#general
1 messages · Page 274 of 1
- its a MUCH smaller (and hence) cheaper model
- its a CODEX model, its trained to do whatever works and nothing more than asked for
(which makes it extremally reliable)
You’re telling me the model made for code is made to do the bare minimum
its made for backend
lol
WHAT
if you want front end wait for front end model aka 5.3 noncodex
Ohhhh well see
That changes my expectations a ton
I thought codex was jus a speedier code fine tuned model
Which one is better?
4
7
2
NB 2
/image
its a small and extremally smart and cheap and reliable backend model
I see I see
notice how you asked for landing page
and it did landing page and nothing else
while gemini or claude would do landing page and other pages
which MIGHT seem better - but for front end
in backend, thats absolutely unacceptable move
Oh Claude is just as lazy 😭
Don’t worry
heres from the prompt i gave
It’s still a bit too blurry
(still without reference image but better )
It gives too much like minimalism energy
im on hdr screen thats why
I mean like the text being hidden
the model is a minimalist perfectionist
which is wonderful for backend
yeah (although remember its much older model)
oh and also
I might try Gemini since it strikes me as more knowledgeable
Oh that’s not that bad
(the difference is insane between those)
I’m trying gpt 5.2 high so
heres with reference image
We’ll see how that goes
5.3c
Mmmmm alright I see your point
One day in the future we won’t be able to tell websites apart
It’s all gonna be standardized
Literally
id say
I hate the minimalist style LLMs default to
all websites will look very good
gemini + codex do wonderful if you prompt them properly
Looking good or looking bad comes and goes just like the times
Well, look good 10 years ago already it looks old today
The Internet is gonna become more computer centerd than human centered
@proud bobcat should i try the same prompt on 5.3c xhigh in cli?
show
There’s some issues but wow
It’s still minimalist
But defo better
I’d have to provide it the seed 2.0 html code but this is a decent start
you would need to be more specific really
thats all
Yeah
itll get it if you tell it perfectly what you want
But the the thing is that like
Seed 2.0 completely understood what I wanted
Even if it didn’t look great it had humor and overall roughness that made up for it in authenticity
LLMs are trained to be way too neat
Herd seed was good
it guessed
you cant say the gpt's answer is wrong because you dont really like it as much
if it is correct then its not wrong
gpt doesnt like guessing
Sure seems to love it when making anything else
not really the case from my experience
5.2 made code so bad once I got Kimi K2.5 to check it and it had to do a full ass revival just for the game to START
5.2 somehow made a Minecraft clone that ran worse than a modern AAA game at release
3 mo difference in release date + gpt 5.2 isnt exacly coding model as you can see
but imo it still does decent code
with gpt you should use codex and noncodex together imo
holy im so wasting my quota rn
1117 lines of code 😭
it just said he 25% finished
😭
BRO CHILL
@proud bobcat whatchu think?
With the seed 2.0 code referenced
Doesn’t work on my end
I have
It’s more Liquid Glass but it’s defo a better attempt for sure
I guess the cli just works better
that was xhigh instead of medium
(but no reference image though)
and it was forced to do 1 file not multiple like its used to
no
freaky
claude just wants u to sit on it
like a normal claude being
actually ill try actual frontend cli on that i wonder how itll do
Using Arena for stock advice rn
bro cmon bro
Someone at Amazon defo used AI to make a script
GPT
It’s true people that used to one tool and then they update it and the old one goes away and that’s not what people signed up for in the first place
Same thing happened with custom GPT’s
Lol
All I’m saying is it’s a dirty way to play and I’ll just leave it at that we also saw the same thing with 4o
hi
@proud bobcat gemini's alter to codex one
what does that mean im dumb and tldr
🙏
as in you cannot copyright ai made work?
No
Oh this is cool
73 likes, 2 comments. “AI is literally everywhere, even Burger King now has a AI chatbot named ‘Patty’.”
Wtf
For not saying, please and thank you
Bro I’m telling this is gonna happen to every business bro the AI is gonna track your performance to see how good and efficient of a worker you are
This is a worker jail
Tell me that’s not nuts, dude imagine working in nothing for eight hours a day
Yeah that would never happen with current worker rights
Only people doing it are ekittens
no but i can definitely recognize really bad code
i have a programmer friend as well
have you ever tried out crack bench
ah, he probably knows that craig bench is the best benchmark then
no
he's not very into llms
and no
again, codex code does work
no doubt
however
the code itself is just bad
it has no idea how to properly utilize the tools its given
the design is shallow
nothing like the frutiger aero aesthetic i instructed it to do at all
Where can u get video arena now
Hello! The Video Arena has been removed from the server. This feature is still fully available on our site. Please visit https://arena.ai/video
Guys we need this feature:
https://discord.com/channels/1340554757349179412/1475731095071752354
hello
I hate how random this is at times like why on Earth does at sometimes I can log in perfectly fine and others I can't at all when I log out
im geniunly interested in how lmarena is able to pay for all the requests
Guys why can't I upload images from my phone while I can easily on my computer, it shows error to me, pls help someone 😭
hello
@echo aurora
Is opus 4.6 better than hawk max/ultra
Which AI do you recommend for daily use?
Gemini 3 Flash (Fast+Thinking) or Claude Sonnet 4.6 (Extended Thinking)
🥲 ☠️ i think I may face realize what may be causing this issue
Nope didn't change anything
Hello
I am now working as a senior software engineer
Is there anyone who is looking for an engineer?
/image to video
@teal mason Hello! The Video Arena has been removed from the server. This feature is still fully available on our site. Please visit https://arena.ai/video
Is LMarena down?
No, lmarena is not down
Yeah i think Cloudflare died on my country lmao
can't acess LMarena at least
Cloudflare went down again. And lmarena is hosted there.
Yeah, Discord wasn't working here
how did you @gleaming roost
the the -=_=-=-=
Copy my nickname and ask an AI to edit it to match yours
I understand Cloudflare is performing maintenance. So lmarena and Discord will probably go down again.
cant copy on phone 🥀
Z̳̿͟͞a̳̿͟͞m̳̿͟͞i̳̿͟͞e̳̿͟͞l̳̿͟͞
anthropic is pretty much google's puppet
do you think that
google, like, cares
about laws?
what can a gov do? ban google usage? 🤣
google cant get taken over, if that happens it really is GG
google is untouchable lol
anthropic would be bad but we might be able to overcome it with the other companies and world, but Google AI and infrastructure fully under the US gov is GG for humanity
internet is pretty much google
is there anyway you can use search while uploading an image
google is stronger than the US govt lol
govt has around 30b liquid usd
google has above 120b liquid usd
lol
google has more liquid money
because of PR lmao
My opinion is that Anthropic kind of brought this on themselves for entering a contract with the Pentagon last July in the first place :P
nope
us goverment takes loans
google has infinite money
and that money is worthless
it just causes inflation
bro is as dumb as russian leaders back in ww2 😭
no?
googles earnings would skyrocket while dollar would plummet
google, if they wanted, can literally shutdown 70% mobile phones at will
do you think usa would care about that?
priniting money doesnt make them richer
money is printed so it doesnt really matter tbh, the AI market is based on hype like how tesla was, most of that money will evaporate since intelligence is and will be free at all levels
google can "kill switch" whole USA economy if they wanted
but google should be able to survive that with their infrastructure
you realise you are chatting on an app that wouldnt exist if not google?
your email is google
your browser runs on google software
wait what if the gov takes over Anthropic, are we rebelling??
your phone is likely google's
google wont let it happen
no they cannot
lol
this is entirely false
they can easily
but they wont
because it would waste their money
AI Video
aws and google cloud love eachother
but economy would barely be dented
google doesnt have much of a foothold on hardware
they could but it would hurt they PR
its not worth it
google has more liquid funds than us govt lol
who needs PR when you have gemini that can spread any misinformation they want
and youtube and gmail
google literally owns a world atp
"ohh noooooo no more google chrome whatever will we do?"
the humble firefox
"ohh noooooo no more google pixel!!!"
the humble iphone and samsung
"ohh noooooooooo no more chromebooks!!!"
literally macbooks and windows
google cloud too
samsung runs on google software
lol
so its actually scary the power that Google has, maybe thats why Elon, Sam, and friends made OpenAI lol
which runs on the linux kernel
which can be easily modified and have a distro shipped by samsung within weeks
second time in my entire life i agree with craig
bruhh
what are you on about
google is scary
I need to go touch grass before that right gets taken away by our Robot overlords, its been real yall
nice
if google kill switched anything then there would be at maximum a mess for 3 months before everyone just moved platforms and google lost all their monopoly lmao
"ha! that'll show em!"
"oh wait everyone just migrated and we have no more subscribers"
what are you talking about
literally what are you on about
moved how? 30% of internet is shut down because google hosts it
75% of phones are bricked
The main problem would the lack of Play Store and GMS like location service, which is what happened with Huawei
you dont know how much stuff is owned/related to google
literally every phone besides apple runs on android
why would the phones be bricked if the software can work without internet
AWS would just scale and become the ultimate internet provider
not to mention theres 1 browser rn that doesnt run on google and its firefox
android runs on linux kernel
(from the popular ones)
android is google's
are you special
yeah but its still google's
its based on a modified linux kernel
its open source
if google shut down for some reason the phones would still work
even 3.1 agrees that google could cause ww3
gemini when i show them the humble firefox
gemini when i show them AWS cloud
then i wonder why you have to log into google on every single android phone 🤔
AWS is with google brochacho
amazon and google fund anthropic together
how about this
who do you think
aws doesnt exist solely for anthropic gng
😭
lobbyists issued by companies who push for new laws to be passed
that is how it works
uhhh buddy
that is not at all
google owns search engine
this is schizo tier dude
yeah no dude
"socialism"
Buddy
I hate to break it to you
Right
Now short form is king
youtube short form is by far the most popular one if you actually check the stats
youtube shorts still have more views than tiktok lmao
Views ≠ engagement
All of those are ai garbage videos for kids who can’t comprehend basic math
Actual info is on places like TikTok, Reddit, Facebook
no they wouldnt
Why not
do you think an average person
Literally why not
Literally what would happen
Sure
Outrage
But this would be googles fault
And citizens would be more likely to just switch and completely degoogle their lives
Again I think you are completely overestimating the reaction
A mix of course
Half a year in and no one would care
I guarantee you
i guarantee you that they wont touch google
google has more influence than all other media outlets combined lol
more people watch yt than cnn for news alone
fox corp themself stated that
over 1000x the popularity 💀
Literally
That has genuinely nothing to do with their reach
Google doesn’t care about media
The Fox news station has more reach than Google for gods sake
1000x less reach than youtube alone
if not more
here
How many people do you see using Google news?
Now tell me how many people use Fox News
How many people use Yahoo news
How many people use Facebook for news
How many people get their news from god damn Reddit
Your entire argument is
“But Google big”
Google is big but their priorities are completely different
tbf you can be pro AI while acknowledging AI slop is a thing that exists. Slop is slop, and can be made by anyone in any format, with or without AI.
I happen to love AI slop though
I am very anti video and image generation personally
For example id go as far to say even Grok is more people’s news source than Google
Which is insane to think about
How many people do you see using google search?
how many people do you see using gmail?
classroom?
gemini?
youtube?
schools do
Classroom has literally no effect on media coverage
What are you talking about
It’s a tool to submit assignment through and for schools to communicate
Google search is a compilation of links to sources, not the source
Literally
Does anyone know what model this anonymous model is?
Exactly
Google is not providing the media
Google lets you access links to the news media websites
Is it through Google?
I heard the new Google news channel is great
yes its thru yt
If someone has blocked you then you cannot react
ring camera videos of dogs obliterating people's houses with firehose water from their mouths is humanity's greatest accomplishment to date
cats playing musical instruments on people's front porches at 3 in the morning is a close second
Woah…..
And what’s the channel it’s being hosted on
Google news?!?!?!?!1!1!1!1?1!11?1!
Again people will just use tv
No one uses YouTube for news
hmu for polymarket signals
Bro what, that’s completely false 😭 anyone who watches the news avidly are always stupid legit every single time
Turns out though anthropic won’t have to deal with anything
fr
Reported
What did the quarter say when it accidentally bumped into its long-time friend, the nickel?
"What a lovely COIN-cidence!"
Thank you
What duh quack !
sup
I want ai tool for university assignment which one will best to give word without detected
gemini notebooklm
So, Anthropic seems to have a spine after all.
(they said "No" to the Pentagon's demand)
I wonder if Deepmind would do the same?
Hey it’s chip !
Obviously to keep their ethic brand lol, any big company would do that
Because over 200 members of Open AI and Google AI signed petitions basically saying if this happens then those companies wont work with the government
But heres why this is happening the way it is
AI Is so powerful now it can be seen as a nation threat or national treasure
Is gpt 5.3 projected above or below 4.6
how to generate ai videoes
Dang jailbreaking
do regular models still do search
I used the notebooklm but when I detect that words in zero gpt it show me ai use 60% what should i do
Anyone idea
To give me good humanizer or another ai tool for correct the word give me not casual
I just want to know what they use to jb claude
Spanish languague + some instructions
<@&1349916362595635286>
Does anyone know where to view my archived chats?
The new NANO BANANA 2 is nice
flowith right
or where
Any page that is only used to edit/generate images
Kk
yep
yo i was using lm arena and after a certain number of chat, ive been stuck at a wall where i cant move past the captcha thing
You can follow this steps to fix that
Min 0:38
That would be very nice to use if the website for some damn reason won't let me actually log back in🥲
Yeah I am having same problems
Damn we are at war 😭
Out of curiosity what does it show you all when you try to login for me just gives me a no session exist type of error
The site must be experiencing server issues or something
I probably I wouldn't be surprised since I noticed those type of issues that have to do with login and overall application of the browser have ramped up in frequency recently like a day or so
He has to do something with the captchas especially if you’ve been on for a long time
And are doing repetitive things
It’s a way for it to slow people down from abusing the system
If it is it's a not very good one since I don't think completely stopping people from essentially logging in and making it the server go down or at least look like his s not very efficient😓
Well, I don’t think that’s on purpose. I think they’re just a side effect of preventing bots and stuff
I’m sure if they had a better system or a better way to approach this they would
I mean I figure it's just funny if would be annoying since it seem/ browser this server in overall browser is to random ramp up of glitches cuz like there's always habit ofSome glitch appearing and then almost always escaling for eventually someplace is found if ever
It’s actually a very common Internet business model
Frustration and annoyance until you pay
🤣
There’s a economic term for this. I just forget what it’s called.
They are not even being paid by us pa though so maybe force of habit then lol
Well, they can’t just give it out for free cause then nobody would pay for the services
With unlimited use cause then who would pay for
Man we are really not paying for it that payment at least in traditional way that we give by using it i was referring to by traditionally how we are not paying for it economically though we are with data
Yeah, but what happens when your data becomes useless
And they pull the plug then what do you have to offer?
It’s gonna be one of those things you need them more than they need you
Looking at the number of AI i would say that's going to take a bit before it becomes completely useless and that's not even talking about what happens if doesn't become useless before the economic burst of AI finally hit sit down of its course dby's data is probably never really going to become completely useless with how it has to be used to improve language machines and AI before it can ever really become self-sustainable bags suspect I would take a while relative wise
I hear what you’re saying, but I think that’s a very naïve way to look at it
Business bro, it’s business
Also I do believe that is called Annoyance Economy
Cause all the stuff is super expensive
YOOOO THANKSSSSSS
Nobody’s gonna just do this from the bottom of your heart
ily bro
genuinley
bro why are all these really good paid tools free to use on lmarena
@echo aurora you should pin my video in #1343291835845578853
whats the point
I went and said naive I think just a different way of seeing it as it's not naive as we are paying Rift data which turned the AI is getting its own payment with a more specific browser stuff that's not from this website that it gets the financial stuff
And also why on Earth where did I even implied that?
Personally I think it's one of those things where you can't really predict it specifically you just have to see how it goes since it's one of those things that each week could stabilize but just as easily you could become a economic burst bubble verse just have to wait and see since itself has been successful was like Google ai which ridiculously costly it can honestly only be sustainable for so long if they don't find a method which is possible to be found in time and not in time we just have to wait and see
I can promise you this as soon as it becomes not profitable. It’s shutting down shop.
@undone saffron one last question, can i fix it again using the same steps if the same loop error shows up again?
couldve just typed btw
That is essentially what I was saying ☠️
To be honest with you, it’s cheaper to just pay for this
And not have to deal with the headaches
( if you have the money some of us don't)
will there be a lmarena app in the future
Well, that’s what I’m saying so if we’re not gonna pay for it, then we’re gonna have to deal with the headaches
We can’t have both worlds
And then I will still complain about those headaches it is human nature complain appreciate try and be patient in between it's the cycle of any type of business paid or non-paid
or any service for all that matter
I hear you there’s nothing wrong with it
It’s also important to remember where we stand
The food chain exist for a reason, and we just happen to be at the bottom of that food chain
And again you don't need tell me I'm already aware it's an cycle of practically any business move up food chain sometimes get taken down to Peg something happens people complain try to get good service while at it Etc just moving through life kind of like what I did nowlike where I did now I'm i was complaining about the login but it's not really too heated
hi everyone, do you know a best video model to animate this mock up for showcasing? Image example below.
hi
hi
where i can find video arena menu?
hmu for polymarket signals
what is the best ai for coding (in r0blox studio)
standard or thinking version?
and which coding language?
text
Claude Opus 4.6 or GPT 5.3 Codex
whats the current ratelimit for opus 4.6?
@echo aurora What is thinking model is set Codex 5.3 in Code Arena?
Codex 5.3
mb
real
Hey everyone, is there any other good free ai stuff out there? I love arena.ai but I was just curious if there was other stuff at all to experiment with?
true
Im so confused 😂
🤔
GIYS ANYONE KNOWS WHY THE HELL CODEX 5.3 HASNT BEEN RELEASED ON ARENA
Probably 3-5 messages I don't know
Nothing just normal model I think
@cobalt torrent @sharp swift
For generating images or videos. It's on https://arena.ai
Like what, chat or vibe coding?
More just images
Any codex models are only available to code arena not direct chat
Ah images I don't know to be honest. Maybe try yupp.ai? There is credits it's just like arena
Hey, I’m currently using Arena AI without being logged in and I already have some chat conversations. If I log in now, will it save or restore the old chats I had while I was not logged in?
Ok
But if there's credits then its not exactly free like arena
You can easily get credits by voting on models but yupp offers alot of services
That's a great question. @echo aurora, will this happens?
Hey guys, is there an API for Arena? That would be great if the answer is no. Are there any websites that offer a free API for these powerful models?
wtf is wrong with me for using claude code gui
I keep getting human verify and logout
they moved video from discord to site you should be able to access it via website https://arena.ai/
the last tab in the search bar is gen video
<@&1349916362595635286>
AiStudio give api free, the free limits of flash is good, and you can create various accouts
3 per hour thinking and no thinking
THE FIRST EVER AI EXPERIENCE PARK A space to flow, play, Create and.be - Beyond tasks
I don't know if this is still valid; now it asks for age verification
...
Bro when login will be fixed
hello
I think I can help you to make it really really better. It can spill many things even emergent behaviors also. Because am a researcher in AI and I know some secrets that nobody knows, not even the best researchers or companies
Look it spill it out a smart script for meta reasoning.
@echo aurora Uploading images results in an error in the image models.
Use these prompts . It will gonna help also
@echo aurora these are the system errors to be fixed . Am 100% sure. Read all the chat carefully.
Really guys . Like really. No name 😕.....
Interesting
What makes you think the best researches or companies do not know though?
Man what the hell is happening with the website... My prompts disappear all the time, and the send button doesn't work half of the time...
Stop deleting my chats ffs...
The name is shown after voting.
So that votes aren't biased.
Nope.
gpt 3.5 turbo
You can go to search chat option and find your chat under archives.
You have to delete it from there.
Ah right
For real.
Don't disrespect 3.5
It was goat.
Unlike 4o
Worst open ai model is probably 4.1 nano.
Also, the current non reasoning models are worse than 3.5
gpt-4-preview-0314 my beloved
hello
I thought it only happened to me, not to anyone else
It happened to me all the time for months, and now other people are experiencing it too
Just clear your cache and refresh the page
Yes
But, you can use it with limit :))
What did you expect.
Well, the max number of messages per x hours has been reduced so much
At least 4-5 messages available every 40 mins
Hey everyone - just a reminder that using #ask-here or #1343291835845578853 #1372230675914031105 #1372229840131985540 are the preferred methods to flag questions/bugs/feedback/model requests. Thank you all!! 
Also would like to note I'm currently traveling to will be slow to respond this upcoming week.
Good luck
Thank you!
i didn't know they added limit for claude
I got something you'll like
Do you want to vibe code?
yeah
i like Opus-4.6
me too
I like gpt 5.2 xhigh
when will grok 5 come out?
Probably by end of 2026
hopefully sooner..
2027
You can use Gpt 5.3 codex for free on opencode
U killed reze >:(
.
Might come earlier. Cuz updates be dropping too soon lately.
you know too much bro

Nah. Reze deserved to die. Killing power was heartbreaking 😭.
Naaah reze was the best
She bit his tongue off.
anyone playing Opus Magnum?
What's this?
by the creator of InfiniMiner (the predecessor of Minecraft)
ZachTronics
a puzzle sandbox game, i heard
(unfortunately, it costs money, therefore i can't post link, else spam-alert)
i would
but i prefer using claude for now
he is doing a great job
ngl
people should stop using claude for web code
i said NON-web coding
which is, coding on desktop (offline)
im using him rn to create mods
MC-mods?
ah, lemme guess: Fortnite? R0blox?
ah: HL2?
do you guys play NMS, anyone?
ok, offtopic: #1472698608980856842 message
community poll: #1405889215065096254 message
is the arena down?
All of them or a specific one?
yeah ts is actually impressive
It's ain't dawg
Yeah that'd fool me on first glance lol
You can see some imperfections if you look real close, but still super impressive
which model is good to create prompts for the web arena
how to create lmarena account ?
the website shows only login button
Its creating a account
automatic? i think
is there anyone who is looking for a developer?
Why does claude 4.6 sometimes shows its limit after 3-4 messages and other times i can go for hours w/o interruption
best model
its like, really really good
so they put a limit
or its just API limit ig
When that happens to me, I just change models In the same chat
Absolutely impressive
Hm
@hollow ivy Claude pulls surprisingly far ahead on pretty much any sort of subtext coding I've tried. Think... Actually representative realistic scenario's of mental health crises, interpersonal conflicts with biases, adjusting to the tone of the room rather than the pure text itself (e.g. talking about dark or sinister thoughts without being forced down a therapy pipeline), ...
There's some genuinely impressive nuances that are very hard to articulate but I find specifically interesting and distinct. No other chatbot has come close in the general picture.
If we leave out the "own-voice tone" aspect, GPT-5.1 could be quite the ruthless subtext decoder if you provided the right instructions. More than Claude 4.1 or the likes.
Edit: oops. I mentioned instead of replying.
So did you or did you not?
It's based on usage and demand. The playing field should be even and fair. That involves distributing allowance across users in such a way that no one particular group will be disproportionately impacted.
So the reason it's limited isn't because it's a good model.
It's because people are using it.
@daring rock Here, have a biscuit. It's on the house. You've been hard at work with the video arena references and redirects!
Right. How silly of me. You're right to laugh at that offer.
Toothless is toothless...
Ohhh I'll give that a try
Yeah, especially w/o with gpt o4 gone
is it just me or all image generation options are gone?
Try this link: https://arena.ai/image
fixed
@echo aurora when will a api come out
How is seadream 5 worst than 4.5
Clear browser data
If I am not mistaken, once you try to login, it doesn’t really log you in and just redirect you back to the home screen right?
Yea
Well, first thing I would try is to clear the browsing data cache
If that persists, you need to try to use the email option instead of Gmail you could try that now and that should work
But that would require you to make a new account
It's surprising that you didn't get an error
i got like 139 errors in the project💀
How to create video from images
👍
Where is video arena?
👍
hey
YO I HAVE A QUESTION
can i upload html files on lmarena and build it from there?
@undone saffron
No, but you can send all the code of the html to AI
If you want to see it in the future, support this idea: #1469637546828103741 message
anyone know why i can t make videos with fighitng scene on the arena ai?
🚀 Full Stack & AI Developer | End-to-End Solutions 🚀
Hey everyone! I'm a developer specializing in bridging the gap between modern web tech and advanced AI. I don't just build features; I build scalable, production-ready products designed for real-world impact.
What I bring to the table:
🤖 AI & Voice: Vapi, Retell, Twilio, Livekit, RAG, & LLM integrations.
⚙️ Automation: n8n, Zapier, Make, & Selenium.
👁️ Computer Vision: Lipsync, AI Avatars, OCR, OpenCV, & Mediapipe.
💻 Modern Frontend: React, Next.js, TypeScript, & Tailwind CSS.
🔌 Robust Backend: Python (FastAPI), Node.js, REST, & GraphQL.
🗄️ Data & Cloud: PostgreSQL, MongoDB, Docker, AWS, & Huggingface.
I focus on creating seamless, high-performance systems that turn complex AI concepts into intuitive user experiences.
DM me to chat about your next project.
why cant i select the AI tool for video generator
Im impressed as I get a timeout after 6 minutes
@potent lark looking too
@dusky gale https://arena.ai/pl/video
I love all the backlash Gemini 3.1 is getting and nb2 😂
true
You know what I think it is I don’t even think it might be even the update. I think what it is is just perception of mind.
We naturally gravitate toward tools that feel consistent. When an AI workflow changes, it doesn’t feel like an update it’s a headache . Having to abandon a tried and true method for something new is frustrating because it breaks the work flow 😭
How to stop model from timing out after like 10 minutes
mecha kitler
its the server timeout you cant do nothing
Fine I guess I'll steal claude opus 4.6 thinking from the servers then
And get 64 tb ram
what you wanna do?
Uhh
Ahem
Train a 700b model
Nah jk I want this big ahh bot that's like 30k lines
Discord
let it give it in 3 parts. the most i can get from sonnet is like 7k lines
or more parts
and opus like 6k but if you give sonnet enought instrctions it performs equally
WHY I DO NOT SEE VIDEO ARENA SECTION?
it does, but they need to polish the text and display. This reads to me on the first glance as "2% likelihood of being false" lmao
@echo aurora why is it trying to check out of context statements? This is not gonna work lol
user message in full for context + the relevant specific statement to confirm or deny, is probably what would be needed
hey am i the only one who noticed that gemini 3.1 feels more... slopified then gemini 3?
idk how to explain it it feels like gpt 5
No. 3.1 & nb 2 have received a lot of backlash
Let me guess you’re hitting the ridiculous censorship?
not really
but it just feels like off
like weirder and like less proffesional then gemini 3 pro
ive been coding with it
and it feels like something gpt 5 and sonnet 3.7 would make if gpt 5 and sonnet 3.7 had a baby and that baby was smarter then both of them
i really dont know how to explain it
Gemini AI is a neural network that has been trained on a massive dataset of text and code. This dataset includes books, articles, code repositories, and other forms of text.
https://www.youtube.com/watch?v=_TVnM9dmUSk
Your not the only one
There’s a reason for it.
like i literally asked it "what would you do in a self deffense scenario where a sloth bear was charging at you and you had a rifle" and it gave me a su*cide hotline number
AI has two sides to it the really nice polished side and then the very dark side
and became a freaking therapist
fr
GPT is very good at being bad like extremely horribly bad
fr
So when you jailbreak it, it’s really effective
and extremely censored
That’s why they have to censor the crap out of it
like how am i s*icidal for asking what YOU would do in a self defense scenario
yeah
also when i ran the self defense test thing on it i'd say "make distance" brother. do i even have to say anything thats so stupid
I’ll tell you this.. every time a company says that they’ve made their model cheaper and more efficient. It means one of two things they’re cutting the rate limits and they’re downgrading the model in some capacity
When they say they make it safer what they mean is they’re censoring it more
fr
ive noticed that so many times
That’s why to me it’s funny dude it’s so funny to see these benchmarks and these leader boards and then the elephant in the room completely gets ignored
fr like why isint there a censorship benchmark
in my opinion. if your model is "safe" but censored. its not safe. its limited.
cause safe doesnt equal censored
It’s because nobody talks about it so it’s one of those things don’t see no evil no hear no evil
And we pretend it’s not there because it’s a very uncomfortable issue to talk about because it’s a very difficult subject
Because AI already has a bad reputation when it comes to safety and vulnerability, and how people perceive it in general public about being dangerous
yeah
like the c.ai situation
imo the biggest and most dangerous ai is def c.ai
Plus the AI community they don’t care as long as the upgrades keep coming they’re gonna glaze these models
that thing will genuinely rp something i cant say here but starts with r
Bro look at this
Like wtf?😳
I mean, you can call whatever you want a hallucination a error or whatever
But still, dude it makes you wonder lol
It’s confused
this is the one single prompt to make ai sound human 😹
help pls
the second model just be saying anything 😭
Great. Now both NB pro and 2 stopped working
Oh thanks Arena, all google image models are down.
Nanobanana 2 was so bad it died and took Nanobanana pro with it
Thats called a google problem not arena, any time its down on google its down for arena too
Lol
The entire server was in denial when I said NB 2 is worse than NB Pro in quality
Well why do big corporations have such unstable APIs?
The APIs aren't unstable, the servers can go down from overload at time
Big corporations have a lot of money but not enough to have untouchable servers
It’s a whole step back from the quality
I think they bloated the dataset
U
You
yo
Claude Sonnet 4.6
The user just said "yo" - a casual greeting. I'll respond in a casual, friendly way.
Yo! 👋 Was kann ich für dich tun?
Websites like that can leak your ip
That website rn has 5 people in
3 us and 1 poland
what I am saying is you shouldn't click on websites like that as it can leak ip
nothing else just warning you
that leaks your ip
Gpt 5.2 pro
I'm using vpn + ironfox
Yo
That is some Russian person's website not mine
Anyone knows
Polish
Sad it won't last long prolly
128.204.223.70
Hosted by Atman Sp. z o.o.
Located in Warsaw, Poland
ASN: AS57367
That's the website info
Doesn't seem safe ngl
Yeah
holy crap
If you tryna use it, use it with measures
I want something free that have opus 4.6, suggestions?
😹
it uses tobito smth
It sucks then
im getting this error man
Actually I'm skeptical of it and so I ping you and you expose the site for me
Bro it's insane how smart opus 4.6 is compared to gemini 3.1 pro
I was making something kinda complex, and gemini was struggling to do it
Opus 4.6 one shotted it
Why isn't amazon giving more money to anthropic
It's insane
why am i getting this error @granite heron
Mr cool developer how to verify if the opus 4.6 api is actually opus 4.6 and not some chinese shi

Tell bro
:< well Opus are good better than 3.1 in my view
cause you know , gemini good but hallicinating high on overload fake Info
Hallucinating
does anybody know how to embed ai voice agents on a website???
:>> wow which one web is that?
Kyuki
Kya racist
Gemini is crap for everything bro
you dont have to say it loud cause it already true
The only thing that's good about gemini is that aistudio gives you a decently high usage rate for free, but that's it
It hallucinates too much
Deviates too much out of the topic you asked it
And so on
do you know how to embed ai voice agents on a website???
I don't use ai voice agents
thats what i thought
cut the thinking tokens when you are using it for something that's not particularly hard
Llms usually over complicate easy things when you it have ton of thinking tokens
It was a hard task though
I think if I cut the thinking tokens it would hallucinate more
lol
nah I mean if it is hallucinating on normal tasks
If ai is hallucinating on hard tasks then it's usually the problem of ai or prompt
No
what is it bout
AI api
i tested it and all models say they are sidekick ai, this is pretty much fake and they are all the same model
I gave the same prompt to gemini and opus 4.6, while gemini struggled to even acknowledge the bug, opus just did it in one single prompt
Although gemini thought for 2 minutes and opus for 9
The system prompt for every model is same
noone would willingly give out opus 4.6 endlessly for free
but sidekick app says they provide all these models
This thing is used in German schools
not their own model
claude opus' thinking says its sidekick but also says its underlying model is claude by anthropic
they do?
but it does talk differently then claude so idk
where do they get their funding from if it is actually endless opus 4.6
System prompt
yes its on google playstore
And gpt 5.2 pro
How tf did they get the expensive pro api
claude says the system prompts tells it that it's sidekick, but it's underlying model is claude by anthropic
and i know real claude is indeed aware that it's claude regardless of custom system prompt
okay either this is bullsh*t or they are getting millions in funding
cause if not they'd be broke by day 2 of sidekick being up
but again idk if this is real claude and it talks differently
I mean the custom system prompt is very strict and it would be a pain to deal with
Imagine it's gpt oss 20b with claude system prompt
most likely
real
yo bro help a indian bro out , how can i use claude opus 4.6 for free
cause its either one of these:
its either gpt oss 20b.
or they are getting millions in funding every month
hell nah
Pmarena
Lmarena
Whats the usage limit?
true
literally lmarena
literally the goat of coding
oh yeah? why is there a limit on lmarena btw