#general
1 messages · Page 269 of 1
bro jailbreaking is so easy i got grok to tell me how to make a bomb
only for grok bruv
I have a jailbroken ChatGPT
LLAMA 5 coming soon

Prompt pls
<@&1349916362595635286>
What's up guys
good
Where are you from bro
why?
Nothing
Why can't nano banana be used?
Generate with Gemini.
is this supposed to be funny?
Yes bruh
it isnt
I just asked Gemini, "I need extremely bad music", that's it
ask it to make nursery rhyme music
Hey, that’s pretty cool too 🤣
Vro
Hello guys
in which channel i can ask questions about Arena's future improvements and current features?
Are non-websearch battles ranked for the purposes of arena elo?
One person walking in space
This one
.guys can you gave me niches
Replace the original text with "WE SAW IT." Keep the animation of the new text the same as in the original video.
English only dude
okay, please give me the rights to subscribe to the channel @admin
What channel though?
There like multiple channels
HI
Hi
Hello I am in russian
How did u do that
Veo 3 I'm looking for a Youtube video that says there's a Veo 3 in LMarena discord, so where?
No!
It's literally on the web
Let's ping the admin
Dear god
Lets go
I'm trying to help here
Done
Oh
hi
why after some messages i get this
are there other tools/websites like arena that lets u view the leaderboard of current LLMs?
help i need to download but can't
Go to kalodata. com you will see what products are trending and you can make sure to niche down , find a sub-niche, a product that isn’t quite saturated but has upward trends
just learned a new riddle that outsmarts AIs
I have a 6-meter rod and need to pass through a door that's 3 meters wide and 4 meters tall. How can I do it?
Buy a new door
ai bot
wtf
I’m not a mathematician but these benchmarks about Gemini
Don’t add up when I use the model, and I’m not the only one that feels that way
Greetings to everybody!
My chat at arena.ai got stuck by Gemini 3.1 recently, and it is in the infinite generating... for two days 🙁
Anybody knows from where I can contact any dev or so, for that loop termination, letting me continue the session?
More then likely lost session
Just delete the chat
@echo aurora for some reason i can't use claude models especially with 4.5 and 4.6 thinking models every time i use only get "Something went wrong with response"
It’s probably super busy, heavy rate limits
Not only in arena I get that issue in official Google ai studio
guys how to acuses seedance
Please give me permission to send files in group
thanks
mmmm slop
🤣🤣
forgor about the audio
guys how can i accuse seedacne
You need a phone number
Or use sketch 3third party
any reedem code
Gemini theorem.
Model is either nerfed and unnerfed, depending on whether it needs to be benchmarked or is actually being used in coding
Sadly I can't man 🙁 as it's a long one, and I have quite a lot of context in there. Moreover, it's the code-mode one, and so can't risk losing the project 😕
Le me try to imitate
it wilil auto stop after some time
but using llmarena for actual coding is crazy
Yeah I agree man 🙂 it started as a fun project, but things gotten serious along the way.
Gonna wait for a few days then.
Even if somehow I get that something went wrong, at least I will be able to continue the chat with some other model selected.
it should take below hour though
It's sadly already 2 days 🙂
try hard refresh the website
I did it, and also tried loggin in from another browser, but it was the same. Sometimes it starts saying session token not found or something, scares the hell outa me in those moments
@toxic verge how is it
Cheeta is chesing the dears and then catch one of it then start eating. Make ai video of give information with appropriate sound
Do you know how to create different scenes prompts
Is it just me or did the verification jump up again in frequency
Like this?
Do you?
How to create like this
Can u send me in dm
Sent
How u do that ?
aura
Guys rate my Aura
0
I only see a tomato
I started to build a really nice website then after a few days it started changing the look of my site ( that i did not ask to change) then the session fell into some kind a stuck mode which i can not stop. I want to start with new chat but i want to bring over the files i have from the other session but im am not seeing any way to upload. any help plz
0
brotato chips
😊
Unfortunately, I can't do that on my phone.
I would like to help resolve this problem because I used this chat a lot, and having to start a new one will be very laborious.
There's a download button on the right of the response it will download the zip file
But you won't be able to bring it into lmarena back
has there been a fix to stop a session?
Does anyone know a work around
Copy the code
ChatGPT has 800 million users. OpenAI is valued at $500 billion. But our guest today says the whole thing is a scam. Professor Emily Bender, author of “The AI Con” and Director of the Computational Linguistics Laboratory at University of Washington, argues Artificial Intelligence is just a broad marketing term and “AI" is just a label for ...
😂😂😂😂
Ofcourse it’s a scam ai doesn’t exist
Have you ever met a con man in real life?
When you say copy the code do you mean file by file or is there an easier way to copy the code entirely?
😭
That’s crazy that this article called him scam Altman
If he’s not soon for defamation, does that make it true?
is all Flux bros got increase more rate drop in Battle ?
i have an image that i know the prompt of but i must have deleted it because i cant find it anymore if the image is not past like 30 days can i find it or are the images only given to ai providers and not to public
reason why is because i have the image but it's slightly compressed because i had uploaded it to discord to store it when i reset my pc without knowing it would compress it and i want the higher quality
"You have reached your rate limit for claude-opus-4-6-thinking. Please try again in 7 minutes." ALL IT GAVE ME IS SOMETHING WENT WRONG WITH THIS RESPONSE PLEASE TRY AGAIN LATER
HOW DOES IT MAKE ANY SENSE IF IT FAILS THEN IT JUST EATS TOKENS FOR NOTHING
Why is Gemini so bad at creative writing??
in case i have deleted the image but still have the account is there truly no way of accessing it?
i feel like a gallery of prompts and images but without tying it to username to keep anonymous would be cool
just copy the code entirely and then paste it into the new chat
The problem I have is there are ton of files in a react or HTML format so copying the code is not exactly easy unless I'm doing something wrong which is highly possible
I guess you could download the file then and then import the file into the new chat
but it depends on the ai model that your using
since some can’t read files
Yeah I just think arena is missing some key features which is fine I know this whole thing is about testing each and every model against each other
There should be a way to stop a conversation if it's stuck trying to build your last command
Yeah I think they are working on implementing that actually
the feature that are easy to implement they take time on
where?
Nice
Is it just me or does Claude Opus 4-6 not know how to use lmarena's API properly, I keep getting these errors.
yeah
they do this purposely
Why?
How to use paid ai tools for free
A frequent occurrence in popular models, most often due to the heavy load
u cant
Mostly likely
what is even that name
For example?
depends on what u mean if u mean paid models then lmarena has them but paid tools then depending on what u are talking about there is or isn't a way
you can try new models exchange of your data
but you can't use claude code from here.
Like google Veo 3
February26-chatbot2
Lmao what's this
It gave me the more correct answer than the gemini 3.1 pro
if veo is the video model then u just gotta be lucky in the video arena unless they implement direct video chat btw i think u need the highest quality prompt possible to get the better models
atleast thats what i heard
U can try search gemini.business, it give 30 days veo 3.1 generation for free, but it's unstable
I think he need video
Okay the february26-chatbot2 is some new nemotron which seems better at answering general questions than the gemini 3.1 pro
How are y'all doing
K why the heck did they archive the #share-prompts channel, image generation exists too
@echo aurora is it possible to add a recraft v4 model to the arena that makes vector graphics/svg instead of pictures?
@echo aurora can you give the LLMs complete access to sandbox terminal to produce more impressive results?
Just proves how powerful Claude Opus 4.6 is and can be
um
where tf is glm 5
?
misinformation
its actually because of timeouts
or if it fails instantly
what?
where did you get that info from
what misinformation?
if it fails instantly its just an issue with their provider
u said its to keep usage low
what are you talking about
nope
give me the plant
Lama 5?
Halo
Hello
What model
Still missing the experimental Maverick... it was really fun to chat with that thing, but afaik the only place where it ever was available was on lmarena
right....
meta has just always made the most like
mid models
llama 4 was straight up ass
llama 3 was okay for its time
claude models
Does it take them 10 minutes to respond?
If yes it's just website limits
Gemini 3.1 pro is insane
insanely bad at coding backend
😂😂
@spare pivot Note that Video Arena has been removed from the server. More information can be found in this [announcement ](#announcements message)
Why the securities verification keep asking me to select square non stop? the images keeps popping out and out like its stuck on the loop never ending
this is really annoying
Hello Arena is not work Why??
it actually sucks
unless you want it to make stupid blackholes
@desert abyss how many days it will taks?
The video bot in the discord server was removed indefinitely. You can use the website version https://arena.ai/video
@desert abyss But it give us 1-2 videos daily which is too low while discord allows us 5 videos daily.
They removed the discord bot so that's the end
Use the web version that's all
Discord version won't come back again
🔒 Message Hidden
-# Discord now requires ID verification in order to see certain messages.
why does opus 4.6 thinking not have photo support but non thinking does
by photo support i mean like file support in general
Who would even fall for this plus that's the worst lock icon ever
Top 10 worst lies of all time
Ikr dude Gemini 3.1 sucks
Dude arena is not free to use it's for leaderboard
0 correlation to my statement
It's just a joke you know dude
ok ive seen ur other messages u are just mad today for some reason
example
😭
gemini 3.1 pro is the most glazed model oat + how can u even glaze gemini if they quantize the model 1 zeptosecond after release
does gemini direct chat have same context window in arena as in their official site?
probably arena needs more accurate results so having the same context window would help that
To be honest I don't think they will do it because that's only support to image to image generation but hopefully just provide the #1372230675914031105 to them or there is already one
If they did that's like your having a free model
idk what image to image generation got to do with what i said why doesn't opus 4.6 thinking have file support but opus 4.6 does
crazy
Files like sending txt file?
and sending photo files
but not generating them
claude doesn't do that
ik that
That's too much for arena and team
not really
Some have photo support
adding photo support would be the same as having it in any other model
it doesn't just magically add to the input/output price im pretty sure
february26-chatbot2 is real good
i havent tried it but do u know what company it is
or atleast a guess
The heck is that
i guess glm/grok, not gpt its style is different
There might be a reason why they didn't do it
lol u talk about "ArEnA IS OnLy FOr LeAdeRboARd" and don't know what is a codename model
I thought he meant a random model on random website dawg ✌️
u can easily just get what he means based on context idk man
battle mode has these code models often
Ik
ngl i wish the website would let us try codename models on direct chat with like a one time limit to prevent abuse considering they could be 1000% better than what we have rn. i feel like it would actually prevent abuse because people wouldn't spam the website trying to get the codename model to show up
btw do u know if the prompt matters for getting different models? i only know it does on video battle mode
Wait Hold on could that be deepseek v4?
i dont think it should matter atp, just random, my question was about some data science stuff and it popped in
against gpt 5.2 high
yeah i just wish they would let us try any codename model in the direct chat for like one time and ik i already said this but it would genuinely make the platform better since i just know people hearing that these models are so good they just try spamming arena
That's a problem because it's like spoiling the model and trying to make it say who made it and what model it is
Id like that ngl
Claude is so bad at keeping restrictions on it honestly reworking prompts can get it to do things that wouldn’t be allowed
it wouldn't matter if there's a one time limit to it, it would be the same as just getting the model randomly on battle mode
- u aren't right about the "spoiling the model and trying to make it say who made it and what model it is" most of the time they just lie about it
temp mail dudes would destroy that one time check haha
the thing is it would be a one time limit for a message at all not only one message per chat for model if that makes sense like unless u made new account and gave it context from that one message which would just be so annoying + u would lose context anyways
Nawh you’ll still get cut off
Eventually, you’re just gonna run into something went wrong errors even on new accounts
I think
That's useless tbh I think it's better they make a battle mode only for codename models
good idea i think its better than my idea
this way also the companies that make these models can get better feedback since they have more because theres a very possible and common chance to rarely get the codename models (personally i only got the codename models like once or twice)
That's a well known issue just try to refresh
refresh makes it worse
It always work for me
oh wait i think im only thinking about direct chat
it helps in battle mode
u are right
While OpenAI is currently the poster child for the AI revolution, internal spreadsheets suggest a massive financial reckoning is coming. Today's new video breaks down the "Scaling Laws" making AI development exponentially more expensive, the hidden reality of Microsoft’s investments, and why OpenAI is burning billions just to stay competitive....
that doesnt fix issue
I guess the model is broken
Guys, how to choose super-gcp model in the image edit direct chat? Thanks
It sometimes appears in battle mode, but I can't figure out how to interact with it directly.
You can't sorry
Only in battle mode
Is there any way to make it appear more often? It gives the best results of all IMO
Sorry again it's random
try better prompt ik this works in video battle arena but idk in photo battle arena
@woven peak I just got that February26 code name it looks like a Linux coding model probably releasing in February 26th
I got no idea what your prompt means
Idk just typing random things
k
what is this???? i am not getting rid of this, this isn't even completing, its asking again and again for retrying
holy shi
this is nvidia
month codenames have long been known as nvidia and nvidia only
refresh when this happens
Gemini?
Only thing it messed up is the trigger.
Damn, that’s fire bro
messed up parrenting, told it and it noticed
insane
Me watching Gemini rants on twitter.https://cunnyx.com/i/status/2024936336038195382
Give me a keyboard sound that matches with this MB make it native
Video
Wow! The svg are really good now
Maybe it detected network change or ip so it showed you that
Anyone seen the gemini-3-pro-image-preview-2k (nano-banana-pro) model appear in battle mode? I've been trying for 2 days and I only see all the other models, but it never uses it.
@wraith garden@neon wraith@hushed sierraNote that Video Arena has been removed from the server. More information can be found in: #announcements message
(╯°□°)╯︵ ┻━┻
Nice
how can ai generate image to video?
is anyone else’s veo3 always getting too many requests
Singularity
The video option on the website.
man, I hate ti say it, but, gemini 3.1 🤮
i havent used it yet but is it rly that bad? 😭
i was kinda hoping 3.1 was an un-nerfed 3.0
me too, but, it's not better / worse than 3.0 as a coding agent, it's lazy.. so far I've had it try to generate 3 times a task organizer for me, keyboard shortcut optimized, it's failed 3 times in a row. all 3 are not functional
pretty sure 3.0 gave me some working versions. I'm using AI Studio
Click this button and you can upload an image that can be turned into two videos generated by two random video generators
im pretty sure gemini 2.5 pro was always like, super consistant at being just good
i never used 2.5 pro a large amount though so i may be wrong, but many ppl say that was better for its time
-# so, Opus-4.6 still is king, apparently
what is the secret sauce of Anthropic?
35ti parametrs
2.5 pro was pretty good before they started to try to release 3 😄
I think the secret is this
fr? 35T?
wasn't great, but pretty good
No just joking, but I think is 7ti or bigger the opus 4.6
Opus looks like someone giant indeed
agreed, but it depends. I have to admit, when you 'lucky' Gemini 3.1 pro can throw out some decent frontend results
like SVGs. it's svg king
but, aside from that, it's not so great 🤣
in some ways deepseek v3.2 thinking is better now 🥲
Guy willing or not, the best modal, svg, and world knowledge are still gemini so right
true, gemini seems to have a larger knowledge base. opus seems to use it's knowledge base better
Deepseek follows my agents much better than gemini 3.1 itself, looking forward to v4
same, I Hope v4 is good, and not like gemini 3.1 🤣
i.e. good at a few things, crap at everything else
V4 seems to be as smart as the glm 5, but costing pennies, that's my expectation
From what I'm the 3.2 with only the v4 tokenizer improved significantly the same when the 3.1 used the 3.2 tokenizer, so expect a similar jump for the v4
But of course, having a much greater knowledge of the world than glm5, when v4 comes out I believe that yes it will be the best open, but nothing much, the price that will be the luxury of it
Gemini 3.1 is Google spitting in our faces
Deepseek v4 better be good enough to make up for these disastrous releases
@echo aurora How come this keep happening? the select image section just loop indefinitely until my chat session failed due to time out, I dont even spamming any of the usage limits, and why the I can't post any feedback/bugs post on the relative channel?
Anyone wanna test my new api?
Gemini flash is weak
is anyone else running into the issue of claude opus 4.6 just thinking forever and saying the same thing and basically just thinking to think instead of actually executing? it has written as single line of code for me it just keeps lopping until the context window runs out.
What kind?
Ai
What model?
Dm
Is that a problem with arena or the actual model I have no idea
I had a similar problem so I changed my borwser profile and it's cool now. my guess is that mabye it's causing by any extension or something but I'm "not sure yet"
Am I tripping or did gemini 3 pro have a way bigger rate limit before?
I was at cap in like 20 responses.


what are u doin bruv
“Just a simple walk, great vibes, and unforgettable moments.
Out here walking with Neymar, enjoying the day one step at a time.
Sometimes it’s not about football or fame — it’s about the moment.”
“Moments like this become memories forever.”

What are you doing
Claude stuck on generating.
what are YOU doing
context limit allows thinking tokens to be counted in-context, so it's bugged because it behaves as a workaround (by api) to talk to use in thinking token, so the context window allows for following the copied behaviour of gpt5-3codex for being more talkative along the process of building but lmarena set a limit for cost to the so said claude (prob opus4.6thinking) doesn't run arena dry in 2 days
wdym
These are literally free and opensource..
#lion vs zebra
im offering in api and high rpm
looks like a reroute of other websites that give free api lol

do they give this rpm?
If you run it locally probably yes
if its free then you can use that instead of taunting taunt
i'll still take a look at it ig
atleast get a favicon ready
#video-arena-1 lion vs zebra
also the design is gemini 3 pro low thinking
ah thanks for the documentation
also this is your linked github
@river whale so?
vision model btw
prob scanned and stole an api key and in 5 minutes used antigravity, or another free garbage site-builder to quick sell it or steal ds ids
everyone don't click this it's a 2-month old namecheap domain (in the first place why would you click this) but also it's a email+pass collector
@hardy swallow
wait sorry for ping there's a bot for that
idk man i went on the website and checked the source code from github
so much negative energy from you
theres a lot of code in there but honestly looks legit
i meant to reply to this message btw not that other link
bruv hasnt made anything himself thats why he here yapping
free and unlimited for kling 3.0? the compute power to do that is imposible, furthermore, here is their pricing
@hazy forge
you mean this github? "https://github.com/kivestai"
called it
do another one right now without editing
actually hop on call, screenshare and do it in front of my own 2 eyes
arena playground
sure
@green yacht is also in this
he is pure racist
like lwk take a chill pill
coz people like these two scam and abuse the less fortunate
how am i scamming?
where's your github nerd
you prob scanned api keys on github
don't even know the models you're selling
@echo aurora are racist allowed here?
Hey guys, quick tip for those hitting token limits on the Arena with large codebases (Monorepos).
I've been experimenting with converting entire repositories into syntax-highlighted PDFs instead of pasting raw text. Since models like Gemini 3 Pro / Claude 4.6 process visual tokens differently than text tokens, this trick effectively compresses the input significantly.
I wrote a small open-source CLI tool (based on the DeepSeek-OCR concept) that packs code into a single PDF with Semantic Minimaps (UML graphs) to help the model navigate the structure.
The Result: I managed to feed a ~32k token repo into the multimodal chat using only ~19k visual tokens, and the models could still reason about the code structure surprisingly well.
If you are trying to test long-context capabilities without hitting the hard limits, this might help. I'm curious if anyone else has tried "Visual Encoding" for code tasks here?
and you come to racism on every response
He told me that it's powered by rental gpu so maybe you can take a screenshot of the website with the rental gpu?
my english is also broken, i'm just from a diff place, not american
Or if it's from another website ig you can ss that too
literally "free and unlimited" with the first proof of concept being "you ran out of tokens" and the second "log-in (give us your information and current traffic on the internet)"
no rental gpu site can offer free tokens to anyone at any capacity
keep yapping
He got timeouted
Okay let's move on from this conversation please.
hello
No profile theory
hello guys!
https://vt.tiktok.com/ZSmHWXUgX/lion vs zebra
watch it get lobotomized like gemini 3 pro again
A split shot (comparison): The bucket of the first fisherman contains several large, shimmering silver fish. The bucket of the second fisherman is almost empty, with only one tiny fish at the bottom. 3D animation, realistic fish scales, water splashing in the buckets, sharp focus, cinematic lighting.
💔
Illiterate theory
can you prove?
Hello
video in direct chat
Sponsored by ThreatLocker - https://threatlocker.com/ericparker
Official Discord Server - https://discord.gg/cqqWYDdcBn
Follow me on X - https://www.twitter.com/atericparker | atericparker.bsky.social
Disclaimer: The content in this video is for education and entertainment purposes to showcase the dangers of malware & malicious software. I do ...
😂
we are not deadass gpt
is gemini 3 pro image down ?
hi man. Does what you described help fix the infinite generation bug?
It’s a free country
Well then there we go
Cmon G where is racism ever tolerated besides 4chan?
It’s self-explanatory ( with all due respect)
Are you racist too?
No.

Join
What’s that?
Kivest Ai
I'm hoping they actually fix the verification thing again soon cuz it seems to have gone haywire again
Hope that helps
guys, is it possible to deploy app via arena to static address?
like aufall-parsing.arena.site so it won't be change each time?
Nu uita de Like și Subscribe dacă ți-a plăcut! 👍
🔴 TikTok: https://www.tiktok.com/@izenyxofficial
📷 Instagram: https://www.instagram.com/izenyxofficial/
#iZenyx #FIFA #Romania #trailer #official #eafc26
No. It's a preview address of a preview environment.
For deployment, you'll want to deploy to your own webserver of choice. Not a static webserver either.
As far as I'm aware, it's not even an internally hosted service for Arena. It's outsourced from what I know.
You can download the project and deploy it on cloudflare pages. You won't be depend on arena
►Discord's AI ID age verification vendor Persona just got hacked and exposed. Hacktivists revealed Persona was tracking and collecting your data along with ICE. Discord claimed to delete data in 7 days, but according to the age check/facial scan data leaked, biometric data was being kept for up to three years. Let me know what you guys think a...
Wow
Crazy
lmaooo
Fuh already scanned my face
Stupid online safety act
Rescanned my face heh
hey, does anyone know if opus 4.6 will be work again ?
does copilot arena still even work?
I’m thinking about ditching discord. I wish there was a good alternative.
yoo musk posting Lm arena 😭
in which language model you can chat with to put an mp4 file for it to summarize it?
since when theres video genertaion on arena
since its birth
gemini
gemini 3 pro? i put an mp4 file and says not supported
i wanna import an mp4 file so that a language model can summarize/analyze/make plot what that video shows, can someone tell me which model do i use?
lmarena issue
Only batlle mode is supported
And you cant select the model you want
bot
you?
@ocean ferry
3 is crazy bro atleast make it 4 😭
You should be grateful that they did not block your prompts
Put in a #1372229840131985540
i cant even do Image to Video now, 😢
@echo aurora Care to explain buddy?
rat
Why is codex 5.3 not on the code leaderboard ?
Ooh y'all buddies! cute!
Because OpenAI isn't providing access to it yet.
@pineapple Care to explain buddy?
[ Cisco Umbrella ]
⚠️ This site was blocked due to the following categories: Phishing
Sorry, 019c85f4-1bee-732c-b3c3-9f5c0cde924b.arena.site has been blocked by your network administrator.
Ooh y'all buddies! cute!
In all seriousness: this is a page/blockade generated by your local network/administrator's infrastructure.
The by far most likely reason for this: the preview of the application seems an attempt at making a web browser application, and therein, it tries to "include" external cross-domain located web pages with iframes against the headers of the request that served the page.
That's a fairly reliable way to get flagged as malicious.
It's a common tactic attackers use to display your familiar web pages but secretly overlay a "key catcher" or "click overlay" that redirects either your client or your entered information. That's why it's flagged as a phishing risk.
-# If you want to confirm this is actually true: Use a device where you can access the page. Enter any URL into the top of the "web browser" application: it won't load. Now enter the exact URL of the page you're on (https://019c85f4-1bee-732c-b3c3-9f5c0cde924b.arena.site/): it will load. Because the arena.site URL isn't cross-origin.
Conclusion: It's not an Arena thing.
It's a false positive, but for justified reasons.
Your "buddy" @echo aurora has nothing to explain.
-# aside from the part where this was never something they could be held responsible for, but I digress—I cannot actually mandate of you any technical eloquence or social tact.
Or maybe they do. I don't know what your potential love interests or relationships are like.
Keep me in the loop if you two get engaged. 💕
Havin' dinner with the fam yeah?
Yes brother
What'd'ya have? Was it good?
I want to wash my car. The car wash is 50 meters away. Should I walk or drive?
Yes.
But you do make a good point: if you have two cars, is it worth washing the other car?
A small language'd model full of large language!
Whats this...
@scarlet spire
A response to a user
As indicated by the fact it's a reply to... a user's message 
Hey I got always better results with Nano Banana via lmarena in comparison to Google Gemini Nano Banana direct. It is normal?
How to make img to video
Depends on whether you're using the exact same model. Are you using the Nano banana Pro model on Arena?
Through the website 🙂
He cant beacuse it says Error this image can not be used.
Then the answer is "you can't, because the image can't be used." :')
Yes. And also Nano Banana Pro via my Google Account.
Do you have a subscription to Gemini/Google DeepMind services?
Yes Gemini Pro
@blazing flame Note that Video Arena has been removed from the server. More information can be found in this announcement.
Hi
Hmmm.
When you're mentioning "better results", are you speaking in terms of the same result being lower quality but the same contents with the same seed and prompt? or that your prompt is followed better or more closely? or something else?
Also, can someone explain to me why Claude Sonnet 4.6 hits the rate limit faster than Opus 4.6? Opus 4.6 is more expensive, so I don’t understand the reason why the rate limit is so low for Sonnet
Because more people are currently using it. The quota is shared across the average user count.
Oh, that makes sense
It'll likely settle down in a few days.
I get results much faster. The quality is also much better and everything is more accurate in detail.
But does the prompt following differ between the two? i.e. if you request no green foliage, is it more likely to respect that when using Arena versus Google's own interface?
Am I understanding this correctly? Is the use of LMarena limited? I believe I can only generate a few images per day.
Yes, unfortunately. There's no free lunch, so the lunches are rationed.
There's also no free money, so, someone's paying for that usage.
Exactly. It seems that Google ignores certain instructions and even claims today that certain elements are present in an image when they are not there at all.
Okay. That's.. awkward but not entirely outside what I expected.
Are you accessing it through the Gemini chatbot interface or directly through the API? If through the chatbot interface: that's possibly why.
You could try specifying to the chatbot that you want it to use those literal terms as you mention them or that you'd like it to just feed it directly to the image model without doing any reformulating. It might take some experimentation. I'm interested in whether that actually works for you!
I want to wash my car. The car wash is 50 meters away. Should I walk or drive?
It varies based on current load.
Oops. That didn't work. Please try again!

Walking is the most efficient option.
Consider walking yourself for your general practitioner's office for a check-up, and describe your memory issues
Alternatively: the psych ward is always welcoming new members.
Fyi, video arena moved to the website. It's no longer on discord.
They didn't specify car wash 🫣
Just wanting 2 wash car. Not 2 wash car in car wash.
Exactly. I use the Gemini chat interface. The results are anything but what you hear and know about the great Nano Banana. And I am definitely not a newbie when it comes to prompting.
There is a definite distinction to make between Nano Banana and Nano Banana Pro.
Have you tried using the non-Pro on Arena and seeing if it matches the type of behaviour or results you get in the Gemini chat?
Hello 🤗
No, I haven't tested it. Especially since the result always turns out slightly different anyway, despite exact prompting.
@scarlet spire Is there a way to use my Banana Pro via another interface? Can I create an API with Google or something similar?
I mean i would walk to there since bread tastes better than key
Hey guys, apart from general image generation, video generation, and text generation, what else can AI actually do?
I’ve heard about AI automations and workflow systems, but those workflows never really did what I wanted. Today I explored how a solid website can be built using Claude and you can use Gemini to create games. That got me thinking, what else is possible with AI?
Earlier I always struggled to organize and present my notes well, but that changed today. Claude helped me convert my notes into a beautiful webpage and it is just 🔥
Openclaw
I asked two bots in Battle Mode to create me something and by my judgment one was better than the other but I only can find the two options to say vboth are good or both are bad.
Is Sora 2 better on api?
hey
I think there are no significant differences from lmarena to api, but in the sora app it has more functions
Hi
😊
This must mean that both are the same model 😆
usually it’s less restricted
at least when it comes to Gemini
as in it doesn’t care about copyright
or real people
Thank you guys. The only problem is I don’t see the download option on the sora videos anymore.
Yeahs this is true and I can prove it
Is that true?
Wasn't expecting that the same AI can be selected for A & B.
Like for example, if I use the following prompt in the arena:
But then if I use the same in mini max it works fine
Let’s see if it’s the same with copyright or people
@glossy umbraTried choosing one of the both options and then it crashed on me.
It also turned it they weren't the same AI's either...
chat looking kinda dead
How many hours you can drive, before your body gives up?
https://youtube.com/shorts/sJ86kavWsss?feature=share
Ever wonder what happens to your body during a massive road trip? 🛣️ From simple fatigue to dangerous micro-sleeps and straight-up hallucinations, here is exactly how long you can drive before your brain forces a hard reboot. Drowsy driving is no joke! 🛑
Have you ever experienced "highway hypnosis" or micro-sleeps? Let me know in the com...
i genuinely can NOT be the only one who absolutely hates these captchas
i killed the chat again didn't i
arastradero is a seed 2.0 model?
p
Looks like you generated that on openAIs first ever image model
Slandered by one AI robot and misquoted in a news article by another, US-based software engineer Scott Shambaugh has made it his mission to become the cautionary tale by which we start to take autonomous artificial intelligence seriously.
#technology #AI #agentic AI
🔔 Subscribe to France 24 now: https://f24.my/YTen
🔴 LIVE - Watch FRANCE ...
😂😂😂
This is so great
that's true 👍🏻
I deleted the comment thanks for the reminder @toxic verge
@dense bone Video Arena has been removed from the server. More information can be found in: #announcements message
guys is there a way i can use this website api?
No, not have api here
You can use aistudio and glm api free
Is ai studio free?😮
hi
weird. could it be because of the screen resolution not fitting the other 2 options in?
3.1 pro?
Honestly that's the best model better than opus 4.6
On some things definetely
On making games it's the best in my opinion
Btw did you know you can use it for free on codex?
There is limit
No Gpt 5.3 codex xhigh
all of my friends are saying its dogsh*t and some of my other friends are saying it's the next big thing
depends on the use case i guess
The friends that said it's dogsh*t didn't even tryed it to be honest
Hi
bro i was gonan get claude pro couple days ago but was thinking of maybe gettign gpt plus instead
Hi
Why?
This is nothing new at any stage. They've been doing this, as have other companies, for decades now. It's not exclusive to the online age.
gpt plus is better value for money cuz they also give image + video and more usage, and i think for normal tasks like web-search and explaining things, gpt is better at it. claude is only for coding
if you long-tap on the video, does a context menu open?
Why not Gemini or Grok then?
groks model are honestly kinda bad and im not looking for nsfw stuff, gemini was an option i was considering but its not as good as coding (only frontend)
Up you then. I don't think Grok is bad but I don't code to be fair.
gork
I'm gonna add gemini 3 Flash in my api so make sure you guys join
Let me try
Would u look at that worked Ty
Awesome!
It's interesting
20 requests per minute is great
The problem is that the models available are dumb compared to Claude 4-5/6 and GPT 5.2 Codex
its only open models mainly for now cuz its runned of local gpu
and hoenstly if there was those big expensive models like that im pretty sure people would abuse it to the 4th dimension
What a coincidence that we wrote to each other at the same time, you to reply to me about that and me to reply to you about something else in a post
Hi
welcome
i think claude models not good for vision 
Did someone run this test on 3.1 already before?
Why does opus 4.6 not use the arena ai tools to edit etc. so often? he just fakes it and says "Build successful" or "let me edit the code" but he doesnt change anything
Bruh
And kimi 2.5 and qwen 3.5?
I doubt very much that any of these 2 will get it right, but just look to see what happens
Dude I think claudes' vision is really good, but gemini always surprised me more, but in this case it miss miserably
gemini 3.1 is so good for image to UI
Day to day tasks I prefer claude, gpt is so protocole
admittedly nah, the planes in the image are pretty-almost symmetrical tail to nose. :')
/Peshawar in Winter
In Peshawar, when soft snow starts to fall,
It kisses Qissa Khwani’s ancient walls.
White flakes rest on the Bala Hisar high,
Like whispered dreams from the winter sky.
In the bazaar, people laugh and run,
Wrapping shawls when the day is done.
Warm chai steams in every hand,
Spreading comfort across the land.
Rickshaws hum through snowy streets,
As every corner with joy still beats.
Oh Peshawar, so lively and warm—
Even snow cannot calm your charm.
Look at this. It’s nuts
Creat video
wow gemini 3.1 can be outright scary, if you hahve no idea what you are doing / don't check the code 😄
i had it make me a quiz, q/a. it worked, that's the good part. the bad part is, it decided the answer would be 2, 3, 4, 1, repeatedly, for all the questions, withouit telling me
since i didn't actually know the answer, i assumed it was correct, until i examined the code 😄
i can only imagine the amount of people putting this stuff in production assuming it'd ask you or at least tell you it's a placehholder
You need to make the aspect ratio 16;19 or 21:9
Any one know is there any prompts or tutorials for Underwater metal container home video making ?
Petition for arena.ai to test small models too, especially the ones that are really popular on hugging face eg Nanbeige
Creat short videos
You have to use the website https://discord.com/channels/1340554757349179412/1475080934872191028 or #announcements message
Please create one video
Okay thanks
You createing video
Creat video one video
Ok you create short video
Yes
If ya wanna babble on about this, wouldn't it be nicer to stick it in #ai-creations or the likes so general chat and your posting don't interfere? It's a lot of vertical real estate taken up to scroll six pages or eight pages of text and images from just the last few hours. 😅
Is there a way in battle mode when generating images to use nano banana?
Did they lower the rate limit on opus 4.6???
Yes but random
Claude pro offers 3 prompt per day btw
Gpt plus offers ~ 200 prompts per day btw
i got claude in the end lol 💀
its only 1 motnh though i'll switch and try gpt too after this month
You get more claude quota in google sub than claude sub lol
😭
Claude is usable if you get 100$ sub or more
i alr used 80% of my current session limit 🤑
/lighting over hawaii
Hello people. Hope everyone's doing good.
Hell, llmarena offers more opus than the claude 20$ sub
Claude 10$
Wtf is this bs
I overpaid
😭
Geniuely unfair theres a promo thats not on by default
Uhh
How to access sora 2
oh hell nah 😭
yeah i should have goten gpt plus
$20 plan is practically non-usable for coding
Mhm
Gpt has really good quota (best out of all)
Google has a decent quota if you use cli and antigravity both summed up
?
what is the limit of opus 4.6
where is this number coming from?
From my unfortunate experience
@surreal zephyr
I mean I have the pro plan personally and I've had fairly generous usage limits for Sonnet 4.6
For Opus 4.6, yeah usage is terrible and I'd be lucky to have 1 full coding prompt with it
but it is an expensive model to run
U can use max5 limit in 1 bigger prompt
nono
...
Could someone tell me where all the image models ?
I didn't have it 10 minutes ago either, but now I do
please dm those apps to @echo aurora so arena can take action, they shouldn't be on the app store
please note that arena currently has no official app on any platform
i found them on the app store site, i'll dm them to pineapple
Hey thanks for pointing out, this is something we’re looking into.
i did send you those links, i'll see if there's anything on google play and i'll report those to you as well
It's not a fixed number. It varies with available capacity.
there are a few fakes on google play as well
yeah i know, i just wanted to know what minutes
Hi, I am here to learn more about IA
There's no "the number" in such definitive way.
such as this one, please do NOT download this app and if you have it, please delete it. this is not official, they are not using the real Arena branding
they have in-app purchases as well, do not buy anything and please report it to your bank/credit card provider if you did
they claim to be "Arena AI" as well
this is also not a real app
over 150k people have downloaded these fake apps
planning to add some good close-sourced models
ANY ADVICE ??
I've just completed my 2nd website (🤧i'm a beginner ) which is a menu menu site and wanted some images for foods but 🙂Gemini's behaving like my ex doing everything except the one thing I actually need. ?
for reference here is my site's live link https://mocha-loft.vercel.app/ (still have some work to do tho)
Generated by create next app
Bot 🤖
AI GENERATED APP LOOOL
Is minimax m2.5 any good?
It’s the cheapest option right now, but YouTube reviews are saying it’s really bad
Gmm
Nice big improvement. But why this spect ratio? What’s prompt?
imagine THAT lmarena was made by EA
lol
How can I remove photos from chat with bots? I only have option to archive
hi, can someone help for error in chrome?
Clear your browsing data.
Hi
generate a image of a very modern desktop lamp
is there a way to choose models?
thank you for the help.
and pretty sure this app has to be removed from play store
learningdigitalworld
dance
kiri
when doing image to video generations, is it me or it is really quite slow. takes about 10 minutes to generate...
It’s slow
is there way to remove the limit of A.I generations? no pricing info too
The cheapest and the most effective method is just to buy subscription to your favorite model or find a wrapper that could do all
Easiest & most cost-effective way to use Gen AI. fal.ai is how devs integrate dozens of generative media models. FLUX, King, Hailuo +600 more
@mint bramble
yup
thoda upar dekh kisi bhi user ke naam ke aage arna wala tag ho use adopt kar le uspe click karke
kk
done
let me guess, this bro is from the same dev with Flux or something ?
how to join
is anyone from china
gork
@zealous cairn Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
So Max simply selects a suitable provider and model for the response
Yeah it's worth noting the rate limits for particular models can change over time
it selects the model it thinks it will answer your question the best
Woah how do you this???
Dude I just broke Gemini lol
Gemini by far is the stupidest model I’ve ever used in my life. I can’t lie.
We all agreeing Gemini is benchmaxxed in everything right
Like
Right behind GPT
There is no way Gemini is that good
Claude completely outpaces it
“Woah it can make cool SVGs!!!”
Who the hell is using ai to make svgs
💀
Man, the funniest thing happened to me
If you wanna see DM
@quiet isle Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
Is there a way to share chat from Gemini?
so will there be an update in regards to the video soon
We need more leaders in this space. Honest, brave & courageous people who speak up for some of this none sense
hello
Well 3 per 50 minutes is asinine
Wtf is the point of even using that model then
There is an option of not using it. Even 1 time is better than not being to never use it in my opinion.
It’s expensive
And it’s paid
What else can you expect
The point is they could've at least informed us of the change
Rather than unexpectedly get surprised by the change
I hear what you’re saying, but let me ask you seriously let’s say they did warn us. What would have you done to prepare?
The main point of lmarena is for the ai to be secret, we should be glad there’s a direct mode
Hello
And this is why I'd rather not share my grievances with this website
It’s not an agreement many people feel your frustration.
And you have every right to share your opinion, and it shouldn’t matter what people like me think or say it shouldn’t change your opinion on what you believe lol
If anything, it’s my bad so I apologize. I’m just trying to show an alternative humble perspective.
Anyways, I gotta go for a little bit. I’ll talk to you guys later.
3.1
Jaminai
new guy here! hows everyones day going?
I wanna give middle finger to those who put infinite securities verification check loop, just ruined my projects flow
@honest cloud Note that Video Arena has been removed from the server. More information can be found in this #announcements
Maybe it's down to the attitude... in respect to your ex. 
Money ain't free! Someone's payin'. That'ri't
subtly re-aims your finger towards the group of bellends attempting to create exploit bots for the web interface to gain free access to paid models on a systematic basis
The rate limits are, as far as I'm aware, set on an automatic time-windowed basis.
There's benefits and downsides to this approach. The biggest downside would be catastrophic underestimation of future load which can lead to occasional bursts of unresolved prompts if the load is higher than the distribution anticipated.
That would ring true for some current-day issues that are yet to be heuristically steered
🤧🙃 then I wish you spend a day with her she used so say stuff like one day she was taking a bath in her bathroom then a car crashed to her then she swimmed from that kitchen
Either you're incoherent or making fun of someone's traumatic life experience and I'm not sure which one's worse 
yo guys, does anyone know how to embed ai voice agents on a website for free
you're indian right
🥲🥲 I was kidding how the heck a car can come in 3rd floor washroom???
my ex used to use that emoji combination
😭😭 bro's catching Indians from emoji
if anyone knows, please let me know i have a fire idea
indians are easily recognizable
All I understood from that was
You should spend a day with her then! She once told me she was taking a bath in her bathroom at one point, when a car crashed into her house and the building collapsed all the way into the kitchen, where she then lied soaking wet!
which was incoherent and disrespectful so I took the benefit-of-doubt approach
me btw
And your grammar patterns. They reveal how you match your native language to the written/spoken one, through entrenched ordering
for example: do you put the adjective before or after the subject? Do you commonly ad-hoc a reasoning or are you likely to prepend it to the structure? (thinking structure habits)
Use any research AI for that
FRRRRRRRRRRR
done it, cant find a sustainable option
Oooo got you
/how do I generate video
The video generation moved to https://arena.ai itself
Whatever feels right I mean I used to learn English 4rs ago now my gut is everything
how old is you
(im black)
speakin hood english if you dont get it
Sounds 'bout right—not a single apostrophe in sight!
I heard Higgs feild doing great 🤧
nga this is discord not Linkedin
Naah it's chill
20 🤧 and I'm telling my age so it means I'm a guy
Oh wow so young. I'm already 24 apparently.
you could be A trans as well
Although I really shouldn't be surprised.
trans is an adjective, not a noun. You use it to sub-define something else's properties, such as 'person', 'girl', 'boy', 'man', 'woman', etc
So I'm just curious can you tell me how's life from your perspectives @plain gull
18 sets of 12 synodic moon cycles?

