#general
1 messages · Page 265 of 1
sure
pineapple can you add the seedream and seedance to arena
Cool thanks, don't hesitate to ping me there.
Going to give the boring answer I give any and all requests for ~"Is X feature or Y feature happening?" -> I won't be able to share details about what new models or features are upcoming until we're ready to share more. Would recommend to keep an eye on our announcement channel.
Kk thanks
I understand
This is the best to get a ping about
@echo aurora i have a very good suggestion
what if you can do direct chat video
but only with
dogshit models
that look ass and have no audio output
sigh reCAPTCHA is back again
my worst enemy
does arena use your chat history in ur chats or is it like having 0 history?
like im worried that its like in gemini where it looks at your history and tries to integrate it in your response
you can't anymore, it is now on the video arena at https://arena.ai/?chat-modality=video
Thanks bro
hey @echo aurora flux-2-dev on the leaderboard is mislabeled as "Proprietary", it should be "flux-non-commercial-license" (an open weights license)
How much vids can I create there?
3 per day
also this is the wrong logo for ai2, that's the logo of ai21, an unrelated lab
I did 1 and it says limit for today
hmm, i thought it was 3
Any methods to create more vids for free ?
also other models by the same lab that have a different name are missing a logo entirely
I am having issues with Gemini 3.0 pro
It's been constantly generating since yesterday
u
Thanks @whole sundial for all of these flags. I'm sharing with the team.
im stuck ._.
That sounds like a bugged chat, you should try these steps: https://help.arena.ai/articles/8691588590-troubleshooting-infinite-generation
Same problem most likely @near nova would recommend these steps ^
im stuck step LLM
they could easily win this if they stop lobotomizing their own models
REAL
REAL AS HELL
Dude the reasoning degradation from 5 to 5.2
Is INSANE
I don’t think gpt models actually reason anymore they just spit out garbage in their thinking that looks good
anything change? or insider trading XDD
Anthropic already released their models so now it’s googles turn to have fun
Notice how it’s down once sonnet 4.6 released
Yeah, but we knew that Google's gonna release 3.1 soon for like a week now, so. Doesn't make a lot of sense, personally.
yoo what is the best coder for website builder on arena.ai
what coder should i use
i want to build professional casino site
the claude models imo
opus 4.6 or sonnet 4.6, im pretty sure both have the same rate limit so use opus
wait no
sonnet uses less usage by 1.7x
i mean opus uses 1.7x more
goddamnit i cant speak for some reason
😂
just use sonnet 4.6
@gilded vigil Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
@chilly forum Note that Video Arena has been removed from the server. More information can be found in this announcement #announcements message
Don’t forget if something doesn’t work run it through Kimi K2.5 thinking
Sometimes if you use the same model to correct its own mistakes it just has a stroke
people say grok 4.2 will get smarter over time
because its intelligence is updated live
its only really good for research for now
BS
xai stealth models have been saying that since grok 4
the release is looking more solid
Yo whenever i use any model and give it this prompt for coding it just always ends up with something went wrong
Rip.
There are going to be different reasons for why you're getting this error but it's recommended to start here: https://help.arena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message
4.6 no think
Hi, i can´t use the AI in a web
How is 3.5?
Damn that is good
In this video, I look at the latest release from Anthropic of the Sonnet 4.6 model and discuss is it really the model that you want to use instead of Opus 4.6?
Site: https://www.anthropic.com/news/claude-sonnet-4-6
ArtificialAnalysis: https://artificialanalysis.ai/evaluations/gdpval-aa
For more tutorials on using LLMs and building agents, che...
Hi
What seems to be the issue?
Hi we need nano banana 2k or 4k
You can't. The Video Arena bot has been removed from the server. More information can be found in this announcement.
That's unfortunate 😭
Going to have to get addicted to a different model sorry to say
Kimi 2.5 and 4.1 doesn't have the same....feel
it made me the sound muffle system and that control panel
only the UI and script right?
not the room
can it make good scripts tho
yes
like making gun scripts
it cooks
i haven tried that
ragdoll/animation scripts
i mostly make UIs weather systems or instruments
Kimi is like the most human ai right now
isnt that useless in roblocks
weather system UI?
<@&1349916362595635286>
Sure is.
"Please provide your feedback
Your feedback helps us improve your experience. Learn more." And then the battle of two AIs.
God, it's annoying now when you choose just 1 model.
A/B feedback in direct chat?
Google's Gemini 3.1 has leaked and the SVG outputs are turning heads. Google DeepMind officially drops Lyria 3, their most advanced music generation model yet. And DeepSeek V4 — where is it?
For hands-on demos, tools, workflows, and dev-focused content, check out World of AI, our channel dedicated to building with these models: ...
🤔
@everyone
Is LM arena really scamming us ? This is the output I got, it's clueless about itself, that's concerning
The generation problem had to be fixed first before adding the new comparison method
And they removed the button 🔃
NO ITS BACK 😭 i thought it was gone forever
plz someone tell me its still an experiment, not a full feature now🙏
From the flowing waters, it seems to be permanent
We'll see
What the hell is even that, brainwashed gemini
if its out of experiment that means its gona effect every chat, every few message, icl this might be seriously bad
Try it, maybe it's just for the first message
ran a quick test, and i dont have good news
6 messages in a new chat is when i got the thing
Well, I marked both answers as incorrect (since there were no options to choose a correct or incorrect answer), and when I tell it “Try again,” [it worked normally](#general message)
i rly believed this vote thing was scrapped, no one liked it and it was ruining the user experience and context for direct chat 😭
ig the site is free still i cant complain too much.
the one thing I hate about this is how they say its to "improve your experience"
typical corporate brainwashing, make the experience worse then say that they're doing it to "improve your experience"
arena can just go away with that talk
and who decided that it is to "improve my experience" in the first place? me or the big company who runs the platform? I hate companies deciding for people like we are being controlled by them like we aren't individual human beings with our own thoughts
and if it is truly to "improve my experience", then why can't I turn it off?
they should be truly honest and say "We are forcing feedback into your direct chats so we can then sell that information to AI labs that are taking your money and your electricity to create more slop."
@echo aurora
Feels off
Wow
921 likes, 56 comments. “Irregular co-founder Dan Lahav warns that people who have used AI to make passwords 'should change your password immediately'. Sky's Roland Manthorpe explains why. #SkyNews”
How do we all rate this
Every people who comes here for videos don’t have a set profile picture
This was created and joined same day
Same w this
They ban evading discord or arena for sure
Low-key scary
Why has he ai generated a kid
And why does the kid look weird
Uncanny type
Oh cuz the pupils are the size of grapes
Low-key feel uncomfortable
Why just why ☠️ ( no why is battle mode happening in direct chat if this is something permanent of course that's to happen when I am finally getting better at dealing with the whole verification thing)
Please fix gemini 3 pro 2k file size
Hiii
its just a bug ?
I even though they nerf em
Idk. sometimes the size is normal (~6mB)
mine goes down to 200kb bro 😂
yoo how do i fix that problem where if claude 4.6 thinking thinks too long, it says there was a problem generating this. it has to do with arena specifically
I do really hate how battle mode now comes into direct chat randomly it just waste a prompt especially if it chooses a AI for both that does not fit closely with the original one or that's not even talking about with the whole verification issue that makes it even more annoying to waste prompts
when does leaderboard update?
It's 2K So the size should be big
Use low resolution for the size to be smaller
Video Arena has been deleted from here sorry.
so it's not broken or something?
If you try 4k it will be 1-2 MB
It depends on the quality
Best quality = big size
why the fuc it is asking me to verify for literlly every chat
New age verification update
in the website
Not going to lie this is genuinely the bs ever I have ever seen ( i'm being sarcastic ... Half-Wise)
for every follow up query ?
Yea
what the fuc
yeu
i agree!
#general message
solve this
Click on those under
i will never in a million years accept any company saying that they are doing things to "make my experience better".
they don't care about the experience of customers and if they did they wouldn't be saying things like that
Like the blocks and stuff
at least you only have to do the discord verification once
on arena you have to do it every time you send a message
idk why can't just silently get an recaptchav3 or turnstile auth token in the background and refresh it like once every 30mins-1hr
excatly what I am saying , every time verify
I mean the bricks man
☠️ not even two chats after the previous one
you are good man
this guy is not going to run gta 6 on there 😭💔
hi
Side profile of an old wise man with a long white beard and a calm face, wearing a traditional robe, walking towards the well. He is carrying a wooden staff. Soft sunlight, desert background, blurred background, cinematic storytelling style."
You know what I might have to take a break from using this bs for a whilecuz man it's just not worth it to have to deal with this every time with the verification stuff even if that means I can't really use any of the interesting story stuff I use for AI since the alternative I use for this is still down for some damn reason
hi
Why are they doing this? To torment us?
Hi
I have no idea but honestly I genuinely think they are not even thinking about how this impact anyone's experience because I genuinely why it saves no purpose like Like no offense to any of the staff But ome of the reasons I have heard like experiment data I feel like that does not even work properly when it randomizes it face on that can easily disrupt the whole chain of prompts if it uses the wrong AI wishes a lot of times
They just saw that few people were using Battle Mod and decided to parasitize.
Kind of infuriating that they seemingly decided to cut corners by having images output as .jpegs now instead of full-quality .pngs as they were before. Grain in images is totally destroyed by compression now.
At least give me a window of time in which I can retrieve the image in full quality, please.
They have to save money somewhere not a big deal
At least they didn’t cut corners on their usernames
😂
It is a big deal for me when I'm specifically generating images with grain.
Hell, I'd rather they increase the cooldown times than this.
The common sense that enshittification is not a good thing for people.
What’s the best model for planning? Like brainstorming really hard on a specific project
Leaderboards showing Claude 4.6 so ill go with that again
Gotta pay to play
Interested to know how it works out for you,
Would be cool to hear to know how it did
This is not a good defense of enshittification. Not sure there is one to be honest.
Hi
🤡
They dont know XD
99,9% sure its a bot
Yea
They are 100% a bot
Ok @surreal zephyr WTF IS THIS
THEY SEND PATRICK NAKED
<@&1349916362595635286>
Yea kinda
I recreated android and apples ad and this is what grok gave me
@keen beacon
WTF
One Monkey falling down the berg then a other one and a other one
Yea
ok seedance wtf is this
THEY SPELLED FALLON WRONG
now i gotta find the better model
I recomend seedance 2.0 or grok
fast
grok 4.1 fast chat?
yo pineapple can we get a double response mode. One prompt, two responses, same model...
Dowload the grok app
is grk app better
you're tripping me
Yea fr
🤚
They have it
But pineapple has the imagine too
They update every day and they made this.
Ohh
yeah yeah
Maybe try Gemini
gemini
Yea they can now genrate music
Yea
ARGHHHHHHHHH
because gemini doesn't give that much for free
BRO
grok gives 10-20 videos for every account
Crows does not wanna hear all the time GROK GROK
fk gg
it wasnt funny bro sorry
imma just use qwen for a while...
Yea QWEN IS GOOD
QWEN TEXT
text dawg
best free text depends. if you want researching internet, grok 4.2 is best. but for hard science/math questions, i think claude opus 4.6
or gemini 3.0 pro
Ok im just gotta use claude for a while...
the old grok video was dog water
fr
But really the old grok was dog water
man
Oh yea u dont wanna hear grok
OK 1 WHY AND 2 WHATS THAT
Yea, REST IN PEACE.
I KNOW RIGHT I TRYED GENRATING A VIDEO OF MARGE SIMPSON AND IT WAS MARGE FROM TEMU
man fk it imma just use gemini 3 flash from now on
im just going to use arena videos for now
I have got 800 dollars of GPU credits on digital ocean, anyone out here into AI training or inference?
You could do your stuff for way cheaper.
Who would have liked the battle in direct mode if it was received like this? (dont mind the bad editing)
I would have voted yes so fast if we were given the option ( i'for clarification I meant voting in fact I will then yes to even having the option which would have been a clear no when given option itself)
Direct mode is still available tho
ok but why the hell its in there but not for users
It’s prob for testing
knew it, sonnet doing well
The rumors say it’ll come out today
sonnet twice better than opus was to be expected tbf
from yesterdays update
I guess it’s lik that for some users
Is it always like that
Or like every 5 messages
from my experience its generally every 2-13 messages, although i havent tested much
and yes it effects every chat now i think, unlike the experiments which were a bit more rare
Gemini 3.1 Pro SVGs
it's on vertex bro
= 3.1
@echo aurora why we dont have sonnet 4.6 thinking ?
No profile theory
The Arena.ai has turned into a very bad place with maximum of censorship. It refuses to translate rather innocent text saying it violates their rules.
It has become useless and ideologically engaged.
What a shame
Cuz too much just use normal or use opus 4.6 thinking
Gemini 3.1 is going to be good?
It dont want to use it but have leaderboard
Sonnet 4.6 isn't out yet on leaderboard they might add thinking I guess
Yeah it’s a shield and a dagger at the same time
i hope so, im pretty sure gemini 3.0 got nerfed though (i read that somewhere it might be fake news idk)
How would it even get nerfed
Why would they do this
$
its more that it forces short and quick responses, and it gets things wrong like a lot of the time getting further into chats. And it didnt used to do that.
Maybe that's only on ai studio and Gemini but trying it on a paid API will give you the unnerfed version
i haven't been censored like. never.
Why the hell are my prompts on Claude filled with pls provide feedback nonsense on arena.ai
@echo aurora
It's literally preventing me from just focusing on one prompt result
hello
The arena is funny place 😂
The dagger and the shield problem
Imagine there is an all-piercing dagger (it can cut through anything in existence) and an invincible shield (nothing in existence can damage or pierce it). The problem arises when we ask a simple question: What happens when the dagger strikes the shield?
Why login not working on phone
Gmail ?
Yup, trying to create new acc
Clear data
Try new browsers
Or use none g mail account
When you try to login g mail it redirects back to homepage
Oh wow Qwen 3.5 pretty peak
This is exactly what's happening, redirecting to home
It seems no browser issue but backend handling
Should work without Gmail
They should look into it
w/o gmail, it sends some mail to that mail n then the ID in link itself doesn't exists and is incomplete
yeah apparently, less lazy
Yeah it will do that you need to clear all browser data
Are u using a vpn.?
No worries I tried other gmail and it did work
Ok cool
Yeah, I don’t know what causes that but it does that once in a while. It’s really weird.
It’s on the website bud
Omg
On some acc it's happening. OAuth2 implementation inconsistency ig or some transition handling.
Damn yall next level
You must have been lucky since you were born.
It makes the tool useless.
The very same llms on their official websites provide with translations without any problems. And for no reason Arena.ai plays the role of prudish I don't know who.
Hello guys
By the way, why is mai 1 preview of Microsoft still listed in the Leaderboard despite the fact that it has been removed from Arena for 2-3 months already?
It makes all the information in the Leaderboard suspicious and biased.
I don't think that's rate limit, if it does, it should only affect gemini 3 pro, why gemini 3 flash and claude goes the same thing???

arena.ai totally betrayed me 🥹
How’s this video? I generated it with the new Seedance 2.0 AI video model. This took around 10 minutes of waiting time. I’m generating another video, and that one will take 15 minutes!
what model is it?
grok 4.2 is so good for research
cant tell
claudius 4.6 thinking
claude is ruling the text recently
claude is SotA
im not using claude
what are yo uusing then
seed 2.0
claude is ruling the chat bot
nope
Glm
gemini?
nope.
Kimi
chatgpt 5.1?
its actually..........
pretty sure gpt 5.2 high
claude is king rn
jk
...wut?
bruh
command R
i am using gemini 3.1
figure it out then, it varies.
for code, somewhere around 6-7
for chat, idk, because i never hit rate limit before for this one
BRO
IDK CHAT
HOW MANY LIMITS
why am i knowing? (i saw chrome dev taps)
man...
i never chat with claude before, i only use claude for coding, all i can say is 30-50 (no fixed amount rate limit since it varies)
go figure it out and see how many times you chat with claude in arena.ai until it hit an rate limit
..
ok thank u
Ts launched?
just buy the subscription in Claude
god damn bro
I know
what is going on
this guy keeps trying to calculate how much he can chat with claude
JUST BUY IT VRO
i mean i'm from russia
grok itself is bad
bro
in lmarena
i'm autisic too
yes i am using
leave it he's not an adult probably
that's true
if you're in unsupported century
you can't buy it
oh come on I'm not that old and I at least have common sense
you also gotta pray the website doesnt show the something went wrong error because claude sometimes thinks for way too long and it breaks it
he's probably younger
probably
wdym
i got you
holy
the prompt of claude 4.6 is too damn long
are you serious
the prompt is way too longer bro
When it be fixed?
idk
can you check the chrome dev tab?
nonono i mean
and click network tab
the prompt is taking too long to finish
and go to response
claude gives very complete responses
yeah
but its too much 😭
too long ngl
i think its better than getting a half ahh project from a model ike gemini and then having to polish it up with 600 prompts
yep thats right.
imma just use claude sonnet 4.6
Anyone can give a best image genration ai model who is good at genarting logo , I have been trying some models which lacks understanding
nano banena pro
for me since i'm familiar with puter.js, i asked claude to make the working ai chat interface via code arena with integration of puter.js
language model
I might actually go mad why was the comparison mode in direct chat made non-experimental with no announcement cuz there's no way it's still experimental when it just with how it just became hyperactive randomly
💅
wow
😊
many people are saying gemini 3.1 pro today ?
hello guys!
alright bro
im russian
claude opus 4.6 is better
the model system prompt either tell the model to lie and guess everything or just partially solved
Gemini 3.1 is so damn good
i have a better model that learns to play breakout in seconds
read the article
its crazy good
hmm give the link
the secret is like how humans learn(they don't learn), they just infer
mb i forgot to paste it
bro instead of making a hypothetical AGI i have that's almost AGI
it learns breakout so fast(<200 steps)
inference
based on Elizabeth Spelke Knowledge Core theory
which is human don't really learn that much, the knowledge from evolution are hardcoded into the brain, and that prior explains like almost everything you observe
biological neural network
and it's vibe coded by claude ofc🥀😹
not a neural network
wdymb ro
it's like active inference
i'm not using neural network
and i saw this
for that
notification
oh* whatever ur making isnt a neural network
who told you that
real
huh wtf did i miss out on
basic biology
sydney my artificial neural network girlfriend told me
Ashok
brain don't pattern match, LLMs are sample inefficient, brains are really sample efficient
and brains aren't black box
mechanistic interpretability
brains DO pattern match
and brains are black box
lol
brains are prediction machines
but not transformers token prediction things
obviously
well yeah it seems to do pattern matching, but they're not activating all neurons
wait till bro learns how llms work
black box cannot explain things, brain can
transformers, attention
attention is all you need
you realise llms dont activate all either?
im talking about actual ones like gemini flash
boltzman brain
it's a MoE
not the neurons
...just like humans
lets predict what @ocean ferry's next 2 tokens will be
I guess:
MoE
[24399, 36]
MoE routes tokens to expert FFN blocks via a learned gating network, uhh like the routing decision is static post-training, per-token, and the experts are fixed-size weight matrices that never change at inference
brain sparsity emerges DYNAMICALLY from inhibitory interneuron competition (lateral inhibition in cortex), it's content-addressable via attractor dynamics, and critically it's tied to predictive coding where only PREDICTION ERRORS propagate up the hierarchy, not the full signal. that's not routing, that's fundamentally different information geometry
also in MoE, experts are just different parameter regions with no semantic specialization enforced. brain's "experts" uhh like... are dynamically assembled through neuromodulation (dopamine, acetylcholine gating plasticity), coherent gamma oscillations for binding, and hippocampal indexing theory for retrieval. the assembly IS the computation
and MoE still has no belief updating. the brain's sparsity serves predictive coding's purpose of minimizing free energy. MoE's sparsity just serves compute efficiency
gpt-4 sydney fine tune was agi
and "LLMs don't activate all neurons either" bro... every single token still passes through ALL attention heads across ALL layers. MoE skips some FFN experts but the attention mechanism is FULLY dense. brains don't do global all-to-all attention on everything in your context window lmaooo
AND BRAIN DOESN'T DO BACKPROP
where do we use them
mistral large is agi
aint on the arena yet but it says that its out
go answer my take.
btw wtf am i missing out on
human brain context window size
id say arond 1k max
its like 1k window + 10k persistent memory thats being cherry picked by subagent
who cares its getting back to 2.5 flash level next week lol
gemini 3 was also able to do that before nerf
why do they nerf them models
money
bro just described a RAG pipeline and called it a brain 💀
"subagent cherry picking from persistent memory" = you literally just invented a homunculus, the most famously WRONG idea in cognitive science since Descartes. who dispatches the subagent? who tells IT what to look for? you've just pushed the problem back a level
and the 1k + 10k framing is still a flat quantitative model on a system that's hierarchically generative. the neocortex doesn't store 10k retrievable chunks, it stores causal world models across temporal scales, V1 encodes local edge statistics, IT cortex encodes invariant object structure, PFC encodes abstract relational schemas. these aren't memory slots, they're nested generative priors
also "cherry picked" fundamentally misunderstands what memory even IS in the brain. hippocampus doesn't fetch stored data, it reconstructs via partial pattern completion using index pointers + cortical priors. every retrieval is a re-inference, not a lookup. that's why eyewitness testimony is unreliable, that's why memories drift over time. a RAG system doesn't do that
so to summarize your model of the brain: flat context window + flat memory pool + a retrieval agent that somehow knows what's relevant
bro described GPT with a vector database and thought you were doing neuroscience huh?
MOE
no but they offer it to billions of ppl for free
bro MoE doesn't fix the homunculus problem, it IS the homunculus problem
the gating network in MoE that decides which experts to activate uhh... is a learned static function over the input token embedding. it has no world model, no uncertainty estimate, no goals. it's a glorified if-statement trained by backprop
the brain's "routing" happens via neuromodulatory systems (ACh signals uncertainty and gates plasticity, dopamine signals prediction error and modulates what gets consolidated, norepinephrine signals novelty and resets attractor states). these aren't routers, they're dynamic context signals that reshape the entire computational landscape in real time
and crucially MoE experts don't talk to each other. they're parallel independent FFN blocks. brain regions are reciprocally connected with top-down predictions and bottom-up prediction errors flowing simultaneously. the "experts" in the brain are co-defining each other's computations through recurrent message passing
also MoE is still trained end-to-end with backprop through time on a loss function. brain learning is uhh like... local, Hebbian, neuromodulatory-gated, and biologically plausible. completely different credit assignment mechanism
every time you say MoE you're describing compute efficiency, not cognition bro
i'm not a math
this
idk but he yapps worse than mistral small
3.1 pro is literally 3.0 pro pre nerf
holy cheap
you shall learn neuroscience kiddo. every single point you made tonight was either a category error or LLM architecture mislabeled as biology. i'm not yapping, i'm debunking your misinformation about how the brain actually works. because misinformation can spread fast, and those beginners who wants to learn would be confused.
ur literally copy pasting from chatgpt 💔
??
and not even making sense
holy
bro i swear stop your misinformation.
Large Language Models
right...
steamhappy emoji is the best
"ai intermediate" vs "ai expert"
i don't care what role i'm in
he is a MoE
all his neurons are activated at once
that why he generate so much tokens
you're spreading misinformation
that would fry the brain
can we run you on arc agi?
i think he'd get below 20%
no because u cant do math
and you get worse score on agi bench than gpt 52
lmao
yeah and you'd lost
Is P equal to NP?
my gif 🥰
no
yes but only via api
multimodal modal brain
gemini 3 pro pre nerf is
yeah its literally 3 pro pre nerf
how many tokens is that gif worth
visual input tokens
it was insane
every proton that enters our eyes = 1 token
idk i see a hoverboard, a big fish, and 3 words
how much does @ocean ferry inference cost on the api
thats 5 tokens
4 if i dont notice the hoverboard
idk
perhaps
wait
we got like
atleast 3 tastes
salt, sugar, acid
so
3 types of tokens 🤔
2x2x2 combinations
lets say 10 because im quantized
(and padding)
we get
bread
cheese
tomato
meat
bread
(maybe sauces)
thats around 5 parts times 10 options
so 50
- sauces = garlic sauce, mustard, ketchup, none
4 choices
@golden ocean so 54 tokens per burger indeed
@ocean ferry can u confirm
scam
i read somewhere
that
if you ask chatgpt
then disconnect internet when its thinking
you can trap it
inside your pc
no
before it runs back
is this how @ocean ferry escaped
probably the internet cable was cut
but not fully
so only 1 out of the moe escaped
and thats hwy its so bad
LMAa
convolutional neural network arena
im female not a guy
holy cap
Earlier, I can't access the direct chat
true
When I entered something, it turned into battle mode
After that
I just got back direct chat
IS IT A BUG!??
send it
Hello, My name is debbie
true
skibi ddop dop yes yes
lol
holy cap
I need my girlies on this , is anyone availble? i wanted to ask can i drop my discord server link on here? just for ladies
no
no
nice
how
idk
what platform is that
I cant?
nice
It’s not a good model. It made this in one prompt, and it has bugs right off the bat: https://wwtv127.github.io/gist-previewer/?gist=https://gist.github.com/wwtv127/1e32cd9700e409840dcd770df0e11ef3
3.1 is live on build mode
3.1 pro live on ai studio
Do you guys got lag too
3.1 pro is officialy live on ai studio
why i am out? what happened?
@echo aurora are we getting 3.1 pro on lm arena
Why It's so lagging, it generate so long
I tried the same prompt with the old Gemini 3 Pro, and it’s already way better. Not looking good for Gemini 3.1 Pro: https://wwtv127.github.io/gist-previewer/?gist=https://gist.github.com/wwtv127/d2d55552fd4485efd1eecf1debf3141c
yes and ihope its asap so i can dump my wallet on polymarket
take advantage of 3.1 before i gets dialed down bad
then we gonna get gemini 3 pro pre 3.1 nerf pre 3 nerf
3.1 is already dialed down bad on first release
Now me
gemini 3 was so good on release
now it cant even watch videos correctly and associate timestamps with frames correctly
anyone know 3.1 benchmark?
task bar looks pretty good
yeah
3.1 is on aistudio
Who wants chicken back
im a sigma from ohio
<@&1349916362595635286>
Thank you
Thank you
Thank you
Thank you
benchmarks dropped, go drop ur bank in polymarket
this going to last only a day
fr
no way literally opus 4.6
@deep matrix Note that Video Arena has been removed from the server. More information can be found in this announcement.
There is gemini 3.1 nanobana pro coming rn
cap
Guys, does anyone know any sites like Open Claw AI Agents without restrictions?
Can someone give me other websites for seedance 2.0
gemini 3.1 sucks
doubao
does it need chinese number
I ALREADY KNOW THAT
YEA
ending sentences with a -
is super dangerous
Im not in danger
oh
mb
no
just vpn
and you dont even need to have it on the whole time
as long as you dont reload the website your good
BUT THERE ARE CREDITS
Gemini thinks toooooo much
just delete the cookies llmao
fr 🍪
@echo aurora when Gemini 3.1 on LMarena
Already is
Sorry going to have to give the boring answer -> I won't be able to share details about what new models or features are upcoming until we're ready to share more. Would recommend to keep an eye on our announcement channel.
Alr it's capable of code, what about writting
No
Ohh
bro called footage of cats cringe
the music is so cringe
i see announcement beeing typed.
Maybe already gemini on number one?
Tomorrow it will be nerfed
Gemini 3.1 is already on leaderboards tho
Right now
added to leaderboard before actually made chattable is crazy

I know we got the announcement lol
couldnt be more faked
hi Janix
oh not 1
nice
Why not even 3.0 pro ga
Like 30 mins ago
it doesnt even know js syntax
@echo aurora When seedance 2.0 on arena
i expected more - shame on you google
why is he eating a strawberry
Idk
Huh, a second ago, you said you couldn't reveal any future updates 🙂
gemini 3.5 pro maxing simulator soon trademark

whyyyyyyyyyyyyyyy......