#general
1 messages · Page 261 of 1
Gemini photo generation works well now ❤️🩹
⛔ @left lodge Whoops! You need the Manage Messages permission to use this command.
Seed 2.0 pro
hello...just a quick question...why I can read this message...
thx for your reply....I would like to fix it..ciao
Hello, I was able to generate videos from my images, but now I don't have permission
Yea, Cookie delete? Cause when i change it to phone it still the same issue hahaha
Did you fix the problem
Did anyone test seed 2 pro in zenmux?
Did you, or how didyou fix the problem
🍪
ok so we will try to delete cookie...thx...
It depends on type
Even some are not fixible on js
Us
Cause there are save on your accounc
Kk
Make sense
let me try it
ok thanks
Check what http status u get when u receive "something went wrong" error in Dev Tools -> Network, on pc
check bro
@wispy nexusNote that Video Arena has been removed from the server. More information can be found in: #announcements message
chrome
@echo aurora just a question are you guys gonna add gfemini 3 pro deep think '
it is , not fully optimal
BUT we can but faith on the Dev
since they are rich some how
to given us lot of perks for free test
so
let hope they can do better
AI are limitless i just wonder where will they continue to test, im here to help
is it just me or gemini has been acting more stupid for the last week lmao
you not wrong
it is how gemini work
the more gemini have more side app , the more user. The more share
Then lower the context
the dump Gemini gets
it become worst than 2.5 in the past
No they won't. It's likely too pricey.
I would be surprised if they bothered considering that gpt pro and x high aren't added.
Sillly question, is it possible to upload audio to arena.ai and have it transcribed & summarized? If so what’s the limit we can upload?
bun here is the thing its not pricy
because google usually asks LMArena to rank there models so they can update and train them from real world votes
Unfortunately, not yet. @echo aurora would be lovely if you can make this happen.
Does this apply to open ai models too?
how i make the video
Tap the video icon in bottom left.
yes arena have a discount but its still pricy especially for ultra tier subscription
and arena has never added any pro model
Closest we got was opus 4.1
How i make video from picture there is no option /image to video??
if they add the pro model of gemini then they must also add pro model of other companies, for fairness
Opus was anthropic's pro model until price dropped with 4.5
or the leaderboard won't be fair anymore
I don't think it ever was fair. I have seen gpt 4o get more votes than O3 mini in math.
also if its already struggling with the long thinking of opus 4-6-thinking the deep think version of gemini wouldn't work lol
on arena
Good point.
i forgor to check discord again
check for this
(have devtools open before triggering the "something went wrong" error)
no google gives them API keys directly
Then i don't know cause from what i've heard before that wasn't the case i mean not for free
hi
they give them credits is what i have heard
like a fart ton of credits
I see, thanks
bro what?????
Thank you!
the funny thing is that the actual message generated fine without erroring
"if it aint broke dont fix it"
hell yea
While the internet obsesses over Gemini 3's benchmark scores, the real revolution is hidden in the 100x reduction in inference compute and the new 'Aletheia' agent. This video breaks down why "thinking time" is the new scaling law and what it actually means for the future of AI development.

I want QWEN 3.5!!!!! NOW! NOW! NOW! NOW! NOW! NOW! NOW!
I want..... umm... "vierra" I think its called.
@dusky vortex Please check my DM
Does anyone know if there is a set list of words that triggers the terms and conditions anyone here?
Can dm me
hi
[Message hidden]
-# Discord now requires ID Verification in order to view certain messages. Learn More
What you gonna do with it?
agentic ~ harnessing 💖💖💖💖💖🧚♀️👰♀️👩❤️💋👩
Agentic breeding vibes 💀
Yeah I just reacted to those emotes.
any anthropic api keys?
ayooo
ill take that as a yes give me one🙏
someone refresh my memory what to add to my prompt to generate 9:16 videos and not the regular landscape type youtube videos
depends on what youre using
if using veo3 or kling generate first frame with nano banana thinking (in gemini app specifically)
Does anyone have a project idea or an active project in progress?
If you need technical support or a developer, feel free to reach out.
Hello anybody online?
The human verification is really really really outrages
Me
I sell AI models4cheap, don’t need devs ( i think )
-# not tryna be rude
glm 5 🔥
H
seedance 2.0
.
Too too too
I hope qwen 3.5 4b is very good as was qwen 3 4b at launch
14b, 30b too
Qwen3.5-9B-instruct is already confirmed.
a Qwen3.5-35B-A3B appears to also be happening, but people aren't sure on that one.
apparently there will be omni-modal models and basic text-only instruct models at launch.
Thats what I can read from the update they pushed to huggingface anyway.
The bots are out today
I'm confident our overall site reliability/model performance will get better over time
I'm late which bots
can anyone tell me how i can make frame by frame traditional anime rather than just 3d like movements with Sora 2? it always generate 3d animation for me
creepy bro
Can anyone tell how cwn i use claud or any other best model to build a application
Will it be at least better than the glm 4.7 flash?
Send again your animation with nano banana
@mental bramble @tribal ermine Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
?help
There is a Chinese website that provides it for free no limits but I forgot the link
Doubao?
Hm ok
Vpn doesnt change
It does
Menu regional restriction wont appear If you use VPN
But If you use their platform too much, I think phone verification appears
we do not know until its out.
never trust benchmarks, teat it urself on a problem u tried before.
qwen3-coder-30B-A3B doesnt have the best benches, but is real nice to use ~ its got a surprising amount of programming knowledge, even in Godot Game Engine (not that popular, much less so than robbbblox or whatever)
Yeah, They updated their security
They noticed lot of americanese using seedance 2.0
I hope you have a alternate chinese phone number
I could use tiktok
ok
A quick look at Claude's generated C compiler.
🥇 Join the community https://www.youtube.com/channel/UCXzw-OdotBUcNA9yhuYQBwA/join
💬 Topics:
- AI agents redefine success criteria
- Autonomous iteration over functionality
- Incomplete compiler dependent on GCC
- Training data limits reasoning ability
- AI hype exceeds practical results
Doubao sucks
When did OpenAi provide codex-5.3-max-high for free on codex dang I think it have limits
I found an unfiltered ai image generator bro
i have a question
for lmarena, does top model scores decreasing
mean that there is a new model coming out that is beating them?
like if its an elo system, and theres secretely like a gemini 3.1 model
wouldnt first place decay?
also how does lm arena update a new model so quickly?
like lets say when 3.1 comes out, will they have already tested it and benchmarked it under a secret name?
i have a question. where can i download lmarena for android??
it's currently website only
I know so much that I think the m2.5 in my opinion is the best open model of all, but I use the z.ai tag because I think it's more tuff
okay. thank you for the reply.
you mean for coding
also long time no see so hi~ 👋
cuz i think for general writing it's k2.5
No man, I use it for everything anyway
Code, law, role play, documents
I found the m2.5 very solid, that's it, I feel it was better than the gemini 3.0 pro
whaaaaat really?
It's really good, I sent a code that gemini couldn't solve and it solved it at first
I also used it in general use with agents for document creation if I do very well
@grave knoll Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
honestly yes, but try glm 5 in my opinion its even better than minimax 2.5
Hey gng what do we rate perplexity
It's not a llm but is it any good
There image gen models lol
Kimi and gpt 5.2 thinking is included W
bro i wanna use seedance 2
Does anyone have problems uploading photos lmarena?
bro by a lot, minimax is good but it feels so benchmaxed. glm is way better imo
Im getting this errors and the image wont upload
me too, but doubao notworking with urban vpn 🙁
lebron or jordan
Seedance rocks what the hell
It’s lame
It’s just not really good at all
oh boy this era gonna be smth else
what is the best niche for 2026
I been saying ppl sleeping on ByteDance
cuz they been sleeping
Is it restrictive?
Wow pretty good
yea so good seedance 2
Only problem is every time I tell myself it’s always gonna get better. It’s gonna get better but then it comes out although it improved it’s like you could tell it still has ways to go. lol
I don’t know if they’ll ever will be a destination or if we’ll just get there one day and not even know it
our standards improve after seeing
Yeah thank god for humans who set the standards high in cinema and animation and film, making in general
9/10 i use it but claude 4.6 opus should be free
Hi guys
Hi
hi
hi
hi
Ig
video
I need video
HOW?!
Hi
hi guys
Hey
I like how he frames it as a repository of human knowledge
But unlike open source software, the barrier of entry is much higher
Is AI an increasing cost industry in the long run from an economics standpoint?
thinking what happen if lmarena.ai was owned by EA
I’m not sure that’s a good question
Possibly... he compares it to open source software, but I'm wondering how much that analogy holds
she's jelous 🙁
@latent fog @rancid knot Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
Jii
I tried to use the new feature of uploading a pdf to a chat, for example with Claude Sonnet 4.5, or even Opus 4.5, but I got the error "Something went wrong with this response, please try again". I looked up in the Network tab of the Developer tools and found that the error stated a3:"Could not process PDF".
Seems like pdf uploading is broken
Do you speak French
is there server issue? i keep getting this error no matter what
I'm writing to French
Create a cinematic short video from this image. Keep all the Arabic calligraphy text completely stat
@brave isle @fleet swallow Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
Rate my video
could you help me pls, I can't create any image on the website cuz of this error
hi everyone wish to have a great time here
this was my prompt, is there any issue with it ?
16:9 aspect ratio gameplay screenshot Kingdom Rush tower defense, Chinese environment, bamboo forest, mossy stone lion statues, template game, blue brick roofs on stone walls, pond with lotus lilypad & gold fishes, pebbles. Bamboo fortress dart archer tower shooting at enemies, Chinese guandao sentinels barrack tower with 2 troops attacking enemies, Chinese Daoist tower with a Daoist sorcerer under a cherry blossom tree summoning formation magic attack, Catapult tower slinging flaming rocks. The enemies are Japanese ninjas, samurais, Edo era swordsmen.
did the sensitive filter catch something by false positive here ? if yes then what ? I don't get it. what's wrong with this prompt ?
Helllo, here for make vidéos
replace the car with my car
ohhhh my godddd can i have them
Can you read English
@floral copper Stop
and Stop generating AI prompts for AI videos with AI
@floral copper Note that Video Arena has been removed from the server. More information can be found in this #announcements
@young wadi Note that Video Arena has been removed from the server. More information can be found in this #announcements
funny how flash beats opus
Inbetween the lines
flash doesnt beat opus
Wait is left the minimax??
Some people have said that opus 4.6 was supposed to be sonnet 5 but instead they just changed the name to opus to get more money?
Also i think sonnet should be released earlier than opus...
It usually is
no its 5.3c
/Animated evening scene near the mango tree. R সিAnimals slowly leave, while a small rabbit stays back looking thoughtful and suspicious. The fox remains near the tree. Calm cartoon animation style.
Dialogue (Bengali – Rabbit):
“কিছু একটা লুকানো আছে…”
Camera: medium shot, gentle pan
Audio: dialogue only
Aspect Ratio: 16:9
Idk i think the opus 4.6 (non thinking) is pretty good advancement
I did this is 3 prompts
I let the first ai create me 3 prompts for the opus and than pasted each prompt one by one
You can litterally craft wooden tools
Craft a crafting table
Needs work but its good
Claude Code CLI is the best though for programming anything, it just puts the files on your computer yourself lol
Gotta pay though
of course it should cause they use algorithmic intelligence while anthropic, gemini, openai are just using raw power
but once those three start also do that it won't be beaten by any small team
cause 1000x raw power is still 1000x raw power lol
@ivory surgeHello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been disabled. Video generation now is only available through the website here -> https://arena.ai/?chat-modality=video
Opus 4.6 has 1 million token context which is huge, meaning it doesn't forget what was said by the time its done with the files. But also have a check your work thing which leaves mistakes out of it
It feels like a lawless land here on weekends
@balmy horizonHello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been
Hey I got to know about this today
I wanna ask how many times can I generate videos here ??
And are there any limits and of generating results ?? Or credits or something ??
And then ??
ok
?? 🫠
let me try
you can't gen videos on this discord anymore, you can do it on video arena 3 times a day
Only 3 times a day 🫠😐 bruhh.....
you know, video generating is a whopping system eater
“ / “
What about I have 100 gmail accounts and I use them all as much as I want 😏👀
Just saying
go on and hit the 100 drunk captcha wall
👀😏😏😏😏😏
TRY AGAIN
come in dm @sturdy mica
try again log out / log in
Use a different model or try again
ok let me see
also check cookie permission and you can try clearing them if nothing helps
ok
how to clean cookie perms
bro can lmarena can also add stop and end button in chat area
to fix this problem
permissions in the browser settings, you can clear cookies for specific pages somewhere here (firefox)
ok
not sure if this would help alot
yes
bro how lm arena get the unlimited creadits of models ?
bro and i deleat the cookies
again show generating
?
?
😜 😜 😜 😜
money
weird huh
I'm the ohwner
who tell lie `is G@y
You really think owner would be here to listen to bugs
There is #1343291835845578853 where you can report the bugs
Some are there for months already not fixed
i neeed to fix it
Do it than
becauce i am making a huge ai project
WOOW
you want to see
??
is bro doing this?
4 new models next week lesgoo
ahh
ok
Nothing is guaranteed it will work 100% here
see this
@echo aurora i know its too much to ask for but can we have one small stop button like if something is generating something and i need to stop the generation mid way is it possible to do that
Your chat could break and everything you've said will stop working because of infinite generation bug
Woow so original
Why can't the video generate model be selected?
Multiple ai models 😮
what is this
you also add this models for free in your wesbite
Sonnet 3.5 💀
I need free ai video generator app anyone tell me app name
where's that
2.5 flash preview 💀
Please
there's no free ai video generator in the next 2 years
it have all
oo
I mean unlocked apk of video generator ai
see on youtube
please, give up, it's for your own sake
Okay now ask it how much is 2+2
even worse
What is this
i have a file to make unlimited video gen even batter then veo
see
Give me
@queen veldt
what are you talking about
why
Yeah but you know you have to pay for API
For those models
And i doubt you know about it
kirk
Since i see some models that don't even exist for 2 years alr
you can also add this models in your website for free
I make education video on scince topics and i need video releted to scince topic
Gemini 1.5 flash doesn't exist bro
but you onyl have 200 per month creadits
make wiht arena
These are all outdated models something is wrong here
Arena is app or website
i also add soon the opus 4.6
Is it free
yes
what
You can't get any model just like that
Arena is app or website tell me
website
I don't think the opus 4.1 api can be called anymore using the API
I think those are pre programmed answers or something
real models
no pre installed answers
tell me what i give to it
What you are doing here?
it also have gpt 5.2 and pro
Say "Divide me the 2 with 5 and give reasoning steps"
Claude 4.1
ok
Which ai
What is your ai name
That's the infinite generation bug
how to fix
Copy the previous text you and ai talked about and open new chat and say "Me and ai have been talking about this, i need to continue this conversation:"
BRO HOW TO DOWNLAORD CODE
And paste as much as possible (it won't handle all at once tho)
why 4.5?
are you a human
Tell me this ai name
what the hell are you talking about
I'm just curious how you have infinite api from Claude
It seems like it's working but you'd have to have some api key or something?
You can't just tell ai to create you a fully working ai models without any api keys lol
😭😭
no i am a man
i am building
?
Ok
no
claude
it have real
some claude model
send the website to talk to ai models for free
ok
Fr he's lying
Send the link
dont copy my idea
Looks cool
who copy my ide he is g@Y
1 more llm wrapper until we reach AGI
i need ot fix api
What is special about claude ai
dont use the api
but we use this models in yur website
like craete api and use it to you wesbite
;>
Why does it have outdated models tho
interface is good or not
It is posible to make unloced ai video generator?
bro
just buy gpt sora
and make video simple
What happen?
2 steps only
I do not have money😭😭 i am poor
so use the lm arena
Do you guys know where we can use seedance 2.0
broo
Como puedo hacer videos?
visit video arena website
likely not even close to codex
My editing skills are intermediate level, can you guys please support me and rate my editing from 10, for voice I used ai tts.
6/10 edit level, but i REALLY don't like this type of some sort AI videos
High > codex upto 5.2
Don't know about 5.3
why do you even need pelican made with svg?
Voice generated using ai, but the graphics were made by humans and ai.
It's a stupid test. No one needs it.
yes, i know but people don't like that type of videos
less AI in AI videos please
ernie 5 is the best model overall? crazy
this test is very idk "from single side"?
ernie 5 >>> opus >>> 3.0 flash > other
looking at that bench
fym gemini's answer is way better, opus is lazy here. both correct tho
He did answer you with simular tricky questions as you asked lol
opus rephrased the same question 10 times, gemini came up with actually, creative, different but similiar questions
Did you ask "mostly code model" for creativity? No?
Your answer is no
0 sense
you mean CREATIVITY-ONLY model?
💀
the whole point of opus is creativity bro
it sucks at coding
Wdym it sucks at coding?
dementia + hallucinations + slow + expensive
i wonder why 5.3c doesnt have such issues
And power = more hallucinations
no
it does have
creativity = more hallucinations
no.
Hello guys, Lmarena have plus plans or pro plans?
Okay so how did i almost build UI chat for AI with claude 4-6 opus?
WIP (work in progress):
that doesnt require almost any real coding
get it to write realistic suspension solver like codex can, ill wait 🤣
html isnt even programming bro
This is impossible you did watch everything in 20 seconds
video is 60s
react
lm-arena restrictions are horrible
i asked codex spark for summary
Use your eyes
Okay bet, write copy of my WIP version and we will see
How your codex is better
thats literally 1 prompt of work
Show me
Opus looks better but Ernie did better for its price. I believe glm could have done better if it thought longer. I clearly mentioned high quality in the prompts. Apparently glm took it lightly.
Did you made it? And thanks for saying this, i was sleepy while making site and i don't know much about react exactly
all works btw
this is just markdown, show me math and code blocks and sliding header of code blocks, syntax highlight, image support, file support, openai compatible api support, models auto fetch
does this got streaming?
Hi
Lol
sending this cz this search filter is too good to gatekeep (list of all REAPed models)
https://huggingface.co/models?other=arxiv:2510.13999&sort=created
and everything i made is just basic chat UI and still not finished, im polishing it right now, like custom emoji (discord one), fonts, like no jitter while generating or regenerating
well the one gpt made has allat xD
rip opus
syfm
every single blocked message i unwrap from you
is saying opus is bad gpt better
opus being mocked by gpt and even gemini is js sad
your life is sad
keep overpaying for wannabe gemini 🤣
you genuinely talk like a 12 year old r.blox kid
And i will add later multiple providers support at once, ui bars at the top, chat history, account registration (with database), python code execution, and site builder, and animations, i absolutely like animations
My third animation attempt, just for test
lol the gpt one already has multiple provider support
gpt? i mean like ui where you add multiple providers and later use them, like top bar in arena with model selection
zachem ty razgovarivayesh s etim chelom
Why does opus always time out
yes it already has that
but u could prolly use openrouter for it
bad model
all he does is say opus bad, gpt better
because "team" can't change time limits they are a bit lazy
он менял своё мнение 10 раз на день
what a shame, it works amazing on poe.com
Good in coding. Mid everywhere else.
It's good at writing I think.
i think coding should be the priority anyway
its the best usecase of ai
not only, think about text writing, graphs... aaaaalll that simple stuff
like people spend time on work to do it
office rats
I want scientific explanations.
GaMiR: Sorry i could not assist you properly, tell me more information about questions you provided.
statements* maybe
maybe, my english is ass
and im forgetting russian over time 😔
(i moved a decade ago)
or somethin idk
sad... Talk with russian people, films and videos are bad
yeah recaptcha is 50x worse
@sleek phoenix The error genericity issue aside; You'll keep running into odd problems and limits/restrictions if you're keen on pasting everything in a log file including everything that is not useful to share. Ever single character there takes some fraction of a token in your context size window. It therefore goes without saying, but: Unless you can't read the log at all and are somehow impaired from cleaning it / making it less verbose, there's no excuse but apathy or indifference that can explain the need to paste your entire log on a free-access service.
Please be courteous.
We all share this space with grace from investors.
sometimes it takes me 20 mins to solve 1 captcha 😀
Robot
Like opus's answer looks like school homework instead of scientific research.
Vs gpt
Search models kinda sus
belive me
😭 google likes to give me captchas that dont make sense (those which arent clear with what boxes i should include etc)
It won't be any better with text. Opus is very hesitant with tokens.
whats the point in that
i understand sending prompts
but if its being abused
then why use captcha for voting
If the ReCAPTCHA is too difficult for you, there's always the accessibility alternative of audio-based challenge.
Is it faster?
Because it directly combats the abuse propensity and potential vector.
You'll have to find out.
i dont even have earbuds in 99% of the time and i dont use my speakers
@thick urchinHello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been disabled. Video generation now is only available through the website here -> https://arena.ai/?chat-modality=video
how?
ужас короче
Y'all trust arena vote based leaderboard over artificial analysis?
🥀 🥞
Hello are you just a mod, or you got same position as pineapple?
Pineapple is admin.
Ok
just discord mod
does anyone ever wonder what the ai video prompts on discord are used for
Okay bye everyone, и пока nickname
ai studio A/B testing again?
You already commented about it: you're having trouble blindly passing the ReCAPTCHA.
That intrinsically indicates you are being required to use cognitive skills to solve the challenge, which bots generally don't have.
@quartz light
It's not about the specific implementation, indeed. It is about the part I mentioned subsequently: the challenge.
The industry term for a user-interacted feature that serves to inconvenience the user but block machines produces individual instances of such to-be-solved "things". Those "to-be-solved things" is a "challenge".
Turnstile to be exact.
Oh btw i'm here again to say this
Hola
btw does any1 know when the next gemini 3 varient will be released?
is LMSYS doing the windsurf arena? or did they just copycat
@echo aurora still no feature of using direct video and side by side, jeez its been ages
Heads up: this account joined here an at least one more AI server just to post this. Twitter account they're linking has 0 followers and 0 following and also happened to make their first post today, this is very suspicious
Going to remove that message & will keep an eye out.
Sorry to say can't say I'm aware of plans to bring Direct/SbS to Video Arena at this time.
Note that Video Arena has been removed from the server. More information can be found in this announcement.
thank you man
i guess the best free video generation workflow before the official release of seedance 2.0 is grok imagine and nano banena pro
@edgy obsidianHello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been disabled. Video generation now is only available through the website here -> https://arena.ai/?chat-modality=video
would be useful
5.3
its definitly 5.3
the progress is real
Idk im using 5.3 in conductor and it's dumber than 5.2
I saw alot of difference between 5.3 and 5.2, i did test, i used it like i used 5.2 and it really did better, its definitly smarter
that's weird, but with the same prompt 5.3 is doing better
than 5.2
idk let me hop on codex and see
Maybe it's me asking over complicated stuff
(proper design)
i don't think its the problem, if your asking for complicated task, using a good prompt change everything
I literally built a prompt upgrader just for that
how does the prompt upgrader work
i say what i want -> it asks me questions -> outputs clear propmpt
I see that's interesting
Wanna see how it asks the questions?
Im very proud of it
Yes how does it ask the questions
is this available on a website
yes obviously
hi
the only ais that have reached 1t tokens is minimax and kimi
this is insane
ill try them out soon
hi
there have been multiple models which reached 1t
doesnt mean theyre good
also closed models are obviously beyond 1t
but for example ring 1t 2.5 is 1t params
WAIT
damn im stupid then
does the "tokens" indicator mean the usage
yeah
or the num of parameters
He is talking about tokens
usage
😭 nvm then
gg
ur not i jus misunderstood
extension for vscode
its like codex, antigravity or whatever
I think the gemini is sooooo used, the most token model used
Everyone likes a little bit more AI uses the gemini app, I'm talking about the mainstream
oh so cursor but vscode
wait vscode is cursor
mf
wait
no
I think gemini today generates more tokens than gpt despite having a little fewer users
But it's very interesting to see people using the minimax 2.5, it's really a great model for the price
honestly i expect minimax 3 to surpass kimi k2.5 in tokens
yes its a fork
Aight i will publish it asap
ok ty for confirmation
Calm down, It will take a while for the 3 to come out
fair enough lel
you shouldnt care about tokens
alr 👍
cuz
this could also mean the model is very inefficient
as in it uses a ton of thinking tokens
The minimax 2.5 is very efficient the thiking of it
The glm who is a thought token spitter
I wanted the models to be like sonnet 4.5 in thinking but opus 4.6 itself worsened the efficiency
one of the best parts about it is it has free unlimited models
search "free"
wait it does?????
i like how gpt 4o is here in its extended form
why are the pro version of chatgpt and the deepthink version of gemini not on the leadeboard ? they should just be available in batlle mode if the problem is usage
because theres no api access yet
only to some research labs like arc
isn't gpt 5.2 pro acessible throuh api ?
ر
thanks magy
real
real
ر
WHAT DOES ر MEAN
its arabic bro
explain please?
They're creating ai for US not to fill their pockets
i think they trying to colapse the economy tbh
especially if you see the trend that AI and SOTA will be free
so all that debt goes where??
i ❤️ bait
Yo what's this
Basically saying that Chinese models just trained their ai's on questions and scored way worse than opus and other mainstream models
i really doubt its wiping the floor this hard
if it is thats insane

though i do think it will be highly better than v3.2 thinking
im gonna say at least a 15-20% increase (overall)
though with how long its been cooking
maybe 25-30%?
wait what
nvm its fake
the "source" is a random twitter post
written by ai
with "hype" language
ah
damn
i knew it because theres no way deepseek would crush coding
deepseek has always been more of an assistant
not a coder
Hola, acabo de llegar. Espero aprender mucho de vosotros.!!
tung tung tung sahur...??
What happened to the server? Why can't I make videos?
HI all. I'm trying to log in to Arena with Google but it doesn't want to click. Platform: desktop browser
.
Hello?.
Plesase answer my question about failure to log in with google
You need login with google, maybe you used all credits of day
I did that and still nothing.
Don't buy bitcoin
It's worthless

Satoshi just needs to touch one of his coins and the house of cards collapses
or the 51% attack, etc
gold is still the best reserve asset
also BlackRock & co already have centralized bitcoin, and privacy does not exist in btc
(ok that's going offtopic, i agree)
It is stuck on "Generating" from the past 3 hours. Refreshing the page won't work. Any solutions/fixes?
try to logout & log back in
@azure turtle @steady elk Hello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been disabled. Video generation now is only available through the website here -> https://arena.ai/?chat-modality=video
..this often works for me
Didn't work...
you could try to use another browser, after having logged out
to have a "clean slate"
sometimes, if the task was very complex, Opus takes quite a long time
have you given him a coding task?
Yeah the opus thinking is fkd
It thinks tooooo looong
Nerf the thinking lol
Or max output tokens idk
Didn't work
Yes, but not so complex.
in C/++?
Python
one method is, to ask him to partition a project into several source-files, to avoid very long answers
also to avoid that error
("Something went wrong..")
But what to do now that it is just 'generating' from the past 3 hours?
i'd start a new thread and slice the project into smaller files/chunks
Thanks

do not look up 4o deprecation reactions on reddit
worst mistake of my life
Has anybody chat actually recovered from the infinite generation bug, or is that just a myth?
for me, logging out often helps, if enough time has passed (for the AI to finish)
but logging out too early might be bad (as the AI might bug out?)
/image to video
Oh, I see. Thanks.
this ain't video arena bro 🥀
LOL the photoshop i cant
😭
PNGs
probably wants to generate a scam
"My brother Missy and I have come to Bangin Barber Shop to get our heads fixed"
💀
what
Buy bitcoin but don't leave it on exchanges if it's useless at all
its AMAZING
highkirkenuinely brotosynthesis, i'm not running that on my windows machine
<@&1349916362595635286>
Higgsfield is redefining AI Filmmaking with Cinema Studio 2.0 — a powerful AI Video Generator built for true Cinematic AI Video creation. From 3D scene navigation and multi-shot storytelling to genre-based motion logic and character performance control, this platform closes the gap between AI clips and real filmmaking.
In this video, I walk t...
@spare escarp @lone monolith @signal cargo @azure turtle Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
Midjourney
Heavy AI censorship degrades the usability of LLM
The hedging is the most disrespectful things to a person by an artificial intelligence machine
It’s an insult too persons own intent. Totally gaslighting
great doubaou seedance 2.0 is ruined
they added a new 'law' that doesnt allow you to make videos with reference images that have people in them
Yeah 🙁 I am having a lot problems with that
it used to be so good
you could generate john wick
walter white and donald trunp
not any more
is gemini 3 deep think the closest public model we have to agi rn
i haven't been in the ai space for a while
that is arguable for some people
but AGI does not mean a model is sentient or conscious
A model improving itself by adding more parameters into its own data?
What a load of bull
It’ll inject itself with slop prompts and data
why i can't write in #video-arena-1
AGI is not possible through LLM.
true
I think you need embodiment
You need to be in the physical world
If we look at all forms of intelligence, they are embodied by a physical structure
if a terrible model attempts to improve itself, it's going to fail
No, not with auto regression token prediction
if researchers started recursive self-improvement right now with the world's current best models they would just become slop
if we want agi we need a better architecture
Yeah embodied systems
LLMs learn from text, not from physical interaction. They don't know what gravity feels like; they just know how the word gravity usually behaves in a sentence.
You’re also not gonna get it with heavy moderation of content I’ll tell you that for sure
is seedance 2.0 in video arena now
Why genini 3 pro result trash now
@uncut basin Hello, as the most recent Discord-related announcement stated (which can be read here -> #announcements message), video-arena channels are no longer available as the Discord Server bot has been disabled. Video generation now is only available through the website here -> https://arena.ai/?chat-modality=video
Portrait: Vlad the Impaler portrait, long dark hair, sharp features, intense piercing eyes, ornate 15th-century Romanian royal armor, red cape, gothic horror lighting, hyper-realistic -
Yo
Ha ha funny 🤣🤣
Already used 10 different Google accounts to generate videos total 30 times 🤣🤣🤣🤣
gawd dang claude💀
@echo aurora Hello , i have an whole clip for report bug that i will slowly given you guys which , so where do i send ?
3.1 pro is amazing oh my goodness
literally can generate copywrited logos if youre vague enough
bro do if yourself arena.ai
How tf do you have access to 3.1?
Did you sign up for prototype?