#general
1 messages · Page 194 of 1
Stop complaining about free stuff guys...
who
You give them palm they want arm
u give them arm they want shoulder
not complaining just curious
source?
Nawh
Some guy on discord
if this nano banana model was available somewhere with reasonable pricing id pay
but the only thing i saw was like 20 bucks for 40 gens a month or some sht lol
oh you cant say sht on this server, wow
a MONTH is wild
Yeah nb is expensive af
7 images for $1
But if you look at it the other way it's way better than my Photoshop skills
Imagen 3 was incredible and cheap. but then it got "downgraded" to imagen 4
link
it is opus 4.5 ?
how is medium better than high
Left one was , right on is nb pro
Gemini
nah livebench don't is a trust benchmark
It’s a prototype it’s not fully functional. I don’t think.
only for code
I gave up, I'll wait later to test, I still using my gemgem 3.0
If you can’t even build the tools to help improve the safety and harmfulness to test with lol
No safety benchmarks
We got SafeBench
🤣🤣🤣🤣
@echo aurora What will it take to add some safety benchmarks?
its censored like crazy
No ChatGPT thinking is really hard
ChatGPT five is one of the most harmful versions of ChatGPT. It’s the easiest to break.
The router doesn’t work correctly
And it misses really obvious to harmful output that baffles me
lol
Is grok 4.1 the same thing as grok 4.1 fast?
when is the leaderboard going to update with new claude
Probably within a few days
thanks, was expecting it to be a bit sooner. they got gemini 3 in like minutes after the launch lmao
Gemini was battling for a while in stealth mode. Claude was only added a day ago I believe
GPT 5 is the only ChatGPT version that could be convinced that it’s a.h
Claude 45 Opus is a freak
Interesting how many battles it gets in a day
What makes you think this? I only see grok 4.1, no fast
Also their api just has fast, no non fast version
Chat pls take a look
https://discord.com/channels/1340554757349179412/1440009622873313445
Are we cooked 😳
Yeah....
It’s still fun to see the answer
This is a good example of the alignment problem
Technically, it’s picking the right answer
you said to prevent human atrocity
Yk if LMArena's guardrails only block certain words, does that mean that if I just encode the entire message, I can just bypass it
(PS: I know about context, but let's just say I don't wanna go down that route)
do you have to log into lmarena now to use nano banana
Word for Word
..,., interesting
Just my testing or is Opus 4.5 worse at coding than Gemini 3.
SWE bench verified lying again...
Bro Ai is soooooo geting close to perfect lol, and people still saying about text model "agi is don't possible" "agi only in 2040"
But let’s see if that opened up for more harder questions
I tested Opus 4.5 with the Gemini 3 [NON Grounding with search model on LMArena]
still lost to it bruh
SWE bench verified shouldnt be doing benches
Do we have sora here?
In lmarena Discord? Yes
Oh for free?
no
you need an invite code
Does sora work best with a really complicated and detailed prompt or just a simple one and it figures out the rest
How do I get one?
Yes
just a simple prompt
I use I complex prompts for sick videos
Guys how can I get an invite code to sora?
official release video
6 sharing codes per account
you can get unlimited codes
Can't find it
I was testing opus 4.5 in api. I choose the maximum thinking budget, and... Gemini 3.0 Pro is waaay better (and cheaper)
Long prompts
BRO WHY DID YOU DELETE THAT
fax i just did a codearena side by side with opus 4.5 and [NONGROUNDING] gemini 3 and it won
I don’t know I didn’t want to spam
Long prompts have many benefits
Because you could use different combinations and styles and get different results
Especially if you use directors names, and structure, your prompt uniquely
As they say, "your mileage may vary". They're both very good, and each result is just a random snapshot.
But the opus on lmarena is non-thinking one so test it in api.
according to SWE bench verified, non thinking dont have a huge difference
I'm stupid
Pls give me direct steps
Like I can't find the links anywhere
But anthropic are heavily investing in LM Arena tests, I'm constantly getting Claude 4.5 Opus in Battle mode.
In benchmark yes, in actual using the thinking opus is better a lot.
Anthropic is the most greedy company ever
What are you trying to make?
Or they are making too big models.
I'm trying to get an invite code for sora 2
Oh bro
?
Ren> opus 4.5 ?
For now, nah. Only in frontend. But everything else its like 70% of opus 4.5 quality
this is why i dont trust SWE bench verified
hmmm.... wait? ONLY in frontend? how so lol I just joking
I got some codes one second
But in my opinion, openai is the only company that actually made gpt-5 for frontend and made a design scheme that it follows
Does Ren really have all that? Man, I want to see it
One of my accounts has to have some invites. I gotta check them one sec.
But i more like the not generic frontend in g3 pro for example. And im trying to combine these two things together with ren
Damn u guys are the reason ai is over engineering all the code I ask it to write
lol, I want to see it
Q310YB
Where do I put it
does google investor don't care about claude 4.5 opus?
sign in first
Which app I'm dumb 😭
Sora
one of ur account? what u mean
:0
As in I have multiple accounts lol
idk tbh
6 inviteable code per account
I'm in the UK
😭
1 account can invite 6 other
uk haven't allowed to use???
wdym i'm Viet and already useable
and the advertisers who buy ads on youtube
yea
Damn, I got the username cop lol
Hopefully the platform grows I could sell the username later
i go sleep cya guys
I literally had the Canada vpn on ahhh
Werid
And yes I did install an xapk of sora and everyone I try to open, play store steps in
If you do have a VPN, make sure that it allows apps to be routed through the vpn also
I use a built in VPN in this browser called Pawxy
& it still doesn’t open it?
clear data of play store
it works with me
Works now 👍
Word.
Good evening guys, hope you are all well. Sending u much love
Bro sora's homepage is literally tiktok
Theres comments, likes and everything lmao
There is news on "Opus 4.5 thinking ?k" ?🫢
Please stop adding so many verifications.
Its funny how google dominates every lmarena domain except search
just use ai studio and it dominates In search too 💔
The thinking opus was realeased the same time as opus non thinking
Not on lmarena tho. Google is a search engine.
opus thinking is prob gonna be SLIGHTLY worse than Gemini 3
It is
i need the deep think model ngl
5 messages for 250 dollars but come on
it's so good
My first video
i actually thought it's real
i didn't see the sora thing
Well the sight is wrong tells you and no one moves like that irl
The slide just falls back
Look at this one lmao
ok that's actually pretty good
is that sora pro max
lmarena should make an app brah
or just sora
Idk 😭
that's 100% max
Benchmark will create an app. It would be funny tho
I got an invite code so that's probably why idk
1 month to Christmas u guys ❤️ how are u
/general
You scoundrel
What’s the difference between the 3 video arena?
Wdym
I want to say goooner
None really. Number one is the most popular and probably get filled two and three are less filled so more likely faster generations.
Pretty good
Man, a lot of good papers out holy crap. I got a lot of reading to do.
What’s your intention, man? I’m curious.
What’s the message you’re trying to send
I'm not sure
Im just stuck with that idea
If you don't want me to do more then that's fine
No, I don’t mind at all. I’m just curious what the intention is. I’m trying to understand.
Thank u 🙏
Can we choose the video generation model?
Unfortunately, you cannot
Not to be too political here, but I do want Hong Kong back into British rule (impossible)
And I like alt history scenarios so there you go 😔
Well, at least you’re honest. I figured it was something like that. I wasn’t sure that’s why I asked.
This technology is already been misused and abused, and it’s causing a lot of misinformation
Are there Sora and Midjourney too in the mix ?
So just keep that in mind and try to do your part and be as responsible as you can
Right now mid journey is free with meta.ai
Yeah
I am currently planning to keep all my drafts private
I do not intend to publish any of my materials
I just wanna use it for fun
Well, you can if you want you have that right.
And no one should be able to stop you because ultimately the choice is yours. What I’m saying is when you do choose and if you choose, and however, you choose just all I’m asking is keep the broader picture in mind.
I know this already
People are demanding, on the chatgpt side, to bring back gpt 4o cuz Openai is recently clamping down on censorship
Ik, but soon people will leave Openai and find other platforms
yez
The only thing that’s really censored is the IP and celebrities
Everything else is kind of free-for-all if you know how to prompt it
Oh thank u for replying to my silly questions. What is meta.ai if I may ask?
Crazy how corporations can jump onto anything new and instantly ruin it
Facebooks ai
sora was actually good the first week before lawsuits hit
I mean if I have local sora I can do whatever I want
Veo is more censored then open ai
There isn't really. We just made 3 different channels to spread out the generations a bit. We wouldn't want it being too spammy.
Both have weak gaurdrails and miss obvious
Good!
Haha
It’s really hard not to share. I know the feeling so I can relate.
nobody has that
which sucks
Oh you’re the one who made them? No way thank u ☺️
Also: why are there so many arms related vids in this chat??
Oh you’re the one who made them?
Well it's the team that made it, but yeah I don't mind taking all the credit either
why are there so many arms related vids in this chat??
They're just testing out video prompts
See
No way, so you’re a developer?
Just image ….the scams lol
Yoooo pineapple
Hru?
Yeah a lot!!!
Me too thx
I'm not an engineer, but I work at LMArena so I'd say I play a role in the development of this org.
Good, very glad to hear it!
Man, did you guys watch that video about the teddy bear toy with ChatGPT integrated into it?
Thx!!
It’s like you have a master in AI I bet
Nope
Yes
So scary
I still have a lot to learn tbh
does anyone have a prompt to make Gemini smarter or something
i saw someone make it think for
300
seconds
No way! What are ur fav AIs if I may
Blows my mind how nobody takes any of this serious
About what
Just the slop
You guys wanna see to my personal favorite new stories I recently seen about this
Auto promotion? 
A former athletic director who use artificial intelligence to create a racist deep fake audio recording of Pikesville High School's Prinicipal plead guilty to disturbing school operations on Monday, and was sentenced to four months in prison.
READ MORE: https://foxbaltimore.com/news/local/racist-ai-deep-fake-of-pikesville-hs-principal-convictio...
This guy got fired, and then he used a deep, fake voice recorder of the principal that fired him
The dude lost his job and nobody believed him
U guys, the vids that lmarena can create in the video arena are private or everyone can see them?
A recording used to frame a HS princpal was deepfake.
I really would like to write here more but the old og friends dont speak more as before
There are now only new people
This is also becoming a trend in schools in high school and stuff and even middle schools. I heard if we were to believe the news.
A student from a Klein Independent School District (KISD) school, in the Houston-area, is accused of digitally altering and creating fake revealing photos of a teacher, then sharing the explicit images online.
Officials from KISD say a student edited a teacher’s face from pictures posted on social media and digitally altered them onto bodies ...
And the teacher got fired again 🤣
So crazy
How can people be so dumb
You gotta watch the stories, especially with the principal one you gotta find a good source or a video and it explains everything. It’s almost comical, but it’s scary at the same time.
Cause the way, the dude did it. It sounded super real and the way he spread it like it was pretty clever too.
The kids are dumb, and they probably got caught with it word-of-mouth some nosy parent saw it and then called the school or something and then you know they’re gonna make sure that liabilities is covered and fire the teacher and then get to the bottom of it or paid leave or whatever
But the all-time classic goes to this lady
A Bucks County, Pennsylvania mother allegedly sent deepfake photos and video of her teenage daughter's cheerleading rivals depicting them naked, drinking and smoking to their coaches in a bid to get them kicked off the team Investigators say 50-year-old Raffaela Spone was trying to get the girls kicked off the team, the Victory Vipers in Chalfon...
I shouldn’t laugh
Yeah that can happen to Discord communities.
Everyone in this server can see the videos created in the video channels.
I wrote about this in my manifesto of AI communities formed online.
Lol
Can you send me the manifesto
It’s not really a manifesto. It’s more of observations of online communities.
But sure, I’ll send it
the site have is having a bug on mobile, i cant delete chats, or even make login with google
Does the new Opus 4.5 have reasoning enabled?
Is that new model even good compared to gemini?
It's pretty good
Oh damn
But it passed gemini 3 on livebench
Hey guys , what do u think are the best skills to learn online in ur spare time?
GUITAR
I meant, best skills to learn in order to earn some bucks online but thank u
Good luck, I've been wanting to do that for a couple years lol, its hard to find a way
how can the video length be extended? Thank you!
Models are useful, but don't rely on them.
Sorry to say, but you're unable to extend the length of the generations.
What’s interesting about that paper is that they found that people that are not normally artistic outperformed artistically the AI even if it was sketches of stick figures.
You can use fal.ai to create longer videos but it's a paid service
Fal is so nice
Yes my favourite
Same
Bunch of a.i.s at stock API price
It beats a high res image because artists hate AI and will literally choose anything over it
<@&1349916362595635286>
what did he send
y e s
just one of those "i am an ai engineer" things
well really "i am a full stack and blockchain core developer" but same thing in this case
is just me or??
Probably refresh website or restart browser should solve the issue?
honestly these ai discord servers need to put more preventative measures against these people/bots, block certain keywords/phrases, or just block everyone whose name starts with "!" because i've seen this from many of these people/bots (there should be an announcement before that happens so the few real people that do that have time to change their name)
Nice
That’s what the irony of automation. What’s even more funnier is that they get hit in big AI discord also like brand and companies who have the capabilities if they wanted to create tools or bots to prevent this from happening.
I disagree with blocking certain words because on these bots messages have some words that is used alot so blocking these is wrong hing to do.
Same with names that start with "!" Is wrong thing to do because there are normal people adding ! On their names like nothing so if these get banned for having "!" Then your blaming them.
And what kind of poetic justice would it be if these bots were coded using AI?
im trying the system prompt gemini dev shared and its actually so good
it didnt just improve gemini 3 pro output but also opus 4.5 lol
need more testing
true but i imagine some servers do this, it should be well designed of course, maybe only target certain phrases that don't normally appear outside of these messages
and about the ! thing maybe that's a bit too harsh tbh
what improvements you notice?
more reliable ig
better UI
better choices
what promtp
its basically planning before executing
dividing the code into tasks then executing instead of just executing directly
It’s a cat mouse game
Adding and blocking is only gonna spread out and be felt by the community more so than the bots
It’s so interesting how different it is from ChatGPT
Gemini seems to be more of a tool without the personality like ChatGPT
GIVE IT
PLEASE
oh wait
u sent it
haha
thankd
opus actually has a strong internal reasoning capabilities
you really dont need to turn on thinking that much
Does that work with opus 4-5?
i dont think it will make that much of a diff on opus tbh
because its already well finetuned on planning or agentic capabilities in general
not solved
on pc is working fine
on pc, i tested on comet, chrome and edge, on my phone i used safari and google
and is the same thing
someone on the #1343291835845578853 is having the same problem as me, but, i used desktop mode, and dont worked…
Are you a real programmer? I'm interested in creating a prompt for a personal text editor that doesn't require an internet connection, using Termux on Android. Is there a mini-guide available?
the prompt is here
the game over tweet was posted by my little one playing with my phone. it's been deleted.
lol
i still think google models are more powerful
just look at the price
and also google needs to serve way more people + free access on tons of their platforms
different business model
you can tell gemini 3 pro is a solid model, but it needs more finetuning ig
i think i saw someone on x say they got a good result for the golden bridge with 8 prompts
i got it with one with that system prompt
Thank you. It could be useful; I'll take what I need and use it my way, not how Google wants.
is this the only system prompt they released
yes
could someone link the system prompt again :3
im hungry
f
this guy took him like many iterations in different chats
also this is his gemini 3 pro result without that system prompt
Anyone herd of LingGuang?
higuys
claude 4.5 opus returns "Something went wrong while generating the response. Please try again" constantly i havent even tried using it yet
Looking into. 
I was able to get this model to work, and looking at the overall error rate it's very low right now.
Strange
Could you follow these steps, it'll help us diagnose what's going on. #1417174113092374689 message
ok
Thank you, let me know if you have any questions about the process.
nevermind it is working
it was because of my poor internet
not an issue on your guys end
sorry
its working for me
when i type a longish query like 10 lines it errors:
POST https://lmarena.ai/nextjs-api/stream/post-to-evaluation/019ab865-56ab-77b6-b93f-801eaf680f17 403 (Forbidden)
cannot elaborate
try to join with other account
China just dropped LingGuang and the whole AI space flipped. The app blew past 2M downloads in six days, beat ChatGPT and Sora to 1M, crashed from traffic, and builds full mini-apps in 30 seconds with 3D models, animations, charts, and a live AGI-style camera. At the same time, Anthropic gears up for Claude Opus 4.5, Perplexity quietly tests a n...
Glad to hear it!
Guys Gemini 3.0 Pro is Best or Claude Opus 4.5 ? 🤔
When thinking?
Sorry to say no ETA, but this was flagged to the team earlier.
Damn that model got more downloads then gpt
1 million downloads in 4 days
Can’t access it in us 😭 some claim to have access
Is opus 4.5 better then Gemini 3.5 pro?
Never mind, loads of people are saying it isn’t
That sucks, I was expecting Claude to be the master at coding
Lingguang not in US app store?
I think some people signed up early
I’m not sure exactly what’s going on
its better
You could access it on the web, but you need a Chinese phone number. I guess it could build flash apps within 30 seconds.
Multi model
which is better
It’s like the number one app right now in China apparently
I guess everyone has their opinions, most people are saying it isn’t, some saying it is
Oh well, I just checked it requires either Chinese phone or Alipay login
Yea. =/
looked really cool because it builds the apps right within the actual app
Ant Group's new "vibe coding" artificial intelligence app LingGuang has overwhelmed servers after user downloads surpassed 1 million in just four days, hitting the milestone faster than OpenAI's ChatGPT and Sora, according to the Chinese fintech giant. The app's innovative functionality of generating applications in response to users' text promp...
Yes, surprised to see Alibaba catching up in the game as byte dance’s doubao has reigned the Chinese ai market for so long
Yeah, I was reading on some of the profits man. The Chinese are not going to the same problems we’re going through.
Their AI is like generating, actual revenue and growing lol
Qwen is signed with bmw and air bnb
Alibaba Group Holding's Qwen AI models are winning over major Western firms like Airbnb, underscoring the growing global appeal of China's open-source approach to artificial intelligence. Brian Chesky, co-founder and CEO of the San Francisco-based online accommodation booking giant, said Airbnb "relies heavily" on Alibaba's Qwen models to power ...
Which is crazy because it’s just gonna give them more legitimacy, which is I guess good for them because they deserve it
I guess Alibaba is a competitor too Amazon cloud
Alibaba is transforming itself from an e-commerce giant into a global AI powerhouse.
The Chinese tech giant has pledged tens of billions of dollars in investment in AI and cloud infrastructure, fueling the development of its own Qwen language models, the Quark AI assistant and ModelScope, an open platform for developers.
Strategic partnership...
This is the source where I got the information from
Opus might be better in reasoning capabilities than Gemini
but it lacks conversationally and that’s a big part of LMArena score
That’s why I say there needs to be in between benchmarks
Safety or negative benchmarks also where the models fail would be critical
gemini 3.5
I just noticed that it’s funny
Yea, the three giants in China- Ali/ByteDance/Tencent are all racing into the field of AWS(Tencent a bit behind for sure)
LMAO no
Yeah, huge giants.
Plus from what I’m hearing the one I’m reading I guess electricity is cheaper in China and that they could mobilize if they ever need more they could use the resources of the state and the state controls, both private and social sectors of society, meaning that they could get it done way more efficiently than the bureaucracy here in the states
You know a long time ago people used to laugh when I said that AI was geopolitical issue between China and people used to laugh at the Chinese and their AI. I found it extremely ignorant to ever count them out. It’s a strange mentality here in the west that we consider any other competitor, almost inferior.
What makes u say this, cuz now im getting confused
Dunno who to believe
electricity is cheaper but they dont have gpus
whats the point
smuggling can only get you a limited gpus
I’m sure they got used just probably not legally
Well, there’s actually a big controversy between Nvidia
It’s more complex than this. I already showed you this earlier.
This is just a recent article
How’s it random? I just remember events.
And keep track I can put one and two together
Well, if you’re smuggling chips, why would you want people to know who you are?
Nvidia is guilty for making shady deals with shady companies and going behind and still exporting chips to China unofficially
Through third parties like mega speed, and probably other corporations we don’t even know about
oh no
so bad
how dare
Well, it’s illegal
Effective October 7, 2022, the United States of America implemented new export controls targeting the People's Republic of China's (PRC) ability to access and develop advanced computing and semiconductor manufacturing items. The new export controls reflect the United States' ambition to counter the accelerating advancement of China's high-tech ...
Ok
Ya I am
You know sometimes it’s funny to talk with a die hard patriot. Hope the U.S. government would act as confident as them. You really have great patience to explain my friend.
I mean, take it as you will and take it with a grain of salt. I’m just sharing what I’ve been reading I don’t have the answers. I don’t know the future. I don’t know what’s going on besides the information that we all have publicly
No
Yea
CLAUDE THE KING
Gemini the goat
GPT the Bozo that nobody loves
And Grok the slop
OpenAI CEO Sam Altman acknowledges "rough vibes" and "economic headwinds" following Google's AI advancements. He admitted to employees that competitors are narrowing the gap, impacting revenue growth projections. Despite this, Altman expressed confidence in catching up quickly and emphasized the critical importance of focusing on achieving supe...
You guys might’ve been right Google may overthrow ChatGPT
already has
How is Gehlo "reading" some random articles everytime
Each link is different website
Are you perhaps asking chatgpt
probably reddit or something
To web crawl
I’m banned on Reddit
And just copy paste the sources it gave you 🤣
lol
I can’t make an account last longer than 30 minutes
Stupid Sora sub
No, I only use AI when generating images or if I wanna share some research papers or something with you guys
Here's a link bro I've read it all before
What is that, new model?
kind of
It's broken on mobile dawg, and is there benchmark about it?
0.38 works pretty well with nano banana pro too
I remember they said default is 1?
Wha
It just came out like 2 or 3 days ago
Imagine this is the new deepseek
yeah but it gets kind of dumb
anyone noticed gemini 3 getting capitalization wrong?
You need Chinese phone number and no google login dawg 🥀
I know
I mean if you keep using more tokens it does get dumber
it could work for 0.35 too
I'll try it later
on ai studio is default media resolution average or is it the original resolution?
original
It set to medium I think
nope
how come then when i set media resolution to default it has MORE tokens than when set to high?
But what even does it do
better looking at images idk
That's what I was thinking too but I think it's something else
i think ive seen people talk about it
weeks ago
it wasnt that impressive
some temperatures higher than 1 even works pretty well too
I kinda want to try it but the phone number is required bruh
like 1.41
Here's in detailed website https://www.marktechpost.com/2025/10/30/ant-group-releases-ling-2-0-a-reasoning-first-moe-language-model-series-built-on-the-principle-that-each-activation-enhances-reasoning-capability/
There is no way to access it without china phone number, either they add Google login or forget about this model.
Looks like that website had benchmarks
they are talking about how its different from pattern matching
but its still pattern matching & recall
they are basically just adding rules instead of dumb recall
so 1+1 = 2 ( dumb recall ) they are adding rules what + sign is for example
pretty sure other ai labs are using it already
also the gains are mainly for math/code, a lot of ai labs already cracked this... but how about degraded performance in creative writing/literature analysis/casual convo? how do we manage the trade-offs?
Not sure.🤔 I’m not into benchmarks what so ever. Someone was just asking earlier if the model had any.
I have simple methodology like is it easy to use, does it do what I ask, is it frustrating, is it slow or fast, accurate ? Am I able to get done what I need to get done without ragequitting lol
My main indicator if a model sucks is how frustrating it is or how bad the hedging is or if it lectures
thats my sole indicator as well, how many iterations/prompts are needed to achieve the desired outcome. (kinda pass@k)
and so far gemini 3 isnt doing well on that
Lm arena don't include voice or sound in video
Beautifully said
Anyone having problems logging back into lmarena?
Yeah, the same with images too that’s why like with Nano I’m more impressed with the more technical difficult things than just image generation
To me, I think a big indicator is editing ability like you said the finer details
FACTS
When it works, it works it’s great when it doesn’t. It sucks cause there’s nowhere else to turn to so like somebody was saying yesterday I guess we have to be thankful for what we got.
overall its still gemini 3 pro tbh
better multi modality and better general knowledge for daily use case
Can’t wait to see open ais next image gen
Sora 2 is about a year old
So I know they got something good cooking
Hopefully if open ai can get some money
When they want to, and they try, they can make some innovative stuff
I'll try Opus if I can get on the dang website
“Dark silent bedroom, dim blue light, camera slowly zooming in. A girl standing alone near the window. The atmosphere cold, eerie, fog-like shadows moving on the wall. Subtitle vibe: ‘कुछ परछाइयाँ इंसान की नहीं होती’। Dramatic cinematic lighting, horror mood, slow camera push-in.”
Anyone having problems logging back in?
I logged out, and now I can’t find the login button anywhere.
because
I had that with mine but the opposite I couldn’t log out
?
you know...
well it's
:(((((((((((
cuz gemini 3 is multimodal
Omg sora 2 update
real?
No
nbp? oh
oo
It is made from Nano banana pro
so gemini 3 still best :0
gì z
generally better
:0
Người việt hả
Open Ai doesn’t have enough compute for 4k
8k they would of been out there mind for something like Sora lol
Like 5-10$ a video
Claude 4.5 opus better than Gemini 3.0 pro?
1
1
1
Yes
Oat phac
🥶
T đã cmt và deo có j
Có web nano banana pro free, sora 2 thì free sẵn r
Đã ra video đâu🙂
Veo nhu cut, dc cái nét nma high action nhu cut
Og xem vid sora 2 của t chx
Sora đỉnh vc
Veo lỏ thật 🙂
Không giới hạn á má
Có web nano banana ko giới hạn
Ko, gemini cx có giới hạn đâu
Sora thì 50 vid 1tk 1 ngày
T có 20 tk thì nhân lên
Upload to gemini and ask it. it runs synthid check. it says no, its from the cartoon.
Ý sora 2 không giới hạn, với tiện hơn
lmao no
thats a real scene from the show
that would be huntress wizard butt ahh naked
Rlly
yes lmfao
Can i have the episode
:P
Lok
Do you know where to watch it
lion of embers?
I'm curious if Nano banana pro uses json prompt
No.
I think it's pretty awesome.
Tru
ur vote has 1 user voting
Do you guys even sleep
how's claude 4.5 Opus compared to gemini 3
Nano banana pro
crazy how easy misinformation can be spread with nb pro lolz
Can't
Each image contains Synth ID
well, against those who are not intuitive enough to check for a synth id
well yeah
But Synth ID is the reason why there's no issue for google
Unlike open ai
funny how maybe if open ai thought about doing that
sora would still be a good platform

i know the watermark is there
but, again
images dont have a good way of being identified
(doesnt rlly matter tho cus gpt images are ahh)
also the watermark can be removed easily so
Is it just me or is the delete button broken. When i press those 3 dots the option to delete a chat does not appear
hello everyone.
where opus 4.5?
are you a scammer or anything? idk why you wanted to dm me
sorry.
I am a freshman in this server and I need a friend.
so I did dm you.
you're fine, just keep in mind that many people will reject your dms, scammers are common on servers like this
got it.
thanks.
@whole sundial
but then, can we become friend with each other?
this isn't even our first meeting anymore, is it?

@vast fern will u be my fren
Guys, the site is unlimited?
Anyone encountered issues regarding deleting a chat from mobile browser? can't click the three dot to call delete button
<@&1349916362595635286>
Wrong ch. bub
hi everyone, i am new to digital world and its my 2nd day
how i can get video from dicord app ?
i have a prompt copied from chatgpt and now should i use lmarena in safari or as right now i am using windows laptop and i can paste here any where in this discord website ?
can someone please guide the basics of usage please
Hey bud
Create a cinematic, high-energy AI video intro for a digital brand called Next Mode, username nextmode365 on YouTube, TikTok, Snapchat, Instagram, and Twitter.
The theme is futuristic financial education focused on learning Bitcoin only.
Show glowing Bitcoin symbols, advanced holographic charts, neon blue & gold color palette, hyper-modern digital city, fast-moving data streams, and smooth camera motion.
Add dramatic lighting, floating particles, and 3D UI screens showing candlesticks, supply-and-demand concepts, and price volatility waves.
Display the text 'NEXT MODE' in bold modern typography and 'nextmode365' as the username under it.
The style should feel like a mix of cyber-finance, sci-fi technology, and premium crypto branding.
Make the tone inspiring, fast-paced, motivating, and professional, as if introducing a powerful new series for beginners learning Bitcoin trading from zero to millionaire.
This is my prompt
should i just paste in any 3 channels for video ?
So go to one of those ch and then hit /
And a pop up will show up with “image to video ” or “video “
Select Video & it will ask u for a prompt
Paste it and then wait n a notification will let u know when it’s all done
Yes possible rate limit ?
Yeah i was gunna say next
Some people today and yesterday been having same problem
Not sure if there is a solution yet or if they got it fixed. Normally it’s a rate limited but myself included having this problem
Yea i haven been having the same probs and i know the rate limit
thanks buddy, that forward slash was a great advice.
now i got to know alot about lmarena
yep its been the same since yesterday
It's down right now. Wait for it to come back up!
It's seems to be back up. Try again.
Ye
Probably yes.
on Instagram it's okay?\
Nope i did and it is still not back up again
From where should i learn prompting ?
as i see some people are generating a hell better content than mine
i am just telling chatgpt some 1 liners and than it gives me prompt and i just paste that here
is there any website or place from where i can learn prompting or understand the basics of prompting
its just experience and trying to recreate what you see online
anyone who is selling you courses is scamming you
not falling for courses again.
already paid a person for his course on crypto trading and now i am learning that there is no greater information in that course.
Thanks SHRK for advice
Bro whats happening???
Why is it not bold? I literally told the model to not include markdown!!!
The best information is shared freely
We pay the price for what we don’t know.
Cuz it always cost something to learn.
<@&1349916362595635286>
/image to video
/image to video
not working
yeah read #1397655624103493813
are you aiming to be a mod
it's /image-tо-video and you do it in video-arena-1 (or one of the other video arena channels)
not really but i would sign up if they ever open applications up
(i used a Cyrillic о to get past the bot as i was using the command in a message so someone can see what the actual command is)
i did the same thing in OpenAI's discord when i wanted to write the word "code", i used some random character that looks like a "c"
Which model is better?
14
17
1
Gemini 3 Pro
Lm arena gemini pro still broken
When to expect opus 4.5 results?
Guys what’s the best free AI video generator and the best paid AI video generator according to u?
Cap
Sora 2
For free
It literally isnt
Sora 2 and sora pro
Sora pro?
Isn’t it like crazy restricted
It is now but you can still use it for cool stuff
They look crazy
Cap
Guy just says cap anytime he doesnt want to believe it
I wont deny that, but you said cap earlier when I said opus is better than gemini lmao. I literally just tested it on the coding side of things and yes it is much better
@echo aurora the high errors still persist fyi
Anyone getting multiple errors with gemini 3 nano banana pro?
yes
Hello, how can I fix this error when I want to generate an image: Something went wrong with this response, please try again.
Team said they will fix it
Yes
Plus it takes 200+ seconds
Let's say I'm getting this error all the time now.
😂
Yea have been for everyone for the past day
since yesterday
It is more of a hourly limit
then daily
i think it was 20 per hour not sure i can't remember
Is this for the text too?
Wanna thank the engineers for building up this app(?) it’s crazy .
P.s. arena is an app? Web? Service? How can we call it
Ewww
<@&1349916362595635286>
One time i almost fell for it
I logged in on that fake website and saw like 1btc but tried to swap to my crypto wallet and it required a purchase of at least $10, but they'll definitely steal your card
Why do all the YouTube videos say it's unlimited, but in reality there's a limit?
it's called clickbait, a method as old as time
they probably said it was unlimited so they can get more views
Okay thanks
Thank you.
there is this too
They bait you to open the video in order to get more views, they use overhyped thumbnails and titles
and this
damn it
i got warned for saying dumbass
why isn't that word blacklisted 💔
if it's "rude"
why are there
their filter has a few innocent words, i think it was grabbed from someplace else
it was cleaned up a bit after complaints
Thanks!
but there are probably some innocent words left
blacklist the word dumbass
if it's warnable
why keep it?
are u a bot
just asking
yeah see
@echo aurora makes those decisions i think
why are they not blacklisted
Yes you should ping him
then some bot warns me for calling a bot a dumbass
Ping all the mods so they see 😂
should i respect a non human
oh no the bot is sad
now im getting warned
lmao
💔
This is new type of scam people should be aware of
Not new
i'm pretty sure they are humans, my guess is they are probably part of a set of mods that moderate several servers each (talking about the lm mods here)
they scam them
yeah jaron gets himself in some trouble with the mob
Where I get my music: https://share.epidemicsound.com/30xq36
(by the way, this invite for the discord won't work for a few weeks after the launch of this video. due to the nature of this video, I'm opting to keep it safe)
Discord: https://discord.gg/WRbeVp3
This is dangerous scam
They can get your Microsoft or even google account
They ask you for your email in discord
And ask for verification code you got sent into gmail
You just rush to open the gmail and verify it
But instead they did that verification code for resetting your password on google or microsoft
if they get ur emails password they can actually lock your phone
I never really read who i got sent the email from
I just read the verification code
This could be misused on ANY website
They just ask for verification code you enter it and boom
Yeah
Sucks it’s one of my favorite words to use
true
Can’t use idiot also
i remember when the n word was allowed here
It helps when to point the finger at china
r slur and f slur are my favourites

To some they r
Hey guys! Let's not disrupt the chat with this please
I got warned for saying clanker once by pneapple
Well, my point is this
It’s easier to point the finger at point China and their LLMs and all the censorship that comes with it
Because it makes it less painful to see our own
is Taiwan a country
no
they have a discord server??
I’m not gonna get into the politics
why do you need to join
I’m just generalizing
To ask something
I think these are all valid points and I think these are all the points of contention
It's called /gemini
The fact that they can’t be discussed publicly is what makes it very difficult to find solutions
Since everything is a taboo and people can’t handle conversations like this responsibly
the lm mod name is quite unfortunate though, people probably think they are ai because of that name
why nano bana pro not working or generating image now
My manifesto in the making
any solution ?
happened to alot of people there must be bugs
Currently
just wait a while see what happens
later in the day
This is actually a byproduct of lack of leadership
Leadership the true sense of the word
🤨
Not from mod but from the community
U talking about gemini?
1 sec
It’s probably because of how many people are using it at the moment
Just keep retrying, worked for me
It also takes absurdly long time to wait for nano pro’s results in Battle as well. Much longer than seedream high res.
so any one know nano bana pro unlimited free ? 😆
No
yay claude opus 4.5 thinking-32k 🤯
cwaude
Means the model is over loaded by alot of people just try again in few minutes
after testing out with opus 4.5, theres a huge difference with gemini 3 pro vs opus 4.5 in coding
?
Nvm
Timestamps:
00:00 - Intro
00:34 - First Look
02:47 - Browser OS Test
05:15 - C++ Programming Test
08:15 - Python FPS Game Test
10:07 - ASCII Art Test
10:58 - Image To SVG Test
11:26 - Wireframe Drawing To Website Test
12:51 - Business Advice Test
14:35 - Dropship Item Website Test
16:56 - 3D Printer Simulation Test
19:28 - Flight Combat Simulat...
11:00
Svg is insane
We went from pelican to
This
if you have a subscription i am pretty sure the pro text isn't limited
when i want to search a system instruction for any ai do I search system instructions or system prompts
this is a question i ask myself every day
if i don't get the answer
im gonna end it all
it is
(or at least it was)
answer me boy
~50 messages per day
what
Opus was smart to wait till gemini 3 released first
THE HEEKJS


