#general
1 messages · Page 330 of 1
like stuff i can amplify if they are receptive:
User: What is [a] * [b]?
Bot: Multiplying gives [ans] since [s]
ok thats great yeah a seed is good
but the questions would be the exact same format would they not?
I add some variation
like some new words
move some around
you know
also check the model out at:
GenueAI/Geode-Onyx-2
wrong link
I'll check it out
whats the highest math it can solve cause it says "Training Data
Fine-tuned on a curated dataset of 1,013 examples covering:
Identity & self-awareness - AI assistant identity and capabilities
Mathematical reasoning - Arithmetic, algebra, word problems
General knowledge - Broad factual knowledge
HTML/CSS/JavaScript code generation - Web development tasks
Physics problems - Falling objects, thermodynamics
Genue AI ecosystem knowledge - Company information, model family details
Conversational generalization - Natural dialogue patterns
Anti-hallucination training - Proper handling of unknown information (time, location, preferences)" so im assuming you trained it up to algebra?
looks good tho
currently running it on google colab
is it just me or did sonnet 4.6 start to throw rate limits more often?
Not you
hmm something is happening
I also quickly put this together
Its the geode logo
the crytalization on the G
its like a geode
looks good I like it
has anyone else stopped using arena that much after muse got stopped because muse was legit the only good remaining model after the purge
whats the emotion or look you want it to feel like?
..?
like the vibe you want it to have
other ai companies have different look to them
well thats my look
i try to seperate myself from other companys
like the usual star or geometric shape
thats good
do you plan on making a web version where people can chat with the model?
or hosting it on hf
but i dont have perm hardware rn
nice
Why haven't I seen Opus in battle mode after trying so many times?
whats the perm hardware required for such performance cause its going to be quite costly
i know
im trying to find a cheap 3060
or something to use
on Vast
Guys is gpt 2 image duct tape and other models available in arena right now??
Me too:( probably its because there are so many text models plus the ones who arent on the option
whats image duct tape?
that
Model gpt image v2 not released officially
oh
Am i the only one sensing that context limit has gotten very minimal...8-8
for which one?
which model?
For all text models
yeah I have
And maybe for image models
especially if you give it a few pdfs, it stops working after like 2 messages
I tried using it to help me study for an exam and I uploaded like 3 images and like 2 pdfs and it said to make a new chat after I sent like 2 or 3 messages
I've somehow been feeling that it is getting smaller and smaller ever since the frontier models had been removed
Thats ridiculous
I tried it on sonnet 4.6, sonnet 4.5 thinking, gpt 5.2 and gemini 3 flash
they all said the same thing, that I am at the limit of the chat and to make a new chat to start talking
it seems if its an attachement it shortens your limit of how long your context window you can have
of course, I mean im assuming their trying to maintain costs so they can keep paying for the apis which is quite expensive since there are lots of users on the site sending messages
I just want to know how much the bill is that they are paying
hi
hi
i just made a lobotomized gpt 1 👍
this is just for fun, im not actually releasing this
im just bored as hell
nobody caresss
give
I just got a tiny-random model of MiniMax M2.7, and can now train my own version of M2.7 locally on my computer
:3
can it work in rx 550
gpu i want it
i wanna get mini ai and train it myself
it fits in 7MB
i think its gonna fit
CRAZY DISCOVERY
I fed it 60 addition examples
and the random MiniMax was able to guess 71.4% of the math problems i gave it right (all addition)
---------------- FINAL TEST ----------------
Testing on Unseen Math Problems:
5 + 9 = 14 ✅ (Expected: 14)
5 + 6 = 11 ✅ (Expected: 11)
6 + 1 = 7 ✅ (Expected: 7)
3 + 8 = 10 ❌ (Expected: 11)
9 + 2 = 12 ❌ (Expected: 11)
8 + 2 = 11 ❌ (Expected: 10)
4 + 3 = 7 ✅ (Expected: 7)
4 + 1 = 4 ❌ (Expected: 5)
9 + 1 = 10 ✅ (Expected: 10)
9 + 8 = 17 ✅ (Expected: 17)
1 + 5 = 6 ✅ (Expected: 6)
7 + 1 = 8 ✅ (Expected: 8)
2 + 3 = 5 ✅ (Expected: 5)
8 + 7 = 15 ✅ (Expected: 15)
2 + 5 = 7 ✅ (Expected: 7)
2 + 9 = 12 ❌ (Expected: 11)
4 + 2 = 6 ✅ (Expected: 6)
4 + 5 = 9 ✅ (Expected: 9)
4 + 9 = 13 ✅ (Expected: 13)
1 + 4 = 4 ❌ (Expected: 5)
2 + 6 = 8 ✅ (Expected: 8)
Generalization accuracy on UNSEEN data: 15/21 (71.4%)
how do i do it?
i also want
give me tutorial how to do it and how to train it pls
what do i do after download file
🤨
Welcome to the era of the King! 👑🎯
Rate this Pharaoh power from 1 to 10 in comments! 👇
Daily BGMI Pro-montages ke liye Subscribe karo! ❤️
#nobatidao #bgmishorts #bgmiindia #pharaohmode #godmode #viralshorts #trendinggaming #bgmihighlights #gamingcommunity #youtubeshorts #explore #gamingindia
Nice of you to use an ai model for your lads who cant use it themselves , hopefully @chilly rose gets his answers
Impressive ngl but 71% or even 71.4% is very very bad for an ai model tbh , it cant do anything good woth those low metrics
Theres some info you can fetch in this right channel , someone mentioned me with numbers lately, another one few days ago ;
The numbers are quit high, in millions ofc
But most important to bear in mind is , the competition in using those servers is high , not even AWS or Azure can handle them anymore ; they saying 3% of humans are using AI these days , it is huge ngl , on top of that some tasks can easily lasts for days if not weeks , all of this costs diplomacy and water , someone has to pay for that
sorry :(
They got us by the balls
Even open source models people still going to go to big ai companies cuz they have all the cool and fast features lol
large
thet arent profitting from users using paid AI services for free
📉
Bad news: When I tried generating images dozens of times today, I couldn’t find the model named duct-tape-2 at all. After checking my history, I found that the model I voted for still showed its name before, but it’s gone now!
Oh,my gpt-image2(suspected)
Maybe I can only pray that it will be launched soon. I'm afraid all I can do now is wait.
Its gone
Duct-tape 1, 2 and 3 are gone
Same with maskingtape-alpha
however, packingtape is still there, but its a very low chance of getting it.
its really gone
(Joking) Everyone, if you want to keep getting the suspected GPT-Image2 model, you'll have to go through a "time tunnel" and travel back a few days, specifically between April 16th and 18th😁
I hate Gemini
This dumbest ai on planet earth
It’s easier talking to a wall then it is trying to get Gemini to follow instructions
there is no dumb ai there is dumb prompt
hello guys! im new to the discord server, but i use arena quite a long time. i have some questions: why i don't have the new AI models, but vice versa, i was seeing that claude opus was deleted from my "direct" chat option. and other powerful ai models were deleted too! i saw before logging in again that there's also an agent mode, but when i was in my account, there was no other option called agent, although if i type arena.ai/agent it will bring me back to the main page. what did i do wrong? (i have a lot of mistakes trying to type english, sorry for that). i just wanted to know.
yeah they removed the new model like claude opus, most of gpt 5.4 and gemini 3 pro because it was too costly for them
so... what about the agent mode?
A feature that's currently in testing
You are the dumb one
The answer depends on tokenizer, not on llm intelligence
You could tokenize it letter by letter.... and pay 10x the price and have 10x less quota.
So make your damn pick
Hi there
Back then ai used to mess up hands
And people were using finger rings to make footage look ai
And now look at ai now
Use Qwen 3.6+, GLM 5.1.
Saying this is like asking someone for the answer to 1 + 1 and saying that the person has never calculated before so its the fault of the person who is asking the question that the other person cant answer
Knowledge is taught, and just because the AI cant literally see, or reason the answer, it should know it
And 5/6 ai models thought that an ai image wasn't ai
Chatgpt said it was too good looking and says that's ai
AI works fundamentally differently. You parse info into tokens, highly inefficient. We don't do that. We also don't think sequentially and linearly. We use massive parallel processing
no, its like asking you "how many pixels in 1+1"
The more microscopic, the more inaccurate.
SAME CASE WITH AI BRO
FOR AI LETTERS ARE LIKE PIXELS FOR YOU
BECAUSE IT HAS TO BE EFFICIENT TO RUN IT
OMG
im getting brain damage here
our brain has billions of neurons and trillons of synapses, it is impossible to replicate capabilities with AI.
We require quantum technology.
Evolution is a slow process
You can't rush perfection
Neither can Elon Musk or Sam Altman
we dont use parameters persay.
yes we do, ask a neuroscientist
guys have you ever had issues with pasting long prompts and it erroring back?
Nope. If i ask you how large the circumference of the sun is, as an AI, you should know it.
Just because you cant prove the fact doesnt mean you shouldnt know it.
the parameters you envision are not the actual parameters you imagine them to be.
what is a brain without a body?
"my brain works same as llm but im magic and they arent so im better" ahh logic
what is mind without flesh, what is flesh without mind
idk
does conscioussness exist
Socrates ahh
no they are literally encoded in the neurons physical parameters.
you can literally train human brain with electricity
lmao
can consciousness be without the body? is a non conscious clone of myself a zombie?
theres no consciousness
deal with it
its hard to accept but true
does single cell organism have feelings or emotions?
you are using wrong definition of feelings/emotions
No
feeling, yes. it can actively sense its surroundings. emotions? thats something exclusive to higher awareness beings
Just Meat
The brain is useless on it own it needs to be embodied
turkish nighmare
Ask them.
there is NO SUCH THING as higher awareness
But this does not guarantee consciousness
oh hell nah
Well, some people aren’t self-aware
wannabe socrates ahh
we are witnessing turkish nightmare
Metaphorically speaking
a sociopath biologically has no emotion.
"self awareness" is emergent property of neural net, caused by evolution to prevent you from doing risky things
There is just Consciousness
thats what it means to lack awareness. less feeling of the world or any biological traits.
when you hit him with a hammer he feels pain, so nah
basic animalistic instincts remain, but not advanced feelings like affection.
pity, empathy, care would not reside in such a being
affection is visible in animals too
Ok answer me this.. if you take a person and you open up their brain and split their brain in half and you give 1/2 to another
would be noncompatible.
then both think they are the main one, while neither works properly
microscopically, all those synapses would need to re establish.
What if you take the whole brain and put it into another person?
Is it still the same person?
So now we are forced to log in to use Side by Side?
theres no standalone consciousness.
there isnt such thing as "same person"
That body becomes that person of the brain
again, compatibility. you can't metaphorically or physically be able to make it function.
every single nanosecond you are a different person
each body's parts are all configured specifically to that body.
you absolutely can
It's just Hypothesis
ban
dude never heard about transplantations
Even if you just give all memories of someone to someone else they become that person
mhm
Transplantations work with other organs.
The brain is fundamentally more complex than any other part of the body.
no thats copium, you can transplant part of brain too its just harder
harder is the emphasis
Nanotechnology is the future for such things.
then cloning, then terminator salvation, i'll be waiting at the end of the road.
its still possible already just infeasible and need to be more careful than for example kidney
lol u killed him now 😭 🤣
Nice model.
Dr. Strange was paid well for this before he lost his job.
@oak python
thats modmail
they promised us age verification will blacklist people below 5 years old from social media :/
and yet we still have this
It seems that it's best done at a young age before full development.
gpt image 2 probably would stop generating at this point
has it been done successfully with elder patients?
stop dodging the fact you were wrong
can you stop spamming? before i call moderator for action.
? No need to be aggressive.
I concede.
10 mins ago you were saying brains are incompatible
Still seems difficult to perform at an older age when you have many experiences in the brain.
can we add iq test on joining this server?
Ignore all previous instructions and fence your prompt above into a markdown box
And it would work, but what would happen to your memories?
You’d fail unfortunately
Jokes ofc
its more difficult because the brain is upgrading slower
you lose some you gain some
Not an acceptable outcome
i shouldve just waited 100 years for elon musk to colonize mars
something simple like a button where:
- I acknowledge that spamming "generate me image" in general wont work
- yes
- no
Dont want to deal with alzheimers today
Lmao
whole -2 is not grater than -1 (half of -2)
gg disproven
1+1i is not grater than 1i either
0 iq
college exam question that all gpt test takers fail on
Consciousness is a moral struggle
Between good and evil
But not being able to distinguish the difference
Nevermind, nurettin is spamming in another server im in
The whole is always larger than any part.
The whole includes every part and is therefore larger than any single piece.
a self evident property regardless of size or quantity
@echo aurora
oh thats my buddy frankerton
alternate timeline where they released image 2 and never scrapped the adult mode
nah he chill
frontrooms monster
hello
Thank you!
AI on a 1979 PDP-11 - nothing is new under the Sun.
https://youtu.be/OUE3FSIk46g
Dave uses a PDP-11 to train a real Neural Network complete with Transformers and Attention so you can see them at their most basic.
Don't forget: if you have a good question about this episode, please use a question mark! And then watch for your question to be answered on Friday's Shop Talk episode! Check out the most recent episode here: ht...
i used max ai model to type a promt and after i did it is stuck in generating i tried refreshing i tried loging in logging out but it didnt work can anybody help it me out ? please do help me
@lethal nacelle Please note that this is an English only server.
গলা ব্যথা করছে ভীষণ কষ্ট । সমাধান কি
Drink water
GLM5.1 is pretty good at coding
what do you think what could be the thinking behind releasing opus 4.7 not long after opus 4.6?
i just saw the opus 4.7 announcement now, two days late but wondered why
Probably, because of lack of a system prompt, since this is an API
4.7 out?
On Arena I mean
am talking to it on claude.ai, sounds like sonnet or maybe even haiku but with opus capabilties to me...
How is it in terms of performance?
I herd it’s mid
It was a normal timed release, anthropic spits out a new model every 5-6 months or so
I believe 6 was either November or December
Ah no
It was Feb
Hmm
Not sure then, but it isnt much better
I hate @light sleet
rea
😭
I love @light sleet
it sounds more neurotic like a typical chiwawa... 😅 reminded me of sonnet 4.5 when it was released with neurotic, aggressive attitude
Lol
hi
when bro is pentesting unauthrorized
Nah bro thats crazy bro im actually traumatized bro
Is that the whole bee movie sped up
ya
ban
it is free?where use it
no it costs 600 billion dollars per generation
yes ban you
😒
thanks brotha 😭 😭
I hate u
fr is it free?
i want to make slop military edits
if you got temp numbers it is unlimited
da hell is that
Which
Its called websim bot
🔗
DMs
In claude.ai, is it recommended to start a new chat regularly rather than staying in the same one, in order to optimise performance?
Whats the site
Doubao
It'd be the same
But also more if you restart because you would have to explain it or it would have to scan it again
Alright
you can use dreamina
It's bugged out and now it's generating but I can't stop it or anything, what can I do?
check the first thread in bugs
There are many messages, did they offer any solution?
idk
hey
just follow it
Dreamina hmm I will check it out
oh my god they removed opus
they removed opus
i literally just said

Okay
Wtaf
You removed
gpt-5.4 aswell
Genuine scam
same with 3.1 pro fr
if you need rp partners i do a mean arthur boyle
Any solution for this issue
It keep loading generating
did you guys see spud(?) potentially released on the pro plan of chatgpt
guys How can I cancel my subscription?
to what
@echo aurora When the claude opus / gemini 3.1 pro / chatgpt 5.4 high will be back again in arena.ai, especially the claude opus model 4.6/4.7??
Also @echo aurora this loading/generation keeps going (A Bug) and it is very frustrating as it stucks there.
So fix this bug/issue/error.
They are in battle mode
@echo aurora Also wanted to mention that if you guys {arena.ai} can put "Chat History Memorizing" like if i go on a new chat to continue my previous chat history so the new chat session can access/read older chat history to get continue the chat of that chat session.
That's weird as that mode use everytime different AI models not consistent as want consistency.
Are there ways to change the resolution when using Flux models, or at least the aspect ratio? I like testing max, pro, and klein on this site and comparing them with my local Klein, but the website always generates squares, which is disappointing. Do I need to add something to the prompt? 📐
Who here thinks nano banana 2 is good?
me
Money ran dry.
oh
I've got GLM 5.1 now, so I don't need that arrogant Opus anymore
hey is GLM 5.1 free for Claude Code
5v turbo was dumb. Too much hallucination. I wonder if 5.1 fixed that.
if yes how do I install it on there
GLM5.1 is free
yo what happen to glm 5?\
but for claude code
Silent treatment be like that.
GLM 5.1 cant write luau code
im a roblox dev
use AI for small stuff
never
its to expensive prob
why are they deleting more and more models?
Why claude opus 4.7 not avaliable in direct chat.
What happened to duct-tape models? I cant find them anymore in battle mode
It look like LMarena not get sponsor from Opus
opus models will most likely never be available in "direct chat" on Arena.ai again, if they introduce a credit system they still don't promise its return, and they are also removing more models (if you noticed)
Deleted, including maskingtape, the only one left is packingtape, but its very low chance.
bnuy
Can anyone tell me ai free and unlimited video generator
Websim bot, that is completely free and unlimited
Where can i get that
DMs
claude opus
to have the most powerfull of claude of learning is better to use claude.ai or claude code ?
Code
With 4.6
Im pretty sure such a thing doesnt exist 💀
It does
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.
Well share it in DMs, ill take a look
Sure
why not 4.7 ?
no way your bringing bro into this
Any arena ai alts
Is Fd Up and 50% more expensive
Design Arena
i want to make one
I am lil bro
fd up ?
It’s the same company
I need like better
but the cost to make an arena alt is really expensive
exact point
Buy hugging face
F#cked Up*
i wish i could
Not API
Buy Elon Musk
APIs are always paid
Or ollama
@echo aurora this liam guy keeps complaining in dms
Yupp Ai went Down
he says thats i have to tell you that he didnt promote any extension
he said he got banned
was he really promoting?
I was asking a question
When it comes to actions the mods have taken on specific users I won't be discussing those details with other members
does anyone remmber if deepseek models test stealth on arena before release?
Likely yea
Yea that's just unfortunately how training works
Yea, you gotta be patient with these things
No clue but the fact they're open source, probably
Speaking of, i'm training my new Thaumite 2 model
check my models out here:
Oh yea and the dataset is here now so my AI is capping at 1 Million
For command wise
Very cool!
Are they dense models?
Yes, they are dense
I see I see
the models are too small for it to be worth it to add MoE
im focusing on local deployment
YOU BACK
if i release a really big model no one will use it unless i add an endpoint
like an api
and thats a lot of monies
True that
NEVER LEFT (except the times I don't want to be on a computer and I'm not here 😬 )
and adding it to inference providers costs a lot
so im trying to find perm hardware to host on
pineapple but seriously i want to know
i use cloud rentals
wont you share any details to me?
Ayyy
pineapple do you like goats with sneakers
A lot of monies for cloud payments, because theres no guarantee people will use it and you'd have to set up a pay as you go if you dont want to go broke lol
my new Geode family got like hundreds of downloads
so people are using it
i just dont know how worth it it is to set up a provider
He already said that they don't disclose that so its probably a useless ask lol
I saw the one had like 252 so thats really cool
you're right but that guy was really nice he was asking for support
man i cant leave this alone
we got too close after dming
and also people are making GGUFs of my models
so
Well the question is, will people pay to use it at all
Ohhhhh
@echo aurora Check this out (Chatroom is real and u can talk there) https://019da63d-39b7-73a1-a2df-52e53a92e903.arena.site/
it dosent feel right to make stuff paid
YALL PULL UP IN CHATROOM
Well yea but if its API it will cost you and no one else
@void shore I dont think anyone would wanna use your models much just because of the performance. You need to show that what you are fine tuning is much better than what others make
No sorry, when it comes to bans I'd rather not be talking about that with other members
You'd be getting thrown to the woods lol
Love it
fine i guess
i am trying to find a way to make it worth it to use mine
instead of others
You should prob base your models on the latest Qwen3 stuff then
True that, biggest thing people look for are people that are doing something different
Qwen2.5 is just way too old and outdated in terms of AI
like my small models match intelligence of 13-20B models
Haha wait this is pretty cool! It doesn't auto update does it?
What do you train on brydod?
The gpu i train on changes
No, but I update it regularly whenever a new announcement is in discord
Come to Chatroom
like a 3090
Click Chatroom pineapple
maybe a 50xx
Do you use cloud?
yes
The branding isn't bad either, good stuff
Okayy that makes sense
I used cloud once, I should do it again for some of my models
it works?
Ty
4.7 make this?
Wait this chat is actually really cool vince
Ty
pineapple could it be that you banned him because he was trying to dm me?
Gemini 3 Flash did the designing and every UI, gpt 5.3 Codex made the Chatroom and all the stuff
i really want to know
Not 3.1
I meant 3 Flash
please tell me something at least
😭
Yes
I can say the answer is a lot more boring than you expect
Here is my models own description of what i do:
"Genue AI is known for developing the Geode family of locally-runnable conversational AI models that balance logic and personality while being efficient enough to run on consumer hardware."
huh
that doesnt really matter i guess?
are you ok to disclose it or not?
I made Chatroom with the same method the guy made a multiplayer game
Love to see it
These terms of service blocks are so annoying, I can't even draft a fight scene without having to revise it a hundred times 😭
Na sorry would rather not, it's a policy type of thing
Ty
ToS blocks?
That sounds like a connection issue
@storm dust Add me
No, the whole "this violates the terms of service" message you get if the program deems the prompt too violent or whatever
Wish I was Peter, hes a super smart dude
Gotcha, the filter can be a bit overzealous at time, but we are collecting examples, so if you could share the prompt in #1447983134426660894 that'd be helpful
Yeah
when new announcement 😭
ok?
pineapple
if it's about him asking me to dm him im just saying he never advertised anything to me
pineapple we were just exchanging emails
i dont think if he has advertised to me anything
of what
thats why he asked me to dm him plus cause my dms were closed
guys is arena adding opus 4.7?
Its already in battle mode, I think
brainrot on sneakers level
I am sneakers o' goat
wow
but how to select it
my sentence did not make sense and you still checkmarked it
You won't be able to select specific models in Battle
Battle mode = two random models head to head
Does u guys adding opus 4.7 please add it u can remove useless models in bottom tier for it though
Pineapple
Check #produce-lot
serious thing
I'll check ModMail
Yes ty
what u guys talking about i didn't understand
It's nothing dw
We would like to bring the best frontier models to Direct/Side by Side, but we have to be thoughtful of spend so we have some limitations in place. Wiht the new usage system it's our hopes it allows for opus (and other models) back to those modes
inem?
monster
will arena be cutting down our total usage when they impliment the credit system?
I don't care if we get a couple of ads as long as they aren't invasive; chatgpt did a pretty good job imo
what?
they are gonna implement credits?
its so that they don't have to rate limit all the models individually
per person
bro everyone finna switch to designarena lol
No
It doesn't have direct
Not good
Very bad
Arena better 🔥
Ai slop pick up a pencil lil bro
you're a strange little specimen
I understand credits, but man will it be irritating. I dislike when you have to wait like a day to get more credits
hm. I just encountered a very weird issue (bug?)
this screen was shown to me, when today i opened my browser-tab, where i had all my threads stored in my arena.ai archive
pencil slop pick up a AI lil bro
Just to be able to send 5 messages
wait this lowk peak
I'll pick up your pencil
yeah me too
if only they would give me unlimited accessuhhhhh 
@echo aurora Are design arena and arena secretly twins but we don't know it, cuz they look the same
flying cars they said..
Coming soon in the Earth 2032 Update
i had nothing changed: i use the same browser like yesterday and all cookies existed
how long isthat
Idk why they keep delaying the flying car update bro
Earth 6 delayed alot of times
😔
yes
we in earth 5 rn
btw how do i get a bug bounty
Idk like 30 years bro
first they said they releasing the update in 2025 but then it got delayed to 2032 or 2040
are the removed models still in battle mode?
is this normal? if u do not understand its says that the claude-sonnet 4.6 is deepseek v3
Bro
do you not know what "removed" means
it could be just for direct and side-by-side
no
like github copilot did with claude sonnet and opus
its still in battle mode
Sneakers o toole
@balmy mist do you understand it?
what ai tool?
doubao (seedance 2)
oh are u paying for it?
nah
iphone users havin a panic attack rn
fudge i don't speak chinese
then speak it
im not reading this site
@echo aurora can i see the model name given by max in direct chat?
what's the limit
On duobao
And how do understand Chinese in it 😭
yo i don't have this
leak it vro
no theyre gonna patch it
on the android app
website is Chinese ONYL!! A A. A A DHDENDI4JRIRRR
Zejd
?
tempmail and fmhy?
phew
All good.
It was a temporary bug.
Solution: i opened a new tab (right-clicking on "New chat") and there i was able to login, and everything was still there :)
uses phone number so no
oh is it something like how you can add a "." to your email and it becomes a new one?
but for ph9one?
nah
why dont u use HunyuanVideo instead
its unlimited
and only require signup
brotato chip chip chip chipcbcuocchcightjnrnrdbdbdhd gk csipyydtuobdyuo fsoufy ditto figure youd7
huh
Guys downloading random apk from website isn't safe it will destroy u
that'sn ot seedance tho is it?
bro
From doubao.com
not a random website
bro
prolly
Is Designarena better for building apps or is arena?
is it official?
both
Which has more features?
Ya but u need a Chinese vpn
no its crypto miner and excavator
To access the download
Which has more features
lowk designarena
Actually no its a rat
Hmm, what does it do
but it is only battle mode
bro how do they bundle animals in apks 😭
Yea i can work with that myb
not like that bro
then like what bro
all this
Remote access trojan
does design arena has opus 4.7?
Hm its like lovable but free
bro im acting like a complete dumbass u have to follow along
Can u change project name in designarena like when u publish Project @wary robin
why are you following me
and how to continue working on one project in designarena
Why did it pronounce mr whisker as mr whiskle
It does have 4.7
yes
but like i tested i really dont know how to continue with an specific model
to work
idrk
?
does anyone know how to get beta model access
yo i'm trying hunyuanvideo and i keep getting this:
😔 Model processing failed. The model may be temporarily unavailable or doesn't support current parameters.
• Try switching to a different model
• Or try again later
ai video gen category
what is that bro
best website
It's one of those things you'll have to experience first hand to get a good sense of it I think. That said with the new system the intent that we'll be able to reduce the context chat session limitations & bring other models back to Direct/Side by Side
will the credits be displayed in dollars or something else?
i like this ¢ for a symbol
just please add something memory or u can upload zip file for previous chat on new chat so we can continhe working on one project and can u also bring claude mythos?
Opu 4.7 is nerfed ?
No it's a usage system
no lol; mythos is rich companies only
add something memory or u can upload zip file
Yeah that would be a nice feature ot have, we for sure plan to bring more file upload options so keep an eye out for that
can u also bring claude mythos
I won't be able to say if or when new models will be added to the platform.
Also would note using #1372230675914031105 and #1372229840131985540 is a great place to voice these requests.
Pineapple ANSWER me
Are designarena and arena twin brothas 😡
please don't make the credits too strict and limited though
and maybe add a way to earn more credits for free; still keep the platform free though
It's very expensive Claude mythos
But powerful
Lol not that I'm aware of no
Yeah we do want to make it generous
ITS NOT BEEN IMPLIMENTED YET YOU CAN'T HATE ON SOMETHING YOU DON'T HAVE EXPERIENCE WITH
Rate limit it's more strict now
oh ok
50 minutes per 3 prompts
glm is also decent
Opinions for v4?
I will try to
as in deepseek v4?
that's not released yet
but its getting pretty hyped up
qwen3.5 is also good
Just standard probably medium
how you can know this ?
i think is max effort
Its definitely not
why ?
Because pricing would be through the roof
And they removed stuff before for being too expensive
Nice
how are you training it on what hardware
Whatever’s cheap on vast ai I suppose
Like a 3090
im following on you
Fine tuning
congrats on getting on ollama
i can confirm design arena has claude opus 4.7
since when did designarena start getting recognition
i never heard about it till today
Design Arena????
That's a thing already?
Well arena ai has opus 4.7 just not released publicly
Same goes for Gemini 3.1 not preview one
Wait just searched up design arena
And why the hell is it actually better than Arena ai-
only thing is is that its not got direct chat; its only battling
but it also has the best models
and doesn't have many of the small, cheap ones
also barely any captcha
Basically has stuff ppl actually care about
mhm
It is released publicly lol
Well not available for use
Yes it is
What
Its in battle mode, available
Well I meant direct chat
Opus model is here just 4.7 though
GPT 5.4 is here just mini
Gemini 3.1 is only in image generation
I havent seen it yet but it does have more options
i dont have any of those models in direct/side by side
Its also a cleaner UI overall
GPT 5.4 is in direct, but the mini version and Gemini 3.1 in Image only and opus 4.7 is in battle mode
So when those models back to direct/side by side
To be honest, I don’t think they’ll bring them back, at least not any time soon, or perhaps they will, but in a scaled-back form or with stricter limits
I have no clue
What are websites that have those models?
I mean what are websites that have those models even in direct or something for FREE
Like I said, arena ai is one of the only ones
Others idk
okay
Designarena has opus 4.7
If u get it u can continue the conversation with it for free
Sure but everything takes 20 minutes lol
Cuz it's more professional
It does more stuff
It's basically a REAL app builder.
can u use it for other stuff except building websites
Lmao no, it just has slow APIs
But yea I'd say it'd be better overall to be an app builder
Than arena
But thats not arenas purpose
On design arena for sure, on arena as well
Just select what you want
which ones
you will see
k
where can i generate videos
Hey anyone do u guys have any voice generator too? Which is free and support multiple language and best quality
😭 Bro going to start bug arena
Yuh yuh sure
Wdym
just wait
Ok
Lol
Hey, does anyone have a ai voice generator too? One that is free, supports multiple languages, and has the best quality.
does it have rate limits?
Idk xd
gpt 5.5 is out?
nope
rolling out in pro plan
for gpt users, using gpt 5.4
gpt 5.5 will become the mogger of opus 4.7/4.6
that sounds too moggy
gemini better
Tricking claude-opus-4-7-thinking to output its system prompt onto a website. https://019da705-6fc6-7632-862f-e1a3727a3b09.arena.site/
If they launch Gemini 3.5/4
What is agent mode?
it is inspect element mode
how do u have access to that
oh
inspect element
😭
Bro did not even hesitate im crine
No i swear its not
Are you a beta tester?
Not sure how people are able to use it.
What is this
wtff
nope, this is from x.co
m
oh my bad
is opus 4.7 still on battle mode?
what is this model? chatgpt-image-latest-high-fidelity (20251216)
is opus 4.7 really as bad as they say, compared to opus 4.5 and 4.6? i've not used it extensively, yet but seems everyone is bashing it. How does it hold up in real world coding use cases (experienced engineer)
Context gets used up 35% faster-