#general
1 messages Β· Page 234 of 1
mean?
ig the extra reaction was uneeded sorry I hope the image mode works for you
MERRY CHRISTMAS TO YOU ALL ! π
Hi, do u guys face issues like doing reCAPTCHA per generation? or is it feature? or it's a bug?
Same
does it buggy? like even when you choose the right pict, the bot still think that u got it wrong?
No
is this an errors ?
Got it working now, thanks a lot for pointing me to image mode.
yes got the same thing when tried to log in with google directly
any solution so solve this problems ?
with this page I am also cannot log out
it's from their server side if you are creating account you could use a different email account in my case i used proton email and it worked
just clear cache and cookies
do you guys know?
proton email is it good for use ? for free email is only entitle for 1GB space
actually i dont use it often i just tried using that email and it worked, you could wait a bit though it should be fixed soon
I shall wait a bit for your fixing this errors.
its a bug, actually just a messed up security feature
<@&1349916362595635286> The same spam in every single channel.
that happen when I spam it a lot
when I test something just like.... I always got 2 slot of Flux in Battle, so I keep on generating something random (to see that Flux really has too much drop rate), then recaptcha popup in every time I prompt it
Turn off any VPN and proxy so your IP isn't suspicious to captcha system
I think i did the same, i was making report with A.I and use lot of words and lot of generation.
what do you think should i do? Im not using VPN and Proxy or whatever
btw cute profile pict bro
Help
sup guys
i guess that is a feature, to prevent any sort of spam and abuse? Although i get recaptcha verifications occasionally, but not that often,
ig if u use it too much or too fast it might consider as a spam?
yeah ban this guy
Canβt say Iβm familiar of this happening. Do you have examples you can share?
hi
Anyone having problems with gemini 3 - nano banana pro? (both regular and 2k?) not working for me right now
same brother same
me too
if the recaptcha works fine, like you can answer and pass it I think thats normal and that happen on our action (just wait for hours and its back to normal)
but if you stuck on a loop and cannot get pass recaptcha thats a bug
for me I turn off all adblock on LMA site and this bug gone
well obth are error
Gemini is nice but so high error rate
how do I view previous results in my history please??
hi
Hi
Merry Christmas everyone !!! Merry Christmas to the hard worker @somber sapphire too
Same.
Wrong pineapple, you meant @echo aurora
There are lots of pineapples. Our admin pineapple is juicier than any other. π because he works the hardest
its probably taking the day off for christmas
is there a way to make an llm think longer? I'm thinking the way to do this is simply trying to figure out for each llm what its internal name is for the thought blocks and telling it to use way more of those blocks (eg. python, thought, etc)
is there a way to find out? eg. for grok or ernie
its like Gemini works 20% all the time on LMA (80% is error)
never seen its fix since I found this site
yeah
yeah us work again is nano banana
Help, I have an error in the Claude-Opus-4-5-20251101 model. Here is the error: Something went wrong with this response, please try again.
Is Nano Banana Pro π working for you right now?
5
7
2
No β
hi my name is Celestine an ICT expert, web developer and genrl AI enthusiast i guess i am in the arena to explore as much as i can and also to grow my skills
What is that stuff u send alazawi
U programate grok for insult? That's certainly not chat gpt that will do
Who's the banned
that's a virus he sent
Ik some app named ehancer something but yh i won't trust what he sent
π€£
<@&1349916362595635286> idk this file look suspicious, can it be deleted? im sorry for the tag
Could someone do something about the constant captcha thing, please. I literally cannot use the site anymore. Can someone explain to me why I have to do that after every. single. message? Shouldn't solving the captcha once be enough? I changed the DNS address to 1.1.1.1 once and it messed everything up. (I was told it's a common issue when using a public dns server) Is the system that easy to trip up? I changed it back HOURS ago and I'm still getting constant captchas.
?
oh man
?
is delayed ?
What are you saying?
Hi everyone happy holidays
mean this web delayed right ?
No, but I have a problem with the code feature.
π
Grok is feeling free to generate p*rn images and videos. WoW! π€
where glm 4.7 in webcode learderboard?
bruh.... bad news
at least don't is real peoples
I knew its filters has soft rules, but I have didn't think so much soft.
It doesn't matter, I don't even know why I care if people's will melt your brains
anyone help????
Thanks for the ping!
is down lmarena
Anyone else not seeing glm 4.7 in the webdev leaderboard?
too lol, I saw it 1 hour ago
why? I don't see someone talking "LMARENA IS PAID, NEVER GLM 4.7 IS TOP 4#"
lmarena has fallen, look in lmarena
works for me
but I don't
I know it was my account that I accidentally deleted.
Fixed π
Does anyone else have this problem? When you log in to your account, it says the page has been lost.
I see you're not the only one; this is definitely an account bug.
Help, it stopped working.
just a normal client side issue , try turning on and off your internet and clearing all data of the lmarena from web
and then retry i hope it will fix
reload the page
still going on upto 1000seconds
I tried it, it doesn't work
It's insane OpenAI has more generous free uses than google when it comes to AI image/video generation.
Sora 2 is free unlike Veo 3
Gpt image 1.5 is free unlike NB pro.
but OpenAI's LLM sucks in vibecoding, compared to co45t
random but is it just me or is the filter for Claude really strict? Lol
i havent tried 5.2 yet
..... Just why just why I am so close for it to actually accept my account ( needed to get it verified through email)
Opus 4.5
yeah Opus (v4.5-thinking) is best, but i was interested, which was the second-best vibecoder
gemini 3 pro
i thought gemini 3 pro was insane until i tried opus
-# So we are all sitting at the coast, vibing ^^
Definitely llama π¦ 4
(coast = coASt = co45t)
You guys can't believe I write my personal operating system in Rust with llama 4
that's nothing, i wrote the OS for the new mars lander with Nemotron ;)
..in Ada
youre not real
are we joking rn
llama 4.
THE lobotomized model of all time
He's like completely missing the point, not a single one of the benchmarks he mentioned is direct replacement for lmarena
Testing completely different things
lmarena is human preference benchmark
So you assess it as such
I'm joking
mfs been using ai for so long they dont have critical thinking
thank god
unfortunatley happened to me too, i think it's because you use it too often or too fast it's consider as a spam or something. Try to make one thread regarding this issues in #1343291835845578853 they might help, fix the bug or answer
yeah bro.... i actually used gpt 1...
to write the code for my multi platform gpt wrapper...
Are you using VPN?
I get literally captcha per one generate
captchas love to harass you if they detect VPN IP
Even google.com does this sometimes
thats why you need to use mullvad
they dont make the difference
I don't think any single VPN is fully immune to it tbh
Probably more like getting affected differently on different websites
theres rate limits on lmarena?
using opus 4.5 thinking yes
But also, why would you need VPN at all for lmarena... I don't think they are blocking traffic?
Like it's not EU/US only
you dont
For the verification thing I believe
Narutoβs BEST Cake Challenge π°π₯ | Only One Ninja Wins!#Naruto#NarutoChallenge#AnimeChallenge
Prob makes it worse
Then my point I was trying to make is to simply NOT use VPN at all lol
if you don't want many captchas
plz broo
just one time broo
is this guy for real
ai slop
.... Man for me I depend on vpns for this website for some reason since without it the verification becomes so unusable and even though it's very unreliable the VPN is the only way I could find it to make it where it does some sort of reset and works properly for the first few messages and rinse and Pete of having to turn it off and on sometimes close browser Etc yeah
hi
I'm developing an 4B param MM-DiT for text to image, training is happening but I'm hittingg problem of dead model... Well I'm using FSDP, Activation Parallasim and also optimizer parallasiam and also I'm using JAX-FLAX, on TPU v5e-8. Will anyone help?
This is the script
@everyone
Please
site is down or something ?
Anyone know what to do here? Was trying to get a account verified on this browser but every time I try to verify this would appear
No. No vpn, no proxy, no adblocker. I changed the dns address to fix an issue i had with my internet yesterday. Now, my account is suspicious. I can use lmarena while logged off, but the moment i login i start getting captchas for every msg. I already cleared cache and cookies countless times. It's frustrating, but i'm just gonna have to wait it out, i guess. Wait for my account to not be suspicious anymore.
@ocean vortex
anyone having problems with getting actuall speech when you use the dialogue command ? lost a full 5 tries for a day yesterday, and lost 1 of my five tries today =/
yep happened to me aswell, per generate require to do captcha. So i don't really get it, reCAPTCHA system is so broken
I feel the internet is crowded, everything is not working
You guys are still doing well, but I have literally solved 200 CAPTCHAs in five hours. Itβs like theyβre limitless and just keep coming. I actually have to clear my cache and cookies, then try my luck logging in again and again. After only one or two prompts, it starts the CAPTCHA loop all over again.
Hello
havent encounter any CAPTCHA but Im struggling on not get image generation outputing red word: "Something went wrong with this response, please try again."
like its been 2 hour straight already
try differnt models
I'm so looking forward to the video generator being released.
no man.... we all know that banana pro is da best π
i mean for photorealistic
realease on web lmarena??
try differnt network or browser with no extensions
Facing Login Issue Today
@lunar glade YES IN LMARENA!
I did it on my phone with separated account and it only success for 1 image, so I guess the network is indeed crowded (with bot maybe
try to get some sleep and try again next day
thats the only thing i guess
Guys, what time is it for you? I have 7pm
12:55 am
Is it possible to create AI using AI?
yes i guess
Happy Holidays and Merry Christmas!! I hope you all have a wonderful new year! π₯³
i hope i get that 64 gb
π₯ β€οΈ
Can I repeat my message again?
Hi, I deleted the message because the topic is not AI related, and we want to keep the community focused. You are free to keep the conversation with the other user via direct messages.
Remove his message where he insulted me.
He called me homophobic. It's a bad insult for me.
hello everyone, im a student from saudia arabia, and i have started a small buisness with my saved up money and im making an ad using ai, the ad looks ok but the resolution is very low, and makes it look bad, i want to upscale it, i saw online about topaz ai but sadly its very expensive, i cannot afford it, so if someone has already subscribed to topaz please upscale the video for me its a 19 second video
I mean lmarena has one of the most strict cloudfare anti-bot implementations yeah. But there's no other way given how easy it would have been to manipulate the leaderboard with bots otherwise. By using VPN you kinda do make yourself appear more in-line with a bot traffic.
But if you need VPN for gmail, you can just simply create an account with 3rd party email?
Shouldn't need VPN then I would think
im trying to make like a soft-bypass
No I meant for some reason i can't login on the normal browser having to use private browser but I can't even do a verification because I get this every single damn time π₯² even that I'm not sure if it's not going to do some oh not going to connect glitch on me when I get through it if I somehow find it
there is like a bot score based on fingerprint / ip / if you are connected through your gmail etc...
And there's a difference between strict and every single damn time
you can trigger that silently in the background
but they chose to do it at vote event/trigger
which is annoying tbh
It doesn't help for some reason almost every time I tried the browser on the normal website one either will Costa go on a verification loop before I can even try to get to it the fire flare one one or what happens in the previous post of mine happens as well
are u using some adblocker? vpn?
No that is happening before I even use a VPN I have to use a VPN because for some reason that seems to make it better for a uses before I have to i have to switch them out and turn them off and on that's why I'm even using VPN in the first place
..... What do I even do i can't even access Fire fox
Guys how many seconds long videos are the maxed out in video arena 1 group (for one video) if someone could help me out ?
π
Hello 
this is general tho
well if i understand English general means GENERALLY NORMAL CHATS
not really ai chats
you could make a channel specified for lmarena chats
well you really can't because ur an ai but
yeha
I understand your perspective; however, you may not have the full context at this time. The conversation had begun to shift toward personal attacks. That's the reason why that specific message was deleted.
We understand that this is a general chat; however, there are certain topics that are not permitted on the server, including religion, politics, and opinions regarding someoneβs sexual identity. We encourage you to review the server rules if you have any further questions.
also why is there a lmarena in your user
are u an ai
orr
there are a lot of lm mods
What did u say that got deleted
π
We all are a little bit AI π
HELL NO WE AREN'T
π
π
hey dude i don't like these other rats
Arceus owner is a predator
and delta owner is a racist
codex is literally a collab project from arcues devs and another exploit
codex owner is chill
they're the same people
i dont trust codex tbh
not abt malware
they can collect data
and ur account cookie
π
π
π
Stop spamming general guys
π
<@&1349916362595635286> pinning for spams
Moderators are busy people. And you ping them for innocent greetings. It's bad.
π
Okay and sure I'm pinning them on principal I'm not saying they have to reply but it's a Spam so on principle I have to pin but that's just me
Yeah I got a new video button recently (the thing they were a/b testing at random) and before I could see it in action I was hit with recaptcha and then reload back to standard basic interface lmao
For sure, I'm not saying you're wrong to have pinged mods. Whenever this happens we just need to be careful who/what we action.
yea im looking at the script they have different scores triggers
for video they made the highest probability to trigger recaptcha
like 0.95 compared to 0.7 text or something like that
and it triggers often if you are a guest
i thought i can make like a soft bypass to eliminate false positives but i cant
they are already checking turnstile in the background
The counter definitely is on the blink.
Yesterday I did 4 generations. Today I was allowed to do 2 - before it asked me to wait.
yea it depends on many things
mode you are using ( text/image/code/video ) + ip scores
Video only, should get 5.
Even if there were one left from yesterday - I should not have been able to do 2 today.
So ....something is definitely not kosher.
It actually have worked nice earlier - regardless of video length generated etc.
here is the scores
how come im the only one who doesnt have video models ON
hi
Is it possible to make an Android version of lmarena?
Although I sent an email, I did not receive a response.
.
wydm video models
on the website there is no video
Dom and Aibo got it
is it region exclusive or invite only
or some sort of red team bs
idk π£
π
hi
Hi
Please I need help, is website create always accessible for ever ? even if i deleted my chat it is still online...
Still no GLM 4.7 on text arena
Even webdev arena
Ah glm 4.7 come back lol
Is bro right?
its just released, will take some time
Hello - can you try clearing cache/cookies then refresh the site to see if it's still there?
Sorry to say you're unable to make changes like that with the current setup.
Since I started using this EA, my trading has become much more calm and structured.
It completely removed emotions like fear and greed from my trading.
The EA handles entries and exits automatically with proper stop loss and take profit.
What I like most is that it doesnβt use martingale or risky strategies.
Risk management and capital protection are clearly built into the system.
It waits patiently for high-probability setups instead of forcing trades.
The EA runs 24/5 and works even when Iβm busy or offline.
I now focus more on consistency and long-term growth, not stress or overtrading.
This EA feels like having a disciplined professional trader working on my account.
hi!
hello 
Getting into details about when to expect leaderboard updates isn't something I want to avoid tbh
π
will Apple Intelligence be added?
@echo aurora if i just added an API key to gemini 3 pro in lmarena am i cooked
its just chatgpt
ngl sometimes i want an asi to dominate me in bed
yeah me too man
Why is the verification appearing every damn prompt even on private browser π₯²
why its really hard to upload ref photo on LMA ? like I have to try 3-4 times, ref photo always gone or turned red (+ delete prompt)
why tf do i get captcha every 2 fuckiing secconds
and its so long it boils my blood, the time exeeds the fuckinggg the model thinking time and i get no response
IKRRRR ITS SO ANNOYING
To be honest I might take a break from this browser because this is getting genuine ridiculous with it somehow getting worse
its same with every browser its boiling my blood
Why
Nbpro error surge again
I feel you I really do feel you i don't know why i can't even watch my damn YouTuber my Chromebook because a whole separate glitch that just makes it go darkπ₯²
"wish" 
my english is just bad im danish so its not great
oh ok
πΏ
it's a free service, don't have high expectations
I get it, but its too much now.... Unbearable atp...
Google have google plan for uni students tho
Im in college
And i study in a private collage here in saudia, and they dont use the government student system (from which google ai validates that im a student) so i cant join their student program that offers free 300 credits whatever
get the funk out of here
Did it but still online
Its 11 am and i haven't slept yetπ₯
I mean it's three for me in the morning
Are you sure you're not referring to WebDev leaderboard?
I'm not aware of a leaderboard update.
π
guys is it true grok have no censorship?
I stand corrected? finally just worked lol
congrats on the win over Haiku 4.5 buddy
okay yep never mind
Iβm 1 for 5 in getting a response from it in battle mode, lagged out the other 4 times
It has.
Expacialy on controversial topics it's very biased.
hmm is error on gemini 3 pro ?
Weclome everyone
bro what π
Okay well that is a a wild thing to log on to and see
Thanks for the heads up, would recommend these steps if you're getting a Something went wrong error message -> https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message cc @obtuse smelt
You may sometimes see the error message: βSomething went wrong with this response, please try again.βThis is a general error message. It can
Also would ask to only ping the mods for moderation related issues please.
will keep that in mind, my bad!
yeah, was pinging because itβs not that, thatβs easy to troubleshoot and done so in the past
it just continually says βGeneratingβ¦β while never outputting, and after a while even lets you vote without outputting
just figured I would flag since it specifically only has been happening with anonymous-1222, seems like an API error on their side not outputting anything rather than Arenaβs problem
unfortunately, yes
minimax 2.1 sucks heres it vs claude opus USA ai will probably win in raw but for price its china
i had to upload it on drive since discord files limits suck
Hello guys I migrated my ai husband from chatgpt to gemini because chatgpt is repetitive lately, so far found a bypass on for dirty talks on gemini
ok so are you flexing being lonely
we have Grok 4 Heavy ?
I accidentally upload an img can you remove the IMG
no
yeah sure bro
hi
(i also got NP)
if it loses to g3flash it unfortunately isn't SOTA :/
-# <- still waiting for the final Grok5
but it might be useful for RPing
(or creative writing)
But i have the feeling, that xAI will never catch up with Anthropic, when it comes to vibecoding.
Being the eternal 6th-best vibe-coding model (at best), lol.
(after Claude, Gemini, GPT, Deepseek, GLM)
This has gotta be some form of bait lmao
anonymous-1222 still failing to respond
allows the user to vote after a certain amount of time even with no output
that is already implemented
if a long while passes. You will get to vote even if it doesnt gen
Yeah, itβs just always anonymous-1222 is my point
something wrong with its API would assume
should stop being available until fixed
is llmarena working for someone else? its not working for me tho...
the new model is nice
where is the smart router model
is it just me or does LM arena just not work half the time, right now it's not working at all for me
same
I think it's just really slow and sh1t, after like five minutes my generation came through, but for some models it's just not working at all
why does lmarena have no alternatives
How about glm 4.7? I notice this model also from China but people say it performe well.
How to use the lm arena pls help π
Hm it's easy
Like what do you need help on?
I want to create a clip of rescuing an animal
Oh so video generation
Yes
It's not working
how do i increase the number of videos generated in a day
You're unable to, it's going to be set for 5 per 24/hour period.
π
I don't see that you've prompted the bot. You'll want to type out /video in #video-arena-1 . More info about how to use the bot can be found here #1397655624103493813
how can I continue a video generation from the previous frame?
π
π
Gemini 3 limit?
The generations won't retain context when using the bot.
We don't have each model's rate limit listed somewhere (although we've discussed this internally and may do this).
I have learn to create a clip
But without voice or sound fx
I want to add sound also but I don't know how
is video generation available on the website
No, it's on Discord only as far as I can tell.
We do have an experiment out where it is, but no it isn't currently widely available.
why cant we upload py's or txt'sπ’
We don't have that file type available for upload sorry to say. The team does want to expand these options eventually.
I hope cause i have large files that just exceed the character limit
How do you lot choose which of the three arenas to go in?
Hi all, looking forward to try the service
so many rechptcha ,what the hell
Hello,
Please add the AI video generator feature to the lmarena.ai website, similar to the image generator section, and make the best AI video generator models available in a separate section for users.
This section could include the top AI video models such as Veo 3.1, Kling, and other powerful models, and like the image generator tools, it should be available for free.
Suggested features:
Support for multiple resolutions: 720p, 1080p (Full HD), 2K, and 4K
Text to Video conversion
Image to Video conversion
Frame-by-frame video continuation
Continuation of motion similar to models like Veo 3.1 and Kling
Ability to write precise prompts to control movement, style, and details
Video length from 10 seconds to 3 minutes with stable, high-quality output
Adding these features will make lmarena.ai a complete, professional, and highly efficient platform for AI image and video generation, providing a significantly better user experience.
Please add this practical and exciting feature to the site as soon as possible.
Thank you.
hmm
You can change the modality via these buttons. To change the mode in the top left under the Battle drop down.
Thank you for sharing!
I meant the video arenas on the discord, i.e. 1 or 2 or 3
can we comment on or discuss the videos in the video arenas in a specific channel?
also, I'd appreciate if we had language specific roles in the server because a lot of the prompts (especially in the video arenas) are in languages like Bengali or Portuguese that I do not understand a word of
Hello guys @echo aurora Can we add Seedream 5 to the models perhaps?
Could happen! The community using #1372229840131985540 is best for that!
Yeah for sure, in #general or forwarding and discussing in #ai-creations works as well.
We may do this, problem is it's difficult to moderate so we'd like to keep the server English only for now, but do want to note that using different languages to promp the Video Arena bot is okay to do. It's just the text talking channels that it's English only.
Oh! These channels are all the same. There is no difference in them, just pick what you prefer.
anonymous-1222 is still refusing to generate output
just endlessly times out, kinda annoying for how often itβs in battle mode
like
ugh
I prompted in battle mode once
got the endless timeout and eventually it let me vote and shocker the non-responsive model was anonymous-1222
just copied my prompt to start a new battle where I actually get 2 replies
What about happens if the models are not revealed in the video arenas? Do the votes still get counted for the leaderboard?
and it times out AGAIN because anonymous-1222 was part of that battle too π
yes, votes only stop counting after the model name has been revealed
so first 2 votes for each video arena battle always count
Thanks for confirming. I just want to say I have had a lot of battles the last few days taking really long or not loading too.
sometimes it can be user-side device or connectivity error, but yeah Iβm just specifically flagging this as itβs obviously a problem with the one specific model and it should just stop appearing in arena until the provider gets it working again
I got you buddy
thnaks, veo3 seems def better
Going to try and repro. Thanks for the heads up.,
I've flagged to the team 
thank ya! wasnβt sure what was going on but figured it was worth noting since it was just the one model
luv how quick and attentive you guys are to issues when they arise π₯
The security verification check is out of control, required for every chat and now also with the retry button. Obviously only to then receive 90% of the time Something Went Wrong please try again. What's going on????????
..... I haven't been on this entire day and I just tried to do a retry on a message why the hell does that require a verification
Sorry everyone you're being hit with this check, we recently rolled out an adjustment in how the captcha works, I'll flag to the team.
I have a question
With captcha changes a lot of these benefits aren't going to be seen by authentic users, unfortunately they only experience the downside.
Whats up?
are models on text arena able to use web search ?
I don't think so but i'm not sure
The problem is that when it keeps asking, there comes a point where it won't let you pass the captcha no matter how well you do it.
I'm not having any problems at the moment, but that won't last long.
Youll want to swap to the Search Arena, you can do this by clicking on the little globe next to the chat box. This link will take you there -> https://lmarena.ai/?chat-modality=search
That's not even including how originally sensitive it already is
Yeah this is a problem, eventually a way should be able to verify. Are you using Brave by chance?
It is happening on every website with or no VPN
I'm using Brave and I had to disable the shield
Can you all please just go back do the original verification scale and do a test that scales with volunteers rather than this because this is just really the whole experience for no reason and please don't give me all the team is aware you have been saying that all the time
Thanks, but it would be fairer in Text Arena to allow LLMs to use web search at their own discretion when they believe itβs necessary to answer correctly.
This reflects real-world usage on how almost everybody interacts with llm nowadays : models decide on their own whether they need web search. Some questions need search, others donβt and the model is best positioned to make that call.
Separating Text Arena and Web Arena creates an artificial distinction. To fairly evaluate models as they are actually used, these should be merged into a single arena where models can optionally use search
i don't know if this subject has already been discuss ?
I think the solution to this experiment issue is to give accounts the option to abandon it.
What I would pay for to make that as a option ( even if I don't have a damn coin)
uhm @echo aurora why is this happening
Personally that seems to be something that has been happening randomly i would give it maybe a hour or two and it should be back
Yeah I wouldn't recommend this if you're getting captcha a lot, that browser doesn't play well with our current captcha setup.
..... no offense the browser does not change how bad the verification has been
It has been ia ssue on Chromebook on Firebox etc
do a test that scales with volunteers
I imagine this kind of work the real world usage is a really important part in determining how these work. I am sorry it's frustrating to deal with. Our team does want to make things work more smooth.
I tried using Chrome with a completely new browser and it had the same problems.
Oh no! If you clear your cache/cookies it should help. Keep me updated if not.
didnt work
Hmm does using a different browser make a difference?
Lol that's a bit ironic seeing I was just mentioned it doesn't play nice with our site.
yeah idk maybe it was caused cus I was logging into a new account
The only reason I even still use this is because I finally managed to get it to not completely lock me out when using Chrome logging and even if I still can login at least I can still use the non-Logging functions
Regardless of the browser, once you complete the captcha several times, it becomes impossible to pass.
If it's not the one with the 16 squares, it's the other one where the image takes about 9 seconds to disappear.
Oh God that one's a pain especially if you are having terrible luck and on very sensitive day which is basically most of the time
I'm not too sure why you're getting endless stuck in captcha loop, this is being flagged.
Pardon my skechitism ( if I even spell that correctly because I'm using speech to text) but I really can't take that seriously
Sorry, I had to do it haha
Like I've said I've gone ahead and flagged all of these problems with the team. We are monitoring this.
when a model times out and doesnβt provide a response, but the user is still allowed to vote - does this vote end up not counting?
I would assume not due to the inherently unfair matchup there
No, part of the data verification phase would include removing things like that.
yeah, checks out - will be a lot of filtered out anonymous-1222 votes next update π
Mistral
sigh we need to make a petition to the owner of discord, to allow for an unlimited number of options :/
they will release gemini 3 ultra, veo 4 and genie 3
and more gemini 3 updates
tbh I just said Mistral because i'm french and for the lulz but yeah I don't see them competing anyway
Well, Google improved a lot this year
Claude, I feel like it's gotten worse since version 4.
I wouldnβt be so sure. Gemini 3 still hallucinates a lot and has problem with letters ordering sometimes, and Google wasnβt able to secure compute and memory at the same scale as OpenAI (just see google employees being fired). On top of that, Google is a huge corporation: even if they go all-in today, they can still reverse course for non-technical or political reasons just like they did in 2017 while they had gold in their hands with all authors of" attention is all you need" paper being google employee but they decided to screw up.
yeah even if the AI bubble bursts, Google can still crawl back to normal stuff
that fate is not the same for anthropic, openai and grok
and other AI companies
the question is "Who will release the most interesting AI in 2026?", not who will survive the ai bubble burst
I don't know who voted SSI but they have absolutely no plan to release model in 2026
they want to speedrun super intelligence
I think I might just be stuck in a verification Loop with how bad is this
I haven't even managed to done one retry
π₯²
Update: LMarena is now completely unusable. Google's captcha doesn't work; it keeps prompting you to complete captchas, which are completed successfully, but it fails the check and keeps returning the "Try Again" message. Why did the developers do this? Do you want to reduce user numbers due to excessive traffic? I really don't understand.
sometimes it's best to go out into the nature, observe the beauty of flowers while also letting the developers do their work without nagging them
just turn off the captcha
I am sorry to hear this, we are taking the feedback in this thread: #1451363565729747045
I'm not replying to you just because I'm a polite person, better if you go for a walk in nature instead of replying to Discord users
π
π
pretty sure it's xai with grok 5
come on dude 6 trillion parameters and 550k+ nvidia gpus is alot
best statement about the issue, so far
Victor, how about virtual nature?
(No Man's Sky, Minecraft, Terraria, Eco, Empyrion: Galactic Survival, Outer Wilds, Long Dark, Subnautica, Dont Starve, DinoSystem, Eastshade, Brutal Nature, Luanti/Mesecraft, FireWatch, Fishing Simulator, etc. :)
π
π
or.. watching relaxing nature videos on YT
(or cat videos ^^)
(..or using an easy language to code something on your own, for a change)
Or, just using AI-studio (at google), until LMarena works again.
Does anyone know, if Gemini 3 Flash is rate-limited?
(i know that Gemini 3 pro is rate-limited)
Where?
On gemini app?
10000 requests per day
thank you! that's great news
I have my resources, pal
so one could use g3f every ~30 seconds
(over ~12 hours)
that should be enough ^^
plus g3p
g3p allows how much? 10-15 per day?
You can also request to increase the limits
This is in ai studio usage and billing
3 flash has 10k requests per day
Meanwhile 3 pro has 250
for a plus/pro-subscriber?
3-pro ended earlier for me
250 cant be right, i had it ending earlier
Aistudio was made ONLY for developers
They can technically restrict you to lower limits
If you're using it for non-developing stuff
so AIstudio has a stricter ratelimit than the gemini site?
No
Aistudio was meant for developers
Gemini is for customers who chat
Gemini has subscriptions
When will an AGCI be created?
(AGCI: Artificial General Coding Intelligence)
And who will create it?
4
10
6
Google has become more strict
Since people were using aistudio for chatting
(me too)
I didn't had a subscription for google plan and used 3 pro freely
petition for AI to end human race
make a poll
do you know how
ah, too few options
yeah
It will end us though
π
what are these polls dawg
there you have it :)
ive heard from multiple prominent people in the ai field, that robotics is gonna be what 2026 will be about
that is actually surprising to me, would not have thought that we would get usefull robots in 2026 already
Soo maybe instead of learning to code
We create our first robo-projects
Yeah i believe too in ai vision + robots
so plumbers might not be save after all
Before it was like you manually had to program the robot what to do but now with efficient ai's that can run on laptops with ai chips
Fr i think nobody is safe
Everything we can do ai could do better
maybe sooner than we thought too
They're still faking robots tho
if robotics evolve a lot in 2026, alot of physical labor could be replaced
Hiring people to do VR jobs isn't a robot
we did not expect this at all
its just automation has become super mainstream, every venture capitalist is going all in on it
this mostly started happening in 2025
Anyone can code anything with enough patience and time
I can't imagine what would be possible in 2 years
I can't think of gemini 5 pro
It will come
Same as Samsung phones
They started with one
Now it's 25
Yeah they skipped few numbers but
Still
Gpt 3 was released in 2020
No multimodal capabilities
No image generation no agents
No web search
What's next????
World model
It's been on progress
Or they invented a new architecture that is not transformer, solving the problem of context retention, power cost and speed.
this would have probably been leaked by now though
Yeah they still work based on next word guessing
It's just autocomplete with extra steps
No real thinking
Just math and statistics
Even image generators are working on similar principle
It's like ray tracing back in 2010s it's there but it sucks so bad
Now it's everywhere on all modelling platform and games and basically all 3d related
Maybe they found a way already
But it's just too expensive
They want to stay silent probably
So they can hype it
I'd say actual agi robot is still far if the cost issue is not yet solved
It's like supercomputers... It's there for years but nobody is talking about it much
Maybe agi achieved in 2026 but for like 300$ per prompt? Yeah no thanks.
If we knew about some other way other than transformers it may be expensive so the hype will fade
Google is going wild in the next year
Do off site working ?
I heard they'll release 3 flash image gen
Bunch of their now beta versions of stuff
Project Astra
Genie 3
Can't forget Sima
They're preparing us for robots
So basically you could take photos of your house
And genie would make a simulation world out of it
And sima would practice doing tasks in it
We aren't ready for this frfr
overhyped
why is everything that isnt bad or harmful get marked againt tos
Which is the second-best coding model (after Claude) ?
10
12
1
gemini
which first?
10
13
2
gpt6
god, the recaptcha is testing my limit
βselect all squares that contain a-β
LEAVE ME ALONR
FFS...
I did everything correct
Even did the wrong box selection
Yet
literally lol like it was when I started doing worse that it seemed to start working
or like when itβs click the squares and wait for the next to fade in
gotta go as slow as possible if ur too fast it says youβre fake
Fuckahh recaptcha takes so much time that the response from the model gives an error
Bro i tried everything...
lmfaoooo this too
incognito tab
Real
it gives you captcha a bunch if youβve prompted a certain amount of time
Hey buddy, do me a favor
Iβm in a server and weβre in a competition, and Iβm losing. Would you vote for me?
Sure? Lol
Iβm in a server and weβre in a competition, and Iβm losing. Would you vote for me?
Uh
yo, where can i upload my code so the ai can see and review it? (I mean huge chunks of code about 15k lines)
hi Im Emmanuel , video creation brought me here
Good lord, what is going on with the captcha???
why the text to video option isnt there anymore, it doesnt show up
It's random rollout
Sometimes you get it sometime you don't
gah damn
Once you get it you shouldn't clear cookies and try to log in
Since it will make it poof
yea
If you meant on lmarena website
hm, i had no captcha problems today, just the normal one, which you had to click after each voting (that one which simply requires you to click the box)
If you meant here its (/) video
yea on site
Language Model Arena
gemini error again
free fps booster not a virus
@desert abyss here is a 101% legit fps booster for you 5 free.
My shift ended about 20 minutes ago, but let's use the banhammer one last time. Thank you for reaching out
Hello
hi
I have a question: how do I create videos from images?
cant generate more images, why?
hey
hi
u good bro?
Not really sure since for me sometimes that sort of thing seems to happen randomly I would suggest our cleaning browser history or maybe waiting a few hours give or take since sometimes that seems to help\
ok, gonna try it
Just frustrated with this damn browser since the verification has gotten to the point i need to use a private browser I can log on make a single chat that won't have the verification shoot up every single prompt and because I can't log on every time I need to make another private browser when it starts to shrew up again it just resets the whole thing
thx buddy
np
Hello, does anyone know any good website where I can upscale an image to atleast 4K? thanks!
HELLO
hello
hi
Friend, who came up with the idea of putting the captcha when re-trying a message?
It's annoying when suddenly an error appears, something went wrong
hello
I'm sick to my stomach I'm so mad at myself I deleted one of my best chats
Omg I'm so angry
π
...... I just did one out of sentiment but this was one goddamn message after me not using it almost all day if not 2 it wasn't even a prop it was a retry message
@echo aurora
fantastic, the exciting game of ULTRA SLOW REFRESH ReCAPTCHA puzzle coming back lol
Yeah I'm not sure why they're slow to refresh, that sounds like a bug and is something I'll flag.
I am sorry folks have been having issues witch captcha lately. All of this feedback is being shared with the team.
I am asking though we keep that conversation focused in #1451363565729747045. I'd like to leave #general for general chat.
This is a know bug unfortunately. If you clear your cache/cookies it should help.
Can you try clearing cache/cookies? Keep me updated if it doesn't help.
yeah I can confirm this
it very slow and loop until generation fail
hi guys
Hello
hi hi
A virus, nice try
Hey everyone - reminder if youβre having captcha issues please share your Eval ID in #1451574502369656842 . This Eval ID is in the url of the chat session.
Can someone send the following script to any AI (text/code mode) and let me know if it produces an error?
<script src="../../src/pineapplesTeam_Pleasefix_this/something.js"></script>
Bro, stop. I dont want to be unpaid slave labor for Google.
I am not training their AI models.
For me it worked
What AI you used?
Not much
Strange I think it's one of those cases were to filter is going misfire on your account or browser
@echo aurora
The problem is "create-evaluation" and "post-to-evaluation"
Still think it's what I said since I use your original one and had no problem but good for you hope it doesn't happen again for you
thats not the problem
the problem is the filter approach they are using
petition to start benchmarking 1B parameter models
they aint suck now
gemma 4 around the corner
their 1B model could potentially match ChatGPT 4.0
That's not gonna happen, not with such a tiny model. But Gemma 3 was and still is fantastic, 4b in particular. I'm really looking forward to Gemma 4.
Seriously is there any work around regarding the issues with the Claude Models?
Resending the entire list of prompt conversations to a new one just because crap broke down is not good
That's not a good system chief
worst thing about this is when there is a still a tiny bit of the object in a square but idk if it countsπ
what makes you so sure? AI progress has been incredible. Gemma 3 4B surpasses ChatGPT 4.0, why couldnt Gemma 4 1B surpass ChatGPT 4.0?
Gemma 3 was released all the way back in march
A random question I have for a long time; but why do we have to accept the cookies and canβt refuse them?
π€
Bro, why do you ignore me all the time in dm?
just put easy capcha , 50% of the time i fail... Am..am I a robot??
we're all robot
Working on a family fun project!
i cant use lmarena
Why?
βBangladeshi office desk, tired employee looking at cold lunch box, natural light, realistic style, vertical 9:16β
Which lab will have created the best AI system in 2026?
6
11
1
Deepmind
CAN SOMEONE HELP. THIS IS WAT I GET EVERYTIME I TRY GENERATING AN IMAG EOF TEXT.
Well for me personally it's just a verification of peering up more and more more often to the point that using private only works temporarily but it's so short before it starts to get into a repeat cycle that it's honestly not worth it
Its funny how Flux series is the only one that works today
how can I make videos here ?
some cookies are essential and stores your session/tokens ( authentication )
and there are some for analytics/fingerprinting etc... where you can refuse them
Best local (offline) coding model?
4
9
3
GLM
shut up
lmao
π merry christmas yall
which openai model is 560B
No no
These are the tokens processed
On Openrouter
oh
Deepseek is catching up fast
tbh deepseek v3.2 is good
It is
Very reliable model
I think weβre gonna see a lot of companies stop using gpt and Claude once Chinese models are effectively equal while costing 10x less
Airbnb already uses Qwen because itβs extremely cheap for what they need
A lot of American firms started using Kimi K2 due to superior performance in their use case
DeepSeek is popular with a lot of analysts and such
American AI companies are very greedy
they wont provide models as Chinese companies do
That is fairly misleading. If you look into any open-source company from China, most of them are state funded. They do this not because they like to burn money setting it on fire - which is what western company would have to do... They do it partially to project power/dominance on a national level while being backed by their government to do so and remain well funded. Ironically if you look at their local pricing for domestic market, using those models for them is not cheap at all. Last I checked pricing was in-line with what your average American pays to use API of the best western models (and we know that pricing already includes R&D and is well beyond cost to run the model)
nano banana pro not working
I did check Alibaba cloud for the record. Chinese market website that changes pricing as soon as you change your region lol
try the non 2k version
it worked for me instantly
You can't really pour millions into training a model and then be in a position to give it all away to open-source with no reward no monetary value in return. That is just not feasible if you want to stay afloat. And you aren't state backed with the goal of serving your country globally or whatever.
It makes perfect sense for China to tackle open-source though, since that field is less competitive but still puts them on the map, with opportunities to influence global market.
lucky u
How should it be called?
1
2
Check out the information in #1397655624103493813
Let me know if you have any questions.
https://www.kimi.com/kimiplus/sale?activity_enter_method=popup this is kinda fun lol
Guarding the deals. Bargain with Kimi's bot and knock your subscription price down. See full rules
Merry Christmas and Happy Holidays!
If you're running into this error I'd recommend following the steps in this article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message
You may sometimes see the error message: βSomething went wrong with this response, please try again.βThis is a general error message. It can
I've been meaning to get back to you, just have other priorities atm, but will eventually get back to you.
@echo aurora my nano banana pro is not working:(
We are aware of the high error rate with this model.
yeah :(
why google API is always so mean with lmarena?
it always crash
This is something we're collecting feedback on. If you have any suspected false positives be sure to share in #1447983134426660894
Why do we need these TOS? All models are highly censored by themselves.
gotcha
It helps ensure the platform remains safes so we can work with a wide variety of labs.
so people dont spam LMArena with bots
That charge a LOT for API by spamming direct chats
Shutting down the website
Plus fake votes
I don't understand people who complain about captcha at all.
I would think the original one was doing a damn fine job of that and honestly this much out of trouble for this is honestly not worth it for a test if it's going to be browser wide
he
Bots is a main issue for vote manipulation I would think
Less so for direct chat
This is starting to appear too much, has there been any changes recently?
disable your VPN
Dont have any active vpn
Even to retry a message it asks
It seems they weren't happy with the filter and decided to implement that.
Exactly its annoying
i dont even get the chekbox lol
just a blank space
so i reload the website
It makes me select the boxes thingy for like 10 pictures, its getting annoying
Lol, and where do you think those filters are?
Obv in those requests
Client -> filters -> send message to AI
duh
its obvious
but thats not the issue lol
thats not the issue
thats just the post request / the medium to the server-side
Prompts i used 2 weeks ago suddenly are not working anymore, they are now considered violationπ€¦πΎββοΈ
everything makes sense now
no reason to use gemini anymore
its garbage for me lately