#general
1 messages · Page 295 of 1
<@&1349916362595635286> uhhhhhhhhhhh
just look up "prism ai llama.cpp", you can download it from their repo
if the links on top don't work, use the release links on the bottom
u should work for lmarena so they finally start fixing bugs rather than make features with 50 more bugs
Quick question, while GPT 5.2 search is the best GPT for verifying things, what's the best regular GPT 5 version for writing an article for research?
Yupp is going down? is it true or just april fool?
I genuinely hate this update like I can genuinely use almost always only gemini-2.5-pro cuz it doesn't burn due my entire damn chat limitation in a few pumps 😭
Is max working with super grok?
WTF, I is has hit and remain faced token limit, WTF. must fix and solution please, admin must listen me, but admin is like dont listen me
Try Temporal's open-source orchestration platform to ship AI features 2x faster - https://temporal.ai
Anthropic accidentally leaked Claude Code's source code to the entire internet. Let's take a look...
#coding #programming #claude #ai
🔖 Topics Covered
- Claude source code leak
- What's in the leak?
- Unreleased features
- Undercover Mode
...
Lmao
This is the 💩I’m talking about all the time lol
I am building one app and it by default start working in battle mode? can i change the mode? and also what i see it says it has added the feature but in preview it does not show those added feature
@quasi atlas if you may
<@&1349916362595635286>
So, this guy was hacked
a couple friends of mine got hacked as well and sent the exact same message
whats funny is it's pretty easy to spot the scam
Does anyone know the max model in direct has limitation?
You're absolutely right
are you lost buddy-
Hi
hi
hi
?
car ai yt automation
so AI channel content
yes
you could look around idk
what do you do
I'm personally against AI slop content
like REALLY against
thats if the content is slop rii
what are you talking about, literally any AI content is AI slop
it's the definition of "AI slop"
I can't understand what you're saying
oh
I do programming idk
thats sharp
sharp?
cool
see I didn't say I use AI to code, usually I code on my own, only times I do use ai is for experiments in languages I dont yet know enough of
all my projects you will see none are ai
I’m just saying in general
I agree with you. It’s all slop.
But I don’t know I like it
I don’t mind though it’s cheap and quick
Bang bang and you knocked out you know
It’s a time saver
other times I use AI is when I wanna set something up and idk how to
We had slop way before even AI
that is true
Go to https://surfshark.com/goodwork or use code GOODWORK at checkout to get 4 extra months of Surfshark VPN.
A riveting, jaw-dropping, no-holds-barred investigation into why corporate America eats slop bowls.
JOIN GOOD WORK'S PATREON FOR LIVE SHOWS AND MORE: https://www.patreon.com/cw/GoodWork
Featuring Restaurant Business reporter Lisa Jenn...
Look at the music industry before AI
a
All the radio music is pretty much slop lol
crazy
Ik
I have admin panel that shows everyone who logged in
Pineapple, are you "testingacount"?
so u can dox us
fantastic
It doesn't use email
It's js simple username and pass
No address in anything
I doubt I can trust that
I can't see location, only the account made in the web, I can ban and stuff
Guys I have a question, when I’m out of token on one chat, is there a way for me to possibly get all those content of the chat in the form of a summary ?
once more I repeat
I still cant confirm it so I cant trust it
kk
nope sorry
unless u dont have an account
then chats are stored in cache I think idk
@quasi atlas I believe this one needs to be removed since either spam or scam or shameless self-promotion, no? #general message
Aight bro
best u ask like a friend or smthn
Thanks, have taken care of.
uhm it's still there is it some webhook thing
alot ppl already tested soo
I thought that was until I saw the fake blank pfp
Thanks, very sneaky user
They were also trying to promp Video Arena, so a bit
damn
Yo btw tell me ur username in Shade I'll give u infinite dredius
did not know that wasnt allowed-
I tired a generation and wasn't able to get it to work
Oh yea
That's a bug
We send a warning first, and if they continue (which happens a lot) we ban
I see
@echo aurora can u check conversation ids and figure out why they're stuck
I switched models to like claude opus and then messaged so I dont get captchas again and now it's stuck
well imma force unstuck the chat whether u like it or not :D.
oh hell yea it works
yea I'm turning this into an extension as a fix for ur users since u guys are too busy making broken features :D.
Yo i fixed it u can try again
@echo aurora Did lmarena just reduced token limit again
I can't even use my previous chat pdf and paste it there to continue my chjat anymore
It hits max token in 2-3 prompts
oh hell nah
It's actually not
there's a misconception going on that sora 2 got shut down but it's the Sors social media platform that was shut off along with the api it gave. The model still works btw
Not like they closed it like gpt 4o
So how will I use it?
hey @echo aurora any update on the generation limit or anything like that? any plans to increase/lower the limit?
uhm why is claude opus 4.6 not found in lmarena?
It is there
@light sleet i dont see it by searching it
They are ending api in sept
i dont see claude opus 4.6 on the model lost [screenshots for further evidence that its not there]
it only sometimes appears dude, you gotta learn to not compare your client to others
see whats funny, I only have the 4.6 models not 4.5 or other models, it tends to happen
dont know of a fix as of now
What was I comparing?
I just said it is there
If i said "it is there, stop lying" that would mean I'm comparing
are you slow?
tfym "compare" twin js said "its there" only lmao 🥀 😭
and how was he supposed to know if theres a bug about it
so basically
youre lit the one who needs to learn smt 🙏🏻
learn to not reply to smt that doesnt include you💔
you wanna stay out of this
you said "it's there" so the assumption was you assumed it's there for you so it must be for them
thats all
aint u not doing that too
no I'm here being reasonable you're here tryna start a fight by ridiculing someone
- why do u care if they do anything
you deliberately tried to get me mad
and youre getting mad
that is a you problem isnt it
cuz
I wasnt trying
you have to learn not to jump into people's convos
now that's straight up ragebait bro 💔
I jumped into this cuz you WERE the one to start a fight between the vince guy
not my fault youre getting ragebaited 💔
11 minutes
thats more acceptable
great sure
- he didnt mention anything about it must be for them so PLEASE stop telling people to learn smt that they didnt do 🥀
im js tryna explain
not "ragebait" u
Guys what is happening
this guy's arguing with me for arguing with u ig idk
so you confirmed it
you were arguing with him bcz he said it's there
lol
isnt it the truth that I was arguing
and he didnt even reply to that
what does that prove
💀
the fact you wanted to argue over someone saying "it's there" on their perspective
like are you dense
now that's what we call u getting ragebaited
cuz u js said u were arguing with the vince guy cuz he said it's there
that's CRAZY
I even justified the argument, they said "it's there" instead for smthn like "It might be a bug or smthn" they were somewhat implying it's there for them so it would be for them too, and then I clarified that was the assumption
you're here tryna argue over nothing, no logic just pure ragebait no justification whatsoever
so what if he said its there
is that your problem or his?
it's his isnt it
is it a crime me calling it out
no it's not
end of discussion
If @analog aspen didnt get mad about it that's your fault
Yep
It kinda is cuz ur jumping into conversations that doesnt include u so u learn before saying anything.
what is there to learn I acted on pure human logic, and I know the context
And he said it's there and you told him to learn lol
what is there to learn for him
im got reach to chat limit of the converstation but when im refresh is back to issue infnite genration
starting to learn how to speak better?
that's his teacher's problem not yours
Didn't we agree to end of discussion?
hey dude u gotta like unlock the chat the chat gets stuck and rather than fixing it the devs make new broken features so like, here's a console command u can use. Press f12 on the website and go to console then paste this in the console
(function() {
'use strict';
console.log("Attempting to force unlock Arena.ai UI...");
// 1. UNLOCK THE INPUT BOX (Textarea)
const inputBox = document.querySelector('textarea');
if (inputBox) {
inputBox.removeAttribute('disabled');
inputBox.placeholder = "Force Unlocked by Script";
inputBox.style.border = "2px solid #00ff00"; // Green border to show it worked
console.log("Input box unlocked.");
} else {
console.log("Input box not found.");
}
// 2. UNLOCK THE SEND BUTTON
// We target all buttons that are currently disabled to be safe
const disabledButtons = document.querySelectorAll('button:disabled, button[disabled]');
disabledButtons.forEach(btn => {
// Remove the HTML disabled attribute
btn.removeAttribute('disabled');
// Remove the Tailwind CSS lock (The "Juice")
// These are the classes that prevent clicks and make it look grey
btn.classList.remove('pointer-events-none');
btn.classList.remove('cursor-not-allowed');
btn.classList.remove('opacity-50');
// Add a visual indicator
btn.style.cursor = "pointer";
btn.style.opacity = "1";
btn.style.border = "1px solid red"; // Red border to identify the unlocked button
console.log("Button unlocked:", btn);
});
alert("UI Unlocked! Check for the Input box with a green border and Buttons with red borders.");
})();
Imma turn this into an extension soon
I cant do this in mobile right
there are special browsers with dev tools on mobile u might be able to
Such as Kiwi Browser
ig so
Will it be available on Chrome Web Store?
I'm not sure yet I needa check like the terms and services of arena
Kk
code arena generated a website called shade that's like code arena so shade generated this 😭 😭
damn
btw I named it amirs rides so it would represent that Uber Driver meme 💀
thats insane
Ikr
Ima go to shade and ask it to generate a website that can build websites 💔
Family Tree ahh
I see
is work really thank you man 😄
np dude, yea it's pretty cool
has anyone found anything better than perplexity for deep search? grok seems to be nerfed.
gemini search isnt so bad
it's ok, but I find perplexity is more through
I see, I dont know enough about that u can ask others
thanks for the input!
but can you suggest any app have dev tool like chorme in mobile?
I had this app for android called "browser dev tools"
I think it exists for android
or just search those keywords
oh im will try download later
kk
your code really work
It's not my code, but it does work yea
If I want to ask for feature wher Edo I go
feedback ig
hi guys i have a question
i am trying to make a video and i asked to make it portrait video but its not working
explain how it's not working
is it stuck
did it throw an error
no no just keep generate regular video landscape
dude be clearer, how exactly is it not working
what am I supposed to understand from "keep generate regular video landscape"
is it generating forever?
no
every time i ask to generate me a video for example a cat dancing
and i want the video in portrait
it generate me a new video but the resolution is landscape
everything is fine i just want a portrait resolution
I dont think, unless there are options to make it portrait, that it would work
like I think the AI needs to support such options
I doubt it can be fixed otherwise
ok thank you
another one got hacked...
<@&1349916362595635286> might wanna kick that guy
oh wow
thanks
You guys wanna know how dirty the industry is I’ve noticed all along, but there’s finally proof
Man people used to say this was crazy nonsense
Neither does Gemini or ChatGPT do this also
Hella deceptive lmao
i was waiting for bro to mention this specific part of fireships video
I’ve been saying this for years and people used to laugh
Because the models never mentioned themselves sometimes when you refer to them, it’s the oddest thing
They will say something like AI or Llm
cration dede
💀
using codearena for non-webapps is actually a w over text arena
it can actually do something useful because it can search web properly and manage files better than the other 2 arenas
Insane performance from Qwen, gotta give it to them
That birdie is a liar
what?
never tried qwen plus
still unable to log into my Google account on Arena and it has been 2 months now. 🤦♂️
Sorry to say I won't be able to share details about upcoming plans, and when to expect them.
Qwen cooked
Are you able to share the Eval ID for the chat session this is happening with?
what does this preview do?
Are you getting the issue where you try to login, and then you're rerouted back to the home page still logged out?
yup
The model is called Preview, its released
i know but what does qwen 3.6 plus do
What do you mean what does it do, output text
it does Plus things.
what about coding?
plus plus
ok
wait...what????!!
yeah... I pretty much gave up on using Arena
what exactly is happening in arena.ai lol? claude opus still send responses despite "Something went wrong" and "rate limit" error, lol
I think their webmasters are just vibecoding at this point.
openrouter for example
what about lmarena?
oh
amazing... people are still trying to use the AI in here.
but its free on openrouter to my knowledge
Midjourney pretty much ruined it for all of us.
Can you fill out this form and ping me when done?
what would be the point? Didn't you already spoke the devs about it 2 months ago?
This strange, what happens if you hard refresh the page?
I don't blame ya for trying to help but I've accepted the fact that they can't fix it.
So I can bump the issue with your account specifically
yeah, didn't you do that before?
I can do it again
do you have a different team or something?
yeah... I've already moved on.
I pretty much just lurk here for leaderboard updates
No, same team. But bumping can help.
That's a shame to hear. I wouldn't want to pressure you into something you don't want to do, but I do want to try and get this account issue resolved.
why do yall complain about something thats free
When things aren't working as they should, that's a valid thing to be upset about.
you are sure it doesnt run at your local machine right? because it sometimes feels like so plus along with sometimes it works if it wants to work
hey I have a question. Are models on Arena legit? If so how are they providing it for free?
Companies fund arena to show their models off ig
Like Gemini is really top on the leaderboard and Google likes that
sd2 out nice
this is the only thing i don't like in arena coding:
they can't show properly their skills if they have to reduce the complexity
due to the memory limit
it also change leaderboard
imagine a very good model but everytime someone ask it to do something it hit the memory limit and then give you a bad result
it'll be low in the leaderboard compared to model that do smaller work
gemma 4??
Gemma actually outstanding for its size
Gemma 4 just dropped today
Gemma are googles opensource models
lmao what are these results
it gonna be same as gemini or gemma 4 is idk?
(don't blame me cuz i use models that i know)
No Gemma is different from Gemini, it is a different model with by far less parameters
i wonder if google gonna release gemma search?
Gemma is for small devices usually, made for integration on Edge devices
and nothing else?
Not only, Gemma 4 is actually top 3 right now overall among open weights
skibid..
Yeah they're the models provided via their APIs. We have a product for AI evaluations, you can learn more here: https://arena.ai/blog/ai-evaluations/
Multimodal reasoning
Develop applications with strong audio and visual understanding, for rich multimodal support.
just to be reminded, qwen makes their models opensource, right?
Honestly insane, I hope google can finally start improving their coding
where is qwen 3.6 plus on text arena 🥲
Dead.
im p sure you can check for yourself
Gemma4!? 🤯🎉
i take it you don't like it much? 😄
honestly i kind of like it
but if you want speed just.. use step 3.5 flash (free)
looking for accuracy/intelligence > speed
wow qwen3.6-plus beat glm-5
but apparently it's proprietary
so booo 😄
if it's proprietary, it's way behind the others
wait qwen code exists??
except maybe mistrl 😄
i tried it yesterday, no qwen3.6 🙁
at least not free 😄
and prior to 3.6, it was the worse of the 4 coding harnesses i've used (qwen-code with qwen3.5 coder vs opencode with Big Pickle, vs claude code, vs codex)
Qwen 3.6 Plus builds on a hybrid architecture that combines efficient linear attention with sparse mixture-of-experts routing, enabling strong scalability and high-performance inference. $0 per million input tokens, $0 per million output tokens. 1,000,000 token context window, maximum output of 65,536 tokens.
yeah, which is why i'm disapopinted qwen-code doesn't have it
are they even trying?
when is pineapple adding gpt o1 pro 🎉 😊
be patient or smth
isnt o1 like 2 years old
maybe they didnt have time?
yes
and really exmepsnive
idk why i even said that
Does OpenAI even serve that model anymore
isnt its output like 1m per 600 dollars
hope so, but it's a shame it's been on openrouter for a few days (afaik) and not their own harness
not sure
its in openrouter tho
GPT 4 COSTS 60 DOLLARS PER 1M OUTPUT??
yall I finally made the extension to fix most of arena's bugs
hi
Tell us more
it fixes
- Stuck Chats
- Copy Button
- Stuck Captcha
- Skip Button disappearance (not yet tested but should work)
the 4 main bugs that has been bothering me personally
what
hi
ikr
you made that using arena?
nice
thanks
depends if arena would let me
@echo aurora would arena let me
copy button also gets bugged too?
wdym
no thats separate
idk
it's the annoying copy button being on the side of user messages
it instead changes it to be at the bottom of user messages
@echo aurora yall should hire me ngl
I'm not sure. I'd need to check with the team. Where can I learn more about this?
about the browser extension?
Yeah
in my dms :D.
My DMs are turned off, can you send one to @oak python
I have got a little problem, even if I finish the Captcha in the required time, it tells me that : Trace ID : b4cffe1b-1a8b
pls
I believe this is a known bug, will check the ID and report to the team if not. 
thanks
Yo pineapple I fixed the bug in Shade u can try now and also, can u gimme your username in Shade
how fix ?
Video Arena currently doesn't support Direct Chat and Side By Side.
Only battle mode.
No problem
@echo aurora
the website with which I created it : designarena.ai
@echo aurora ur sure anyone checks the modmail thing
it's just a web app
Yes, just working on a few different things right now, will get to your message when I cna.
Thanks, will keep you updated
is there a way to buy unlimited use of nano bannana pro on arena
Would you mind posting your question in #ask-here ?
kk
i did sorry im impatient asf
No problem
Can u gimme ur username in it
I'll give u credits to try it out
do you know if this is a thing because i cant find a pricing page or is arena for ai research purposes
I'll keep you updated when I have more free time 
We don't have a subscription to gain access to better rate limits
Ig no
Kk, also u can reply with yes or no or js a number, Are u one of them?
keep it that way 🙏 !!!
i am very thankfull to ur services
yo @echo aurora if i provide trace id of my chat can you fix the token error?
he cant fix it pretty sure he can just tell u what it is
but I can fix it
:D.
If it's for the Session Token Limit error message, that's going to be working as intended.
were your servers crashing before this? i'd say everything was going fine.
If it's the Something went wrong error message, then yes sharing that Trace ID in #1417174113092374689 would be helpful. With these I'm able to either tell you what the problem is, or file a bug ticket with our team for fixing.
can you see the chat if i send it to you
I'd ask to not advertise your extension until I give you the okay. It's unclear to me if it's ToU breaking or not.
fine :(.
just curious
I'm just looking at the reasons why the generation errored out.
The program I use doesn't show me specific prompts, just overall information on what is causing the error.
i mean i guess the messages the ai gives is pretty long.
i think your 'session limit' should be more of a message count based. not on tokens otherwise this is why people are complaining
by chance did ur ai stop generating after tryna generate long amounts of text
no. it just didnt let me prompt next time
Why the heck didn't arena test the smaller gemma models? Petition for arena to stop ignoring smaller models. It's not always about the trillion parameter overexpensive models
is it stuck generating forever
We want small models listed
it used to be after i refreshed it during the error but now i stopped it and the chat isnt bricked anymore
right now you keep getting errors?
the token limit error
what model
i think they were already doing fine before this
claude opus/sonnet 4.6
no i mean the chat session limit
yea that too
u cant prompt it anymore
it ends way too soon
I have a half-solution: create a new chat and send a message. When the “stop generating” button appears, go back to the old chat (the button will show up there to stop it). After stopping the old chat, try again...
is it good now or, like whats the current status
people are still complaining
I wonder whether it holds for claude family of 4.5. I didn't notice a lot of difference between these once 4.6 rolled out 🙂
they did nothing
eh, that does not happen for me also he's past the bricked chat issue I think
oh yea they do that they tend to publish broken features before making fixes to existing ones
it's just what they do :D.
Bro? Even I who don't do code I noticed a brutal difference in intelligence between 4.5 and 4.6
I haven't specified which areas interested me and what I was doing with 4.5. Plus, again, I just wondered whether the difference in access is significant, if present at all. 4.6 did not exist until a few months ago, and somehow you all coped, right 😄
this is why I made my extension, which I needa wait for pineapple man to approve 😔
but if i were to provide the trace id could you or the team see what the chat is about?
ooh what is it about
Oh
it fixes a ton of bugs :D.
life is soo much brighter FINALLY being able to use arena.ai :D.
how do you use it?
We are interested in including a wide range of models. If you have any specific ones you'd like to see, adding to #1372229840131985540 would be ideal.
I'll explain once pineapple man approves it 😔
is it a chrome extension or a different site?
how good is glm 5v turbo
Will you post it on GitHub if it's released?
probably chrome webstore
Ok
idk
who else thinks claude's rate limits are terrible?
I mean, if only they had a choice
sonnet and haiku are still usable tho
i thought claude was thriving
nah the problem is actually sonnet
wat
i cant have sonnet generating a long paragraph without telling me to wait for 5 hours to generate another one
btw liam tell me if there any news about the extension
awaiting approval from mister pineapple man :D.
uhh
sorry i didnt understand what you meant
i can only messsage it once before it tells me to wait for 5 hours
Honestly for me it looks like it usually just goes to 30 minutes every like five prompt
Which model are you using?
just why did they add this session limit? nothing wrong was going for them
Have you ever considered they may be millions of you junkies now instead of, like, hundreds of thousands a few months ago, lol
Fair enough even if I personally have more issues with the context token limitation changes then time
It’s always been there, they just made it more precise now... and from what I understand, some of those 'something went wrong' errors were actually token limits, but it just didn't say so before. But it's true that chats have shorter session limits now
junkies 😢
Shorter is a vast understatement but okay
i get the error 500 wt
Has anyone noticed that Gemini-3.1-pro-preview is getting dumber and dumber, and that they've increased the token limit per session?
📉
Just look at the popularity of models on huggingface. The qwen 3.5 9b has 70% more downloads tham the 27b version, yet it's not listed on arena. You don't need polls to figure out what should be added next, just look at the numbers
Smaller models are far more cost effective. That is why they are often more useful than big bulky slow "frontier" models
You should simply look at huggingface to see what models people use most, and add those. Rather than go by polls in a sub-channel of an obscure discord with 7 total votes
Yeah that's a good idea. I think it's important we still hear from the community for what they're looking for. But yeah I do like your idea a lot.
NO WAY 🎉
WAY 🥳
are u saying that my claude opus 4.6 romance chats are public now
The team deserves the thanks 
For the ecosystem
The community has come up with many ideas, but it seems that only 2% of them are actually implemented
There are many great ideas that could improve the platform, but right now the priority is to implement restrictions/limits than community's ideas
We have a very capable team that’s thoughtful about how we prioritize our work. Our goal is to measure and advance the frontier of AI for real-world use, and we believe that brings meaningful value to our users. We understand that not everyone will agree with every decision we make, or the timelines it takes for these changes to be live.


hi
Is the website built or assisted by ai or do you not know, also do you think ai could take over website building?
opuses are vaporized?
Opus just went poof
it's an experiment where some users don't have access to certain models (seems to be just opus for now) to test rate limits and gating models
bruh
Honestly hasn't happened to me y at least I haven't notice when it does
I just noticed it suddenly used Max
They are removing the moudle that you use emost from the user that cost them the most
That why when you enter another account it's there bec you don't use it as offen
They might make it up ban im future so no account switching will help
That my theory
They are trying to cut costs as I think
honestly the whole experiment system is stupid, we should be able to opt in and out of them. they are treating users like guinea pigs at this point, even though the whole point of arena is to treat people like guinea pigs, using their data to evaluate and train models (and sometimes releasing datasets to the community, also side note: arena please do another battle mode dataset release, i would like to see an image gen one as well, maybe make it a bit bigger. i later want to try to fine tune a small local model on part of one of your previous datasets to see if it improves the model)
You already sold you soul and data when you clicked I accept terms and service
And they offer paid moudle for free
why experiments is active on main site?
If you don't want your data being shared well it's best to pay £100,000 to get a supercomputer then host Kimi K2.5
to keep everything free their has to be some drawbacks, way better than paying months for premium AI models still
"free" doesn't exist
because not many people are using beta arena so I guess the best way to get feedback is to do experiments on unsuspecting people
You pay with your data
Google search etc harvests data and uses it for advertisements, however they don't sell it which ig is a good thing
And Microsoft but I believe they sell it
Officially anyway
Well no your data is more valuable to them then selling it as they are an advertisement company
It still counts as Financial free to be fair even if it's not actually free in the broader time
I do understand, I used to use arena a lot and to be fair very little data gets released, and it's not much different from normal chatbots like chatgpt where your data is automatically shared for training
No company don't sell your data even games at this point the internet is cooked
True
Lmao every company uses your data
Most sell it
All
It's in there tos
Nope not all
these days data protection is kind of a joke anyway, like even childrens games wanting your face and ID 😭
They won't be sitting on free profit and not take it
Google doesn't sell your data as they value it more as giving it to competitors will put them at a disadvantage
No offense I honestly don't have much trust in companies after all even if you somehow on Earth trust them t for whatever reason be real I think we all know there is going to be probably some idiot who accidentally makes it get stealed anyway
For there advertisements
What's better than using millions of faces to train ai
Yup
Companies are evil
Just some less then others
it is part of using free ai services (and in fact most internet services, free or paid), you are paying with your data and at this point we have to live with it
Gine
Or you try to break the system
Poison your data etc
Removed sadly
Anthropic ain't shutdown lmao
Ohh you mean the model on arena.ai?
Yes
I thought you meant the company lol
I honestly almost got a heart attack when I was about to get the infinite generation two chat in a row almost at the exact same prompt 😭
there’s no opus
They lose money anyway
all opus is gone
for me its here?
What is the difference between those six and five one anyway
Newer models with better data
I already tried it still here for me even hard reset
still their
in my opinion anthropic is the least likely of the 3 major us ai-only companies (anthropic, openai, and xai) to go bankrupt due to spending nothing on things like image and video gen and most people that use claude pay for it, mainly for coding, and their api costs are expensive (google and microsoft don't count as they have other things that subsidize their ai costs)
I think you just got hit with a random experiment
And of course for some reason that chat still made it go into rate limit despite not generating anything 🥲
I agree
which is also btw why arena pulled opus first, it is the most expensive model they offer ($25 per million output tokens for 4.5 and 4.6, $75 per million for 4.0 and 4.1)
Lmao 75 per million???
Yeah Precisely
opus said bye bye
Even gpt 5.4 high is 1.5x ~ cheaper
is claude rly gone for some ppl, check this?
yeah, they must've greatly reduced the price to get people to use it, i expect claude 5.0 mythos to be the same or similar price
Np is there
Press F for Opus
Old Models just were Expensive
Newer Models are More Efficient
is anybody gonna ACTUALLY tell me how to get opus
Isn't qwen like pennies per million?
?
try a diffrent web browser
It's gone till further notice.
Use different account buts against rules I warned you
They removed it for being too expensive I believe
@echo aurora can you confirm this?
claude opus is 100% here im using it rn
gpt pro models are more expensive than old opus, which is why arena never had them
I say they add ads and remove limits
some people have it, some don't
Thing is arena.ai isn't profitable
What are the rate limits btw?
5
And they will likely be sued if they try to profit without licenses
How??
Per hour?
yeah
💀
Exactly I say they ads for evry message that will genrate tons of money
They have 5m users
the fact there’s better models then claude opus
claude opus isnt for general chats though, for me, its for 5 very well thought awnsers
Yup
they live off of their vc money and the fact major labs are paying them to have their models evaluated on arena, yupp died because they had much less vc money and less major models tested (and the fact they gave money away to users who used it enough and cashed out their credits)
why do i have the feeling that half the people not having claude opus is a part of an experiment 😭
GPT 9? Gemini 15? Deepseek 1922?
region maybe???? try vpn to the UK
you may get opus if you make another account, but this may not work so you may have to make a few accounts
....... I hate this I genuine hate this two chats in a row
bro chill out
it is also gone for me too
is your reigion in the UK???
no?
let me try my arena accounts rq, haven't used arena recently
my theory is that some places maybe dont have access to it now, im probbaly wrong but id try a VPN
Of course of course context had to make this to happen when I reset my browser
reasons why i’m NOT using claude sonnet- to restricted- thinks it’s smarter then you- littearly refuses to make anything accusing you of murder
Test Qwen 3.6 Plus
😮
that would be cool 😭
my main account doesn't have opus, my alt does
That's Odd
Why do I have opus 4.7
oh it is accounts
nada
Might be account base or rolling out cuz it's not appearing on mine
bro's trying to trick us
Won't be surprised
Bro I just tricked 2 people 😭
im not worthy of the power
Y'all are so gullible
I mean does it really count as surprising when you just casually look don't really think about it i mean if you got me to look at it sure but honestly not really much reaction lol
Orginal btw
no
Mee
lowkey need it frfr
i thought it was just inspect element 😭
Nuh uh
Lmao that could of worked
claude opus 5 will be the real kicker
Asking ai was easier though
Mythos comes before it
honestly they may never make an opus 4.7, they could but mythos 5.0 is coming in a few weeks after cybersecurity and safety testing and opus 5.0 should come not long after that
one thing i like about new strong claude models is they dont release temporalily good, and nerf to massive levels later on
mythos will come in 2030
Opus 5 should be cheaper
claude's brother
I honestly hate how communication tight they are when it comes to development because God damn they honestly should have come up with something replace this this year when it comes to tokens dude they have like three suggestions already don't they not from the community
it is a model that got some stuff about it accidentally leaked a week ago, it is a new tier above opus and supposedly it is very powerful, so powerful it is raising new concerns internally about its safety (anthropic's goal is to make safe ai products (safe as in safe from nsfw content and safe as in that it won't harm people))
Taking Opus away showed me how dependant I've become on this model. 🙂
The Rumours say is about 500$-1000$ for Million Tokens
Mythos is very compute hungry
My theory is claude is on Compute Shortage because of its testing
Didn't they lose some data centers in the middle east already?
that might contribute to the compute shortage
😯
forget sonnet 5, check out claude opus 6 💪
👍
lmao
i'm sorry I just had to stop myself from laughing when I read that since I was just doing a chat how about role-Playing / making of a country in the Middle East on claude lmao]
What are sonnet rate limits btw?
I think they must be higher than opus, at least 1.5x for sure
slop experiment
That would be about 8 Prompts per hour
Will we get the claude 4.6 opus back, or not?
I again weird for me they still exist
Is not gone
log into different account
hmm...
What is the even the difference between thinking and Ultra thinking?
I hate and love how that explains everything and nothing at the same time
once you press that model your wallet is done you dont even have to chat for your wallet to get empty...
God damn took way too much tries to get hate to register properly on speak to text
lol
So we're getting opus back eventually right....right?
📉
If they add it to Arena, it would only be available to users who signed up with an email address, but those who signed up early using a Google account won't have access to it
Does arena support api keys?
No
it will use a million tokens for a simple hello, don't worry though the tps is very high due to using google tpus
hopefully they get enough feedback saying that nobody wants or likes having opus taken away because if they take opus away, what's next? nano banana pro? gpt image 1.5?
this experiment allows arena to take away any model from any (or all) users for any (or no) reason
wan 2.7 just dropped
im crine, they taking away models😭
📉
seems to be image gen only for now, it's pretty impressive though (will post this in #ai-creations too)
yessir wonder how their video gen gonna be
kinda silly not even 2.5 is open source yet
hopefully one of them will be, their most recent event they had for qwen mentioned having more open source wan models in the future
honestly thats the only good thing they are at otherwise nobody choosing their models over sora veo or kling etc
yeah ltx and hunyuan are giving pretty tough competition to wan for open source currently, alibaba needs to strike back or wan will lose its niche, they are only winning in terms of small local video models but kandinsky is competing pretty hard there as well
i don't use that reddit account much anymore lol except for browsing (i saw that btw, i have message logger)
hi! Video creation is available on the website now: https://arena.ai/video
@echo aurora
PINEAPPLE.
WHERE THE HELL DID THE CLAUDE OPUS FAMALY GO. ITS JUST SONET AND THAT ONE HAIKU MODEL NOW
Here's your answer: #1487734609851449374 message
why is it always claude opus
Hey @heavy blaze I'll be sharing an announcement in a bit with more information.
btw guys i did make a feedback thread here (https://discord.com/channels/1340554757349179412/1489394191435042996) for arena
If you want to use Opus, you'll need to use an account created with email
Again weird for me since I made both account by Google not emai l i can still see it
Sonnet only
that makes no sense, i have 2 accounts made with google oauth and one has it and the other doesn't
Can still see it
lol
Anyway hope you all are doing good
seems to be just opus for now, no other models in any modality are affected
@shy mason what did i miss?
what do you do when a long response gets interrupted by - Something went wrong with this response, please try again.
is there any way to tell it to continue from where it left off or is it al lost? it was generating a bunch of files and I dont want to lose the woerk
I have the same problem T_T
Whenever you come across the Something went wrong error message I’d recommend the steps outlined in [this message](#1417174113092374689 message). Since you mentioned a long response this makes me think it's generating/thinking for more than ~10 minutes, which is what the cap is currently at. Sorry to say there isn't a way for the model to pickup where it left off. I'd recommend trying to reduce the expected response of the output.
We are working on expanding this timeout limit. I wouldn't be able to share when to expect it, but we are putting effort into it. cc @fervent silo
this happen when i have 2 active chat session with same model and yeah i using it too much, don't worry, hard refreshing fixed it.
Nice
Opus has now been completely removed, regardless of the account type
like for all my friend
The funny thing is that the model replied to me 6 minutes ago
Well, time to use kaggle to run AIs locally
At least you could do that I can't cuz Chromebook sucks for that type of stuff 🥲
...
Hi there
The AI runs on a kaggle server instance
I don't mean literally on your local disk
so we still get opus 4.5 and downward?
Yes
There will be more I guess
for gemini 3.1 google ai studio is pretty decent
Correct, it's just the models being mentioned in that announcement for now. I'll be sure to provide further updates.
I predicted like 2 days ago they will start removing models cuz it's too expensive
the hell is Yupp
It makes sense tho I would of done the same
to be fair we have had the most premium models since 2023 on here, thats a long run but it deffinetly cant last forever
Will shut down, they told it.
$
I still have access for some reason
Only thing left in direct will be very small models and battle mode will be more of a priority
Oh wait deadass
it was a site like lmarena, it was a bit different and worse but it was popular too, they just shut it down
clarification on your announcement: are gemini 3.1 flash image (nano banana 2) and 3.1 flash lite going away too?
im geussing its either money, or the models not running reliably. ima hope for the 2nd option
well for poor people arena.ai was probably the only way to run premiere models for free. opus and gpt-5.3 are fantastic. i understand they have to survive, prob anthropic cut them off?
Man I can't no more with all this happening
this was the last enshittification i tolerated. 1.7$ billion evaluation and then this. you were once the cool guys
and it would make sense since CLAUDE RUNS ON A COIN FLIP
i assume it's just pro, but the way you phrased it suggests both of these could go away too
not coming back
It's money these models are too expensive
Ilimited AI won't be real at this moment
damn
wdym not coming back it’s not like yoh paid for it bro 😂
also similar thing with gpt 5.4, are mini and nano going away too?
You can still use Gemini 3.1 for free and sonnet is good too and free in arena
i know, but still
damn
you are always paying and if you dont pay with cash you pay with something else (in this case data)
wtf how
I'm kind of annoyed on to that every time someone mentions about free AI someone else replies about repaying with something else I'm pretty damn sure everyone else is aware like it does not need to be stated again
money cant buy happiness. said some people. they were wrong, in my opnion
@echo aurora are we getting sonnet 4.6 thinking in direct mode? It never came out but 4.5 has thinking
idk guess I got spared
welcome to 1984: you must accept the anti-user changes, you can't say no and you can't disagree
you need money just to live
agreed
Even with money all these comapanies will use your data to train it, they all are american companies.
me too
nono fr, how do you got access
this is a path we are going down, and us people cant do nothing about it.
im not gonna refresh my page
refresh it
there is no happy ending.
just in case if if its still there
deadass I got no clue bro I genuinely think they spared me 😭
April fools was yesterday btw 😂
Bro it's expensive ai models calm down
if i refresh it it might dissapear
i refreshed it
im just pissed theise models have to be so expensive in the first place
It's normal for me.
before i log in, the opus 4.6 was there
are these numbers correct??? because sonnet is more expensive???
Soon there will be $1000 per Million output models
yeah see, its here
plus, you cant even delete your account
For upto 200K.is key
okay its still there
ah yeah
yup
100%
i think its the bug
Tbf I hardly use those models anyway since I always use the web search models
Like even o1 pro was $600 per m out which is crazy
Search models are good ngl
some people dont have the money for it
what really bugs me is every sigle youtuber just assuming you have a $200 max plan. no one ever talks about cheaper options
the only time I use the normal models is really for coding or for like “look at this image”
Jensen confirmed there will be these models soon and I'm assuming Gemini deep think so close to that too
I dont the money also
I made an app from the search model
Android app
Over this past week, my hard drive got fried, we had to euthanize my cat, and now we've lost Claude opus in direct and side by side.
We all don't care about gemini cuz there's aistudio for that.
gpt, opus, gemini all have $200 tier for any decent work, i mean no one uses gemini but still
Ai studio limits for 3.1 pro aren't unlimited
Less you want to use gemini cli or antigravity a lot.
meanwhile these companies are crashing the market and causing massive layoffs
Any good cli ?
I never got to the limit.
yeah
It's not very high I'm guessing around 20 messages a day
Hmmm ig
But its limted
"anti-user changes"
Lol