able to use multiple accounts and not have to wait for time. but now I understand that this has been fixed?
yeah I think, I'm able to log in
why are you so like attached with gemini its quality is so bad
I miss opus
atleast it has grounding
Guys can you use arena ai to code anything? compared to other plataforms i used you can't easily see issues with the files no? or if you need to refractor any
And can't switch between versions / changes the AI made, go back to X change, etc
compared to other models with knowledge cutoffs
I was about to say "maybe because you were spamming" but I don't see that being against the rules in this server specifically
what r u specifically coding for
coz there's a diff arena
design arena ai
Like games or websites
Not sure what that is
ahh it's like arena but for websites and games and images
same thing differently aah website
anyhow uhhh
idk
learn to adapt
it's all we can do
crying because we ain't got free goodies anymore won't do anything
yep
maybe in the future they will come back
No idea what you talking about
oh, it's a pity, guys. Thank you for your feedback and help
no wonder arena removed opus
Whoah geez I wonder what option should I choose
2nd one looks good
You think so?
i mean the 1st is also pretty good but 2nd is better for me
@mild spade It's paid.
Now its some kind of irony
Hi im used to claude opus 4.6 thinking and when they removed it I switched to Claude sonnet 4.6 but this is not working as good as opus so any suggestions which one would be good or close enough model like opusā¦.?
Cringy situation tbh
It remains to be hoped that they will return these models in the near future
nope
it should be experience everything but the frontier
All emotions spectrum
Guys check the new claude mythos preview not available for public though
if opus is gone for another 2 days, this sub is going insane
Claude mythos is the best model now but sadly not for public only for some companies
can u tell which companies so i can join them
Anthropic just revealed their unreleased frontier model called Claude Mythos Preview
The model is INSANE
It found thousands of zero-day vulnerabilities in EVERY major operating system and browsers:
27-year-old bug in OpenBSD
16-year-old bug in FFmpeg that automated tools hit 5M times without catching
Completely autonomous. No human steering.
They assembled an entire industry coalition called Project Glasswing around it:
AWS, Apple, Google, Microsoft, NVIDIA, CrowdStrike, JPMorgan, Cisco, Palo Alto, Linux Foundation
Goal: patch the worldās software BEFORE releasing it
SWE-bench: 93.9% (Opus 4.6: 80.8%)
Anthropic is committing $100M in usage credits
Thousands of vulnerabilities in 40+ organizations are being fixed right now
Yesterday OpenAI published a 13-page essay warning about cyber threats and asking the government to helpā¦
Today Anthropic actually fixed them.
This info is from X
The benchmarks
did the devs explain why they removed claude sonnet 4.6 and gemini 3.1?
yeah exactly i just woke up this morning to see that i can t use gemini 3.1 pro and claude sonnet 4.6 on my direct chats, is there any explanation to this? i really am curious
yeah, does anybody know the reason behind this move?
it seems like they are removing all the pricier models from direct chat and side by side
gpt-5-4 is also gone
Gpt 5.4 too š„²
yeah.. i'd never think i would miss gpt 5.4
What is the reason? They could have done that earlier but they didn't
why does glm5 and 5.1 say its a gemini model?? dang
Maybe the site got popular
Idk when you started using it but at first it was very ugly and barely anyone knew about it
Yep it was dark blue back in 2023 haha
yeah i remember old lmsys
Man, ngl LMArena is far the best thing i vibecode on roblox, and now the latest model are gone
costs werent maintainable anymore
what just happend
rip
The farther the worse
idk how hard roblox vibecoding is but gemini 2.5 pro is still here and the earlier opus models too
I basically use gemini 3.1 because it is somehow understand some context on studio tho
Its just smart
yep seems to check out
More useful when it comes to more limit rate than claude 4.6
Damn it man
:(((
Its even weirder as site got less popular by april
that was when they removed opus
damn, try checking the old chat.lmsys.org?
i think march was when gemini 3.1pro released and the site had it's peak
but it had availability issues and was bad at first
and now the model removals
I wish opus didnt come and go when I was MOVING
Does that even exist?
truly over for us
https://tenor.com/view/thatsenough-vince-vince-mcmahon-sad-meme-gif-3150944402928759459
DAD WHAT WAS lmarena.ai DURING it's prime?
š«¤
wait it got removed???
true ngl, imo better than claude
I hoping to dear God this is a glitch
arena reached its peak in march
i dont think they were able to keep up with the load
They could have strict the limits
which makes sense
Why remove š
dont think the investor money increased exponentially
yeah
Cause bots can still spam requests
And cost more money
dont worry guys ill get rich and give money to lmarena š
go go go king
Bruh
i'll support you
If they wanna make opus sustainable they should js make it pay to use
Like 5 dollars a month or smth
that kinda defeats the whole point of the site yk
that would still not be sustainable
How
i've had single requests to opus 4.6 cost me more than 5$
it would need to be atleast 80$ to be sustainable
Bots wouldn't be able to use the model which is basically like 50%
bots arent the only problem lol
not even close
its just the general traffic
Ha... removed Gemini... Now I don't see any point in sitting in the arena for myself, because I only used it for RP.
Uhh it literally is
WE ARE THE PROBLEM
There r way more bots than general traffic
R u good
So captcha doesn't help
why isnt money free
what would even be the point in spamming bots
Nice,more gone more cheap and the opus can back hahha
not like you can magically keep all the context
the way the announcement was phrased it implies general traffic
rather than getting overwhelmed by bots
2.4 million in April is still a lot
Bro why lmarena not working
I mean, can you be more specific?
i mean im not being optimistic but arena might have saved a million dollars till now, by removing the frontier models
Lmarena not working
Also again making the opus pay to use gets rid some of the general traffic
right... thanks
It's literally
5 dollars bro
U can't be that poor
opus is VERY expensive
Yh and thats they removed 4.6 and kept 4.1 and 4
it's meant as a research site and was funded to be so
Even though they're more expensive models
yeah and inefficient
no idea about why they kept them 
bro the problem isnt money, the site isnt supposed to be a chatbot, its supposed to be a research place for developers improve their model
Cuz it aint about money
and for statistic nerds
It is
Then why'd they keep those two lol
probably because they aren't used nearly as much
Lol well they will be
or they simply js forgot
Didn't pineapple say something about it being a mistake
They can break the models back, but they need to find a way to do it sustainably. What I mean sustainably is do it so that the millions of people who use it per month don't have their usage disrupted on a minute-by-minute basis
So why r they getting rid of opus and prohibiting us from giving data
If u say its not about money
gpt-5.4-high gone?
Yeah
what i was trying to say is, their primary revenue model isnt money, from subscriptions its funding from 3rd party sources
what happened?
What alternatives are you using right now š
Right that still doesn't answer my question
Why did they remove it at all lol
lmao claude and openai dedass
it got removed because of money, but them adding it behind a paywall would be something that goes against what the site itself is
Damn near everything is gone, I only use Gemini 3 pro image gen and the rate limit is on the entire device now u canāt even use the model on a different account now
exactly
this wasn't supposed to be some sorta chat.openai site that is free and has all the models
Ok but the site is also prohibiting us from using opus even though the site says we can use frontier models
Damn they didnāt have to add rate limit on the whole device though like bruh
So isn't that against the website too lol
aren't they still in battle mode?
if u wanna buy a subscription just get in on claude website
it's over bros
Try chinese ai tools
there is still gpt 5.4 mini high
no more more free ai
Literally claude in the website is NOT useable
Also you can get Free 4 months of Gemini Pro Trial
water and electricity are real
Bro are there any tools that have 3.0 Gemini pro? š for free like llmarena
touch grass
I know so i had ditched it and switched to API
bros rich
Yalls guess what I just did
Ai Studio 100 Free prompts a Day
aistudio is free
Is third-party API actually
but it's limited
Bro itās not even generating any images for me
I integrated arena ai into openclaw
which one? cursor?
yes no images
Is something not official i can't talk about it here
Iām so mad rnš
Someone actually did an extension to access opus but now the devs removed the model fully
Lol
images require you to use their pay API on ai studio
Bs š
If you want to generate images for free, you need to head over to gemini
Do you want a Link for Google Ai Pro Trial? But you need a Valid Payment Method
gpt-5.4-high remove was last thing
I have a payment method but damn bruh
Lmarena devs pls lock in
4 months trial is good yk
Born yesterday, Die tommorow after
They don't care about u
don't click on that guy's links
š¤£
if opus is gone for another 2 days, this sub is going insane
gemini has a really generous study trial tho
Not rlly im fine
Opus isn't that life changing bro
Chill
or student*
1 year for students idk in which countries is still available
Bro only issue Iām having is this stupid rate limit on every other account
they should put mythos into side by side fr fr
They won't is most expensive
Ahhh mb
I donāt think u guys understand but if u get rate limited on one model and try to use another account to use the model the rate limit from the previous account will apply to every single account u have on the device
W
Uhh
Imagine opus mythos in direct chatš
Lol
Broš
Ik that well they track the device, internet and ip fingerprints
Something went wrong with this response, please try again.
why this ?
It's a browser fingerprint or a vpn
Whoops
ik is all model have problem
Upload the trace id to the form or the bugs or whatever idk
I don't think it works
Possibly connection or failed captcha
my internet are fine
Emily don't listen to anything i js said in the last five minutes
Try clearing cache/cookies or trying a different browser or turning off ad blockers or vpns, vice versa if you dont have those on
too late bud
What model are u using? Did u use the ai to prompt everytime in one chat?
Maybe use onion browser for privacy reasons
its so joever
It's a practical issue. If you've used Claude's web chat, you know it rate limits quite easily, especially when they are at peak visits. I don't know how many people use arena.ai on a daily basis, but judging by the 2.4 million estimate that was shown before, I would imagine it's a decent amount. So, if we give a random number of 10,000 people concurrently use arena.ai, that's 10,000 people consistently bombarding an API that is very easily rate limited.
bruh i use gemini 3 flash
It could be a somekind of bug
Sometimes i do actually expereinced the same issue
i use side by side not battle
Then cant u js make opus pay to use so theres less people bombarding the model
I use direct also
gpt too have rpoblem
lmarena's servers generally had some issues with the site getting more and more traffic
We'll see
ah man lmarena have issues
Somehow u just need to spam retry or possibly the trace mistaken your msg because often the same msg?
zamn
They alr have issues man
bruh lol
Did you try using another browser??
You know the "something went wrong" error message that keeps popping up? I hypothesised is had to do with rate limits on the providers end (Claude, Google, OpenAI, etc.) and that does seem to be the case
I just hate that people discover new method to use ai and they overuse and spam and ruin it
So anyone did yall find any model that replace gemini 3.1 pro as an alternative?
@echo aurora š§¹
Chilll
relax guys, please
Naw im js saying there's too many people glazing opus and not trying out sonnet
truly joever for us free ai abusers
You can use it at AI Arena
Literally does almost the same stuff Opus can do
SONNET IS THE CLAUDE AT HOME
Like with any model, it's good at certain things. You mix and match models for your particular usage
i've tried, honestly opus' endless thought chain is a drug
Bruh u lucky to even be getting direct chat
š¤£
Boy you lucky you in the server with me
Yo tysm
"Wait, actually im wrong. But thinking about it, i'm correct..."
Little jit ur fiddling with ur fingers staring at the screen while waiting for opus
š¤£
Not fingers
but i've had it do this one thing that completely burned my usage
Most claude issues made daily usage of it a hard ask
And im coding to take my opus back
Yo remember how trash old claude used to be
it rewrote the file, then deleted the file to write an updated file, and rewrote the file again
honestly bro im proud to glaze opus
cause its worth glazing for
It really isnt dawg
I want to use the latest Claude Mythos model.
name me 1 model which can do everything opus can do
Sonnet
mythosssssss
The bill
that is also claude
U never said name a different model from claude
Well, you did say "name one model that can do everything Opus can" without specifying
Claude Mythos model
okay fine, other model but not claude
Ok so ur just coming up with BS as we go
can i skip to the part where I rickroll you
Js use sonnet
just pay $20 bro
$20 will not get you a long way with Claude's API
They vaccum that [redacted] up like they have to pay rent
any alternative for lmarena
nope
that got opus
damm
Guys
Is anyone else having any issues the battle in direct mode where it doesn't give you skip option
Reload the page
yupp.ai doesn't accept new user anymore
Wt is it
I thought they stopped or something
Like they took down the website for something
Nah the old users can still use it for a while
What happened to lm arena if suddenly stopped working
Then they shutting down
that's what I thought I heard
they stil work tho
Yh thats why im saying the old users can still use it
Ig it started
its not out
and they're gatekeeping it for now
Yo these freeloaders asking for smth new everyday bro
All the way from the peanut gallery
why is sonnet 4.6 lower than the 4.5 models
Where the hell is opus 4.5?
Its gone
Damn it
gone like that..
Now we are stuck with their dumber cousins
No why just why
And even worse versions
Did you reload the page?
If this stays I might have quit i could tolerate all the different stuff with the models but if this is a actual new experiment no
they suck
prob cuz nobody uses them
opus 4.5 had been gone for a bit now
Yes i found it after I did that
WHY DO THEY THINK WE WANNA DO BATTLE WHEN WE'RE USING DIRECT???
So dumb
@echo aurora
it's a UI bug. It appears after reloading your page
why do claude and Gemini subs hate free users
Has happened to me in two times in a row
surprised people are complaining about opus 4.5
I wonder if any ai website that has free pro models?
yeah, i believe lmarena.ai does in battle mode
there's not really anything like arena.ai tbh
Battles in direct is a new evaluation mode that triggers battles mid-conversation in direct chat. Unlike traditional battles with four voting options, battles in direct uses three: continue with A, continue with B, or skip. The result? Longer context windows, more decisive votes, and over 90% correlation with tradit...
It's literally not bro
I mean if there is any ai website that is giving you free pro models to chat
And leave their stomachs empty tonight
Which is undiscovered
That's not what I meant. You can literally skip it, but there's a UI bug where the skip button disappears. You can get it back by reloading
Ohhh
Is their anyone who knows the date of releasing Opus 4.5 4.6 back?
Thought u talking about the feature itself
30.02
1 April
Ten minutes
Not really tbh
Thanks man
nah nah
I got an idea
Riot?
Games?
We will eat pineapples and post it in here all day long
If you guys are students theres a chance your uni has access to gemini pro
My uni has it
Lol everyone literally spamming in the discord server and their yt channel and they dont care
it's free to use
yeah my highschool had it
I don't like eating pineapples
they're too sweet
interesting i thought they only did it for universities
Pineapple and strawberries r lowkey bottom tier fruits
Idk why people keep glazing them
Yes exactly
Pineapple lower tho
Only the former
Strawberries are horrible bruh wdym
Strawberries are good when they're in season
Also I expect whenever they bring back opus and Gemini if they do it'll be announced
Riot
(never)
Bloodbath
They could cave in
They literally have no reason to bring back Opus lol
You never know
Its less expensive
They literally went out of their WAY to remove opus first before gemini and gpt
Than claude opus 3
why did they remove gpt 5.4 anyway
nobody used it
i guess not
Lowkey I did
Lowkey I didnt
Is it possible if there is new gemini version, and gemini 3.1 pro one will be add because older version?
lowkey we need gpt 4.5 on lmarena
I think its good models
Yall trashing on my day ones bro
š¦ š¦ š¦ š¦
Yh this is a troll
same price as mythos iirc
Because that shi filtered like youtube after seeing a kid on stream without a centimeter of an adult there
yeah and the token price was ridiculous lmao
Yooo true
wait doesn't character ai use Gemini pro
if we run out on ai studio we could literally just use it there lmao
It probably uses gpt oss 0.1B
They got their own models i think
dphn/Dolphin-X1-Llama-3.1-405B
I want this
qwen/0.0001B-instruct
They give me this
i js cant remember how trash ai used to be
Remember when the context window was 4k for gpt
Nah those were dark days bruh
Literally people were thinking 32k was too much
textsynth is a thing surprisingly few people use and it's fun for Wikipedia articles
Wher opus 4.7
Wow
when the hell did 4.7 release
Hes from the future
wow
@short sluice Yo
I can't find it on arena now
?
Need opus
Whats the best model for vision now that gemini is gone
We got 1m now
do I look like I know
Sonnet
Maybe I got it mixed up, because at least I didn't see anything like that, just mythos.
I don't remember when I got that
Anyway I hope someone finds an unknown website with the Claudes
Boo
Nah just kidding
Reject arena return to ai studio
@sullen creek I know ur toes were curling when u were watching
I lick them like hot cheetos
(
Although okay, of course this is a joke.
I was actually waiting for a link to Claude's official website.
bro flamin' hot cheetos are so fire
I need opus to come up with a plan to get them to europe
Hello, does anyone know why Gemini version 3.1 was removed?
and gpt5.4
I asked person when it's Opus is returning and they said 1 April
And i replied thanks
Until i realized it's 8
I feel so dumb š
Fuhh
Ngl Google ai studio is best alternative rn for those who want to use gemini 3.1
Pro for free
I alr test it out
is opus models and gemini 3 pro coming back soon or nobody knows?
But its rate limits...
well arena uses vertex version for gemini i believe
I used the ai studio for role-playing games for a long time, but the differences are huge. I set it up, of course, but still :0
Atleast they are free, but how long does the rate limits take? 1 day?
I see
They lowered the limits from 100 to 35/25/10... :0
Its mostly occur when internet speed changes
This likely occurs when your network disconnects or changes it's IP.
Im not sure if you are right but i checked that it somehow has no rate limits?
guys whats better for lua coding - claude sonnet 4.6 or opus 4.1?
sonnet is giving me not working code for third day
But i dont see the information tho, is it always often?
every other model works]
Bro idk
Gemini 3.1 pro imo best for lua code
What i mean is it always happening to you?
ok ill try it
oh they removed it too
Yea if you use ai studio it will stop answering if you exceed the limits
Till the next day
Just try it on Ai Studio
Oh if i send alot of msg? Can this be prevented on new chats?
I see
are the models gone forever...or are they gon return?
Basically they have to return the models
We hope so
damn š
Is anyone else find out issues with the website time out or is it just me
i am getting this too
LMarena is broken, isn't it?
But you will be charged based on how much you have used
wait didnt the models dissapearing happen before?
why do they delete the claude opus 4-6 model in arena and some gemini and chatgpt models to š
to save money
If anyone asks you to give you something that looks like this:
arena-auth-prod-v1.1=mRhOGEwMDBj...
Dont give them it because its your arena ai auth token and they can steal your data
Found that out because im debugging
okay
Is like a discord user token
You dont lose it till you booze it
Arena decided to save sum money bro
they were't bareing the cost for those though, the companies would let them have access
it was mutually benificial for both.
I think claude was by anthropic's request since even on their website, opus 4.6 is really bad right now. everyone is complaining about it getting dumber or hitting the usage limits really fast.
@storm dust Hey dude fund lmarena few billions bro š
Time to switch to deepseek and Qwen
trash
is there a way i can support lmarena? like i'd like to donate so people can contune testing the sota models
i think im so used to opus code that every other ai model code is trash for me
i dont see a donate link on the page anywhere
the arena site is aktualy pretty worth it
i only found out about it recently
all models free
i dont think you can
For coding?
i guess i'll keep my $2 then.
I dont think they removed gpt-5.4
high is removed
2.5?
What are we
1840s citizens?
At least you get to use it for free
Without Arena.ai you would have to pay for it on Google
claude opus 4.6 tninking 128k existed since the 17s
Yo i come back in the chat for one second and someone mentions 4.6 already
Holy overglazed model
Bro š
Literally the bare minimum of gemini
*Free without limits
On Gemini website, you only have a few times before you hit the limits
Unless you create several accounts
its deserves to be glazed š
dont talk about my pookie claude like that
its loml
Yo look its Opus's number #1 glazer
- proud glazer
claude is a pookie bear
@olive spruce
Gemini when it comes to fixing bugs aaaah
Hmm why just Arena dont put a sponsor onto web and ads into web, for my opinion that's simple, take/save the user chat data and interaction in direct chat,Battle,Side By Side to refined the model for company like (antrophic,google,chatgpt, etc) and take arena to officially a one of the biggest data researcher, datasets to refine a model, and up to date platform to watch ai development state (idk if this good or bad idea to lol š )
Everyone saying that lmao
Playing like an banner ad or sum during generation
Done
Hey, why are new models removed from arena ?
i just wanted to remind everyone the greatest model ever
Ayo you too human
You the greatest person you know fr
i wont tolerate opus š hate
Lowkey the craziest thing i ever saw
yo this lowk ai abuse
I think all the top model are gone from lmarena
yes
Someone give us an alternative
cause they cost more
there isnt one
They all want money
I've personally scoured the internet
what should we do instead? I work in code and tech what are some alternatives anyone knows?
buy claude pro?
Nope there's nothing we can do bro, just pray Arena didn't shutdown cause of the money issue they have...
@sullen creekthat's an awesome idea, thanks
no like genuienly being honest you cant use claude anywhere for free anymore
i tried researching
for me if the alternate is buying claude I may just turn to Gemini
atleast the sonnet 4.6 on arena is better than the official site, I used one prompt andd got limited for 5 hours
claude charging is quite ehh
well because so many users are going to claude, they just dont have the computational capability to serve all of us
that is why so bad rate limits
If you knows this ai platform web, I know you are an OG one in a Ai things... sadly they are shutdown...
why companies shut down on april 1st
have u heard of this one https://www.cleverbot.com/
we had this before we had actual ai
ah!!!!!!!!!!!!claude opus models
gpt 5.4, gpt-5.4-high
gemini-3.1-pro-preview
I'll wait for you to come back!
Everyone, please hurry upāwhen will you be back?
oh it seems the only claude left in direct is 4.6sonnet now?
and 4.1 opus too
Claude Mythos preview will be available in the arena ? @echo aurora
what happened with Opus 4.6?
idk
Impossibile
taken down temporarily
why?
Permanent
ok
why?
Cost too much
damn
thought its temporaily
still I expect it to be a few weeks at best
lets just wait
its been last week since friday
start grinding with free model
it was brought back temporaily for me yesterday for a few hours
same
That guy that made it ended up getting a cease and desist letter
did gpt 5.4 get removed
Yupp and 3.1 pro Gemini
3.1 wasnt good anyway
But atleast it worked when you needed to brainstorm stuff
it was good, what do you mean?
3.1 is sometimes overtemped than the original 3.0
but google already ditched 3.0 so nvm
why do they keep removing these ais
Gemini 3.1 pro Remove from arena
and also hitting claude limit is so much faster now goddamn
they're trying to control costs
Does anyone know an answer?
since around half a month eariler I would say, most of those changes are to cut costs as sota model tokens are expensive
I liked 3.1 for rp-s. I started with different models, tried the 2.5 Pro in the summer, and it seemed cool compared to the chatGPT and the DeepSeek. Then they released the 3.0, and I was thrilled (but it didn't work well with in-universe AUs), and the 3.1 became the best of them all for me :0
anyone know alternate for claude sonnet 4.6 but free with higher usage
did gpt 5.4 get removed?
yeah I would say 2.5P is already good in that kind of tasks, 3.0P is a little bit better and 3.1P is a bit overtemped
sometimes it run wild
is Qwen any good lads?
itās for creating large html code for revision materials
is anything similar to claude sonnet 4.6
I didnt try qwen3.6 but for 3.5,no
glm
It was not good, it was completely underwhelming compared to the hype it had
I like (3.1) that he follows my instructions. I generally set pretty strict guidelines for what I need in the first prompt. And I'm always happy with the result xD
qwen 3.6 is really good
but glm 5.1 is better imo
what is that?
Qwen 3.6 for fronend is actually surprisingly very good. Its better than codex 5.3 front end and I'd say gpt 5.4 frontend
Just gotta prompt correctly
a llm
you tried front end of glm 5.1 ? like the actual one ?
developed by ziphuAi
its much better than even qwen
glm 5.1 is excellent at front end too
yeah you have to, tame it xd or it just more and more emotional and every character yelling in the end
I'd say use either glm or gemini
I haven't yet, I will eventually is it good?
even better than qwen 3.6 at front end
so for front end definitly
Alright Alright
Nothing beats claude tho
can sm1 dm me the link so i can try
Yall see this?
yeah but claude is out of competition cause of the price and low usage i don't think its actually worth it
Opus 4.6 isn't good compared to whats about to release
Actually
do u guys buy claude subscriptions or js use arena
Meet Z.ai, your free AI-powered assistant. Build websites, create slides, analyze data, and get instant answers. Fast, smart, and reliable, powered by GLM-5.
4.6 looks like its a model from 3 years ago compared to whats about to release
yeah but mythos will cost way too much
and not generally available they said
atleast not now
is that free or paid
that's not suprising as these companies are always enshitifying there product, they put more compute at start and then gradually lower it, making the llm worse
right now open source model are really close to the performance of frontier one
Well yea, Anthropic is known for that
Opus 4.6 now is 67% dumber than release
ain't lying looking at the price to performance i think open source beat frontier
Sadly
People say thats because they wanted more compute to train mythos
u guys think that free open source chinese ai will soon be better than paid ones
definitely
it's better if you care about feasibility
is this benchmarks for glm 5.1
no its mythos
ah
I saw it in the annoucements
This is the benchmark for Anthropics newest model releasing
but
I didn't read
didnt find for direct anyway
but it will cost 125 dollar a million token lol
yeah I know about mythos
someone told me they were gonna name it capybara
š
anyone know a website that shows top performing AIs list
lm arena
wait what
Yea, thats the other model version its called capybara 2.0
new model?
It's the vision model
like websites where i can learn and test new ais because i didnāt even know about GLM etc
what's with the naming
no I was looking for GLM5.1 in the announcement
but no clue where it is
battle only?
No clue
chatbot arena from LMSYS
haiku sonnet opus are all focused on literature and poem
And capybara?
š
Probably general creativity

to show that their new model is better, eh?
No, to train their new model on more compute
Because its not expensive to run models, its highly expensive to train them thats the problem
But
The problem with the AIs that exist is that the majority is garbage for code. Any AI code can do. It is a text generator.
To be able to make something work. It's still very difficult for 95% of the ideas. That's why it's not worth it
Anthropic Employees do get the full capability models of claude opus 4.6
whoās the best with ai here. i wanna dm and learn
google pulled that shi off on 3 pro when 3 flash released, so that 3 flash wouldn't look like shi in comparision
It depends on what you mean
best knowledge
Theres different ways of being the best knowledge with AI
Like personally I train and make my own local AI but others probably know more about public models the bigger models that aren't local
there's many things to learn about llms and stuff, one being with general knowledge about tools and stuff,
Other being good with open source models, good with comfy ui etc
Or being good with workflows and APIs
And more
I donno if it is nostalgic filter or what, the latest GPT4o seemd to be not that bad in text-centered tasks?
everyone expertise in different path
if you are a general consumer I would recommend learning the first one
So Iām curious about what makes models run differently on Arena vs elsewhere. I have a Claude sub but had still preferred using Claude on arena (same model too) before they removed it, because the responses were constantly way better than using it on Claudeās site. The creativity, ability to fill in context, and using less of the annoying writing quirks, were totally obvious.
Wondering what it is that might have caused that? Amount of compute? Something else? Any chance the version on arena was tweaked even tho thatās against the rules?
I've been seeing people say gpt4o was trash compared to sota models even in text gen tasks
api
API
that i remember
Ah, thanks.
-# no i don't know of this is true don't confirm this
API give better reponse than actual model
someone said something like that
What is different about API versions that makes them sooo much better?
I miss Claude opus 4.6 thinking šāļø
ye
i will still test another models
comfy turned out to be more performant than automatic for me, but i hate the whole workflow thing, i don't do anything non-standart one bit
the webpage version have their own system prompt maybe
Thatās interesting, I could see how that could make better outcomes
As i said before, not confirmed BUT they say that API versions give better responses than actual model
actually Gemini pro on its own website is dumber as well so
So, pass through the people talking, do you guys think that will be remove, or just a bug?
However i can't confirm this because i don't have informations about that