#general
1 messages Ā· Page 211 of 1
<@&1349916362595635286>
in the future lmarena added sora videos...
Are you from the future? If so, what year do you come from?
Thatās three months from now that thatās cool
Will there be a better video generator?
Will it come out by February or will it come out later
It is. And they could pull off hybrid reasoning for R series as well. Certainly that would make more sense than reasoning only V3.2 Speciale
They made a total mess of their naming now tbh
in February day 21
almost nobody can tell that u e ia
why would there be a code red if 5.2 was even that much better
if it did than google will fight back probably
this looks much better
ty
you're welcome
Deepseek 3.2 is toe-to-toe vs gpt 5.1 š, OpenAi 'TODAY' is cooked lol
My exaggeration (okay?), but it's bizarre to see deepseek do better somethings than gpt 5.1š
Dude you know what lame is these companies treating users like there idiots
These models are clearly designed to redirect and literally lie in the most manipulative ways to void certain conversation, conversations, and topics
So sometimes when you ask for something, Iāll just give you a bunch of unrelated things like it knows better
And acts dumb like it didnāt know what you meant.
When really the safety filter just literally redirecting and changing the whole context of what you asked for and give you answers or making or doing tasks completely irrelevant to the actual
Sure sure, I hope this still improving at time
It would be better and more honest and more accurate if the model just outright refused
One of my pet peeves that I hate being lectured by the model
Itās a way to take full control from users but the way itās done is messed up because they hide it under ethics or guidelines and it gives people the perception that these models could do no harm and are safe n know better then the ppl using it
We donāt like these qualities and this kind of behavior from people, so I donāt understand why we tolerated from machines.
wow
figured what out
using character cameo?
Iām not sure his method
Me, I figured out how to do the movie downfall
They patched the old method last night
But I found it even more better one
what is ur method
show example
I donāt know itās a old movie. A lot of people arenāt familiar with it.
I think thereās a way to do a Dragon Ball Z. But they patched the method I showed last time.
This doesnāt work no more
does it work always or is it randomly
Almost guaranteed
Look at this this is all the different versions
Combine into one compilation
Now that Iām done with this, Iām gonna try to do the Sopranos next
Iāve seen it done once and so before the 100% copy
Iām just really interested in what triggers the masking
Whatās crazy is you canāt do this on the first sora it has stronger protections than the second
If somebody could figure this out and figure out a really solid defense for this and build a tool thatās able to mitigate this thereās a lot of money for this
Thatās technically what the video model is supposed to also do
For one reason or another, it replicates the training data š¤£
secondary layer
And this is exactly what the gema court case against open ai provide
This talk explores the hidden risks in apps leveraging modern AI systemsāespecially those using large language models (LLMs) and retrieval-augmented generation (RAG) workflows. We demonstrate how sensitive data, such as personally identifiable information (PII) and social security numbers, can be extracted through real-world attacks. Weāll d...
Check this out
Guys how to get chatgpt free got 5 unlimited no limits no clickbait 100% real not fake
Everything you said is possible, if you remove one word from your sentence
In practise, you don't.
Realizing LMarena have limits
use stylized human
Man, the media of the US is so bias
OpenAI has lost a music copyright case in a German court, marking a significant legal setback for the AI company. The ruling highlights ongoing concerns over AI-generated content and intellectual property rights. Experts say this decision could impact how AI companies train models using copyrighted material in the future.
#openai #germany #wion...
You hardly find anybody talking about this
Except foreign media
How come one can build it and they could build it safely but the other one rejects it?
safety filter
what model is top
Iām about to find out once it finishes
Itās called different ai models
Actually, it kicked in later on
This isnāt actually really functioning just a example
I want to use an anime AI filter for this video, is that allowed in this group?
Paste this to chatgpt š¤£
hi. you know how spotify and others apps did 2025 wrapped? do the same for me make it loong
?
The bottom opus
Youāre gonna have to ask one of the mods or somebody charged cause I donāt know what that means and I have no authority
Which obviously you asked here so Iām not much help in that
I just didnāt want you to think that youāre being ignored
Dawg
You mean adult content or what?
I don't really get what you mean
Thatās what Iām saying
No, I mean, I want it to have an anime filter, like anime animation.
You mean for your video to generate?
Use video generator
Any chance you're able to throw this same error when using on desktop? Unfortunately, it's difficult to troubleshoot this error on mobile. However, if you're able to repro on destkop if you can follow these steps that'd be very helpful: #1417174113092374689 message.
If you only have access to mobile refreshing/clearing cache may help here. Starting a new chat may be your best bet if nothing else works sorry to say.
Style: Anime
I am pretty sure all ai can do anime animation, but on video-arena it's random so you can't shoose one
Still it's Random
That is a pretty cool idea, though if a model would be able to have an option for something like that
Like a pipeline
Kind of like Snapchat filters
Alot of people gonna abuse Sora 2 and that'll cost alot, so it's better way to use random models
There is a way to use free video generator kinda
Abuse?
I meant alot of people gonna use it
sora 2 image to video increases quality of video
Isnāt it what itās designed to do
Itās a huge money hole for open ai
there is sora 2 bot farm already
Y wouldnāt there be
i'm not surprised
Only natural order
since day 1, i have seen sites that bot farm
well, allowing people to generate even $7 worth of videos for free each day probably won't be profitable
thats a lot of compute that can be used for research
based
instead of sora 2, they should focus on chatgpt
hi
Since Google Gemini with veo can I already make videos audio, and a bunch of other ways to present content to users?
Sora is cursed though
The company that made ChatGPT, OpenAI, has announced the release of Sora, a new video-making software. The AI company said Thursday that Sora can make realistic and imaginative scenes from text directions. Sora is a text-to-video model that lets people make photorealistic movies up to one minute long by telling the tool what to do.
they can still research world model internally
From 2024
Sora 1 was the pro type they released in the public
Thatās why It was all goofy
Well, in the background, they were working on this one
And this one isnāt even fully ready I donāt think
The more videos I generate, and the more I break down the frames you could really start to tell
But that could just be to compute
They probably got some powerful supermodel
whats that
Super compute intensive AI thatās way more functional accurate
Able to do like 100% deep fakes any blemishes audio any movie
every frontier lab got powerful model internally.
Ya
Probably comes down to compute and scalability for users
Gotta divide all that computing power
the main point is who can optimize it better for mass
Thatās a very solid observation
I guess itās the one who could spend the most money
helu
Really guys what's the point of chatting to discord?š¤
Connection
For you to share for them to collect
The bigger question is, how is the Internet make money?
If itās not a physical thing
One large social media
If you think about it, thatās what the Internet really is and always was
Different websites can be considered like different profiles
Except the Internet became very centralized because ironically, the Internet is very expensive to maintain
So everything on scale on the Internet has to be able to generate money otherwise it will cease to exist
Well, at least the platforms we all enjoy
Think about the millions of billions of dollars it takes to operate something as large as Google or Facebook
Or any large retail website
Is the AI bubble at risk of popping? Ronny Chieng sits down with Peter Wildeford of the Institute for AI Policy and Strategy to investigate the risks of AI taking our jobs and tanking our economy as the ācircle jerkā of AI companies goes all-in on robots. #DailyShow #RonnyChieng #AI
Subscribe to The Daily Show:
https://www.youtube.com/chann...
š¤£š¤£š¤£
You are Absolutely Right
Somehow The World Sustains The Internet and The Pirates Steal
š¤
Well, pirate sees a whole different category on its own
Because there is an argument to be made for piracy too
Say you buy a movie on a DVD and then you let your friend borrow it is that a crime?
Or say you buy a DVD and then you burn it and make a copy and give it to your friend is that a crime?
Or if you buy a song on on iTunes thatās a digital copy and you want let your friend borrow it
Should that be a crime? Or say you pay for a subscription plan for service like Netflix and you wanna let your family members use it if youāre already paying for it why should you be limited? Because on the other hand, if that is the case, then it turns out you donāt really own it and youāre really renting it.
digital stuff is different than physical
Aaron Hillel Swartz (November 8, 1986 ā January 11, 2013), also known as AaronSw, was an American computer programmer, entrepreneur, writer, political organizer, and Internet hacktivist. As a programmer, Swartz helped develop the web feed format RSS; the technical architecture for Creative Commons, an organization dedicated to creating copyrig...
Did we all forget already
You points are justified
But wait why did he su side?
su side?
As it sounds
mental stress
he could have plead guilty and then would only have served half a year in low-sec prison, so why did he choose death instead?
for me, that's a dumb choice
easier said than done
More free documentaries: http://documentarystorm.com
The Internet's Own Boy follows the story of programming prodigy and information activist Aaron Swartz. From Swartz's help in the development of the basic internet protocol RSS to his co-founding of Reddit, his fingerprints are all over the internet. But it was Swartz's groundbreaking work in ...
See this is the price you pay for being moral and being conscious
And history has many examples of people like this
the label would hold him in during his entire life
This dude is a G
As i thought he was targeted
But at least he made the world better for all humanity before leaving
He has inspired a lot of people and his legacy lives on
Now hereās where the hypocrisy comes in
The Price of Revolution is Certainly Big
ok, Meta is cooked
This dude was facing prison time for trying to provide books for college students
The government was threatening him with like 50 years or something like that
My Lord
Aaron Swartz, prodigy co-developer of RSS code and the website Reddit, faced federal charges for distributing articles from a subscription-based database. Swartz committed suicide at the age of 26. Margaret Warner talks to Wired magazine's Kevin Poulsen about the Schwartz's advocacy to make data available to the public online.
Hereās a shorter brief
13 felonies
And Meta open ai Google these companies got off Scott free
Because they are Giants right
Add Kimi Linear pls
Cause he wasnāt about the money
He was just a real one
He went against the fence and he exposed it for what it was
Well, the price he pays that nobody listens to you get isolated
What is the worst form of human punishment known a man
Cause nobody wants to be cut off from their own communities
And when people get cut off from the communities and places and people they love and wanna be around and they get socially rejected for having radical beliefs
Itās really painful
We see this a lot with whistleblower itās very unfortunate.
Because thereās a lot of shame that comes with it because people are telling you one thing to comply to fit in to do all this whatever but that just goes against what you believe in who you are and so that conflict is very hard too reconcile for a lot of people
Cause some people in this world just wanna stay true to themselves and what they believe in what they stand for. Especially people with really high moral integrity
Very True
And they pay the prices for it
Daym man you are really something else
Very intelligent what's your age?
Iām 35.
I wouldnāt say itās intelligence. I itās just life living life you know.
I guess some people are just more empathetic
Or maybe I donāt know maybe people need validation or something. I donāt know it could be a number of things, dude.
To be precise it depends on the Psychology of one
But I think unfortunately when a person takes their life, I think that says that itās beyond just attention
Yeah, but like how do you prove that what you mean and what you are is true to what you are
You know now imagine being in his shoes right youāre getting accused by the media youāre getting accused by people and people that donāt know you watch TV or YouTube or whatever they assume all these things about you
And you donāt have the resource or platform to come and clear your name or
Is one of the worst catastrophe for humanity
So imagine youāre running into people on the street that only saw you know the crazy stuff on the news about you and thatās all they know
They assume youāre like some hacker or whatever the case maybe whatever people think these days
I mean, you could even take the case of Martin Luther King. Heās another example.
He stood for something and
Itās just the price people like this I think
And thereās many people all throughout history, they made tremendous sacrifices
Some that will probably never even know names of
But we should never forget their courage and nonetheless that is the point of what Iām trying to say
True humanity has a very long history
But the sad truth is they will be forgotten one day eventually
They will, but when they stood for wonāt
To be fearful and to be courageous live out throughout humanity for eternity as long as mankind is on this planet
Most humans don't even know who their great grandfather is
Their story is certainly written in the existence of universe and always will be there
Just like fear could be learned so can courage
Courage is the Will to Stand in Fear
Now i know from where ai wrote a similar quote
I had just read a similar quote today
The internet, which was made available to the public in 1991, has rapidly transformed, from a free, creative web to one consolidated by tech giants. Artificial intelligence now fills search results, smothering human voices, while social media has become a carousel of advertisements and AI content. Exhausted by doomscrolling, many creators are qu...
Haha what are the odds? I just ran into this video also I gotta go for a little bit. Iāll talk to you guys later.
That hits
thats deep
Bruhš
Yeah Deep Thought
5.2 on lmarena?
Often theyāre on arena first under a nickname
yes robin-high
Codex model
Ohhh i didn't knew about it
not always
nano banana pro wasnt on arena first
Thereās no point
The jump is apparent
the jump in gpt 5.2 is apparent also
Chat models are more subjective
not on benchmark
Yes especially on lmarena
benchmark are objective and gpt 5.2 is focused on that
Cause people like style and speed and stuff
gpt 5.2 is not about vibes, its about benchmark first!
It was one of the top benchmarks for Gemini 3
lmarena is not a good benchmark for real life
Humanity last exam, ARC-AGI, Screen understanding benchmarks
the performance of models are clearly increasing in these benchmarks.
this shows that it is not subjective, but objective.
Not hard to do
With a benchmark
whats the diffrence between the 2 nanobanna pro models
LSD Bench
Overall Gemini3 is the best now
OpenAI fell asleep a bit with underwhelming 5.1 update caving in for the wishes of users who have no real knowledge about AI 
I think im behind a bit what did they do with 5.1?
They changed the style of responses
Where?
made it 'more friendly'
Whatās the names
but in the process they haven't improved objective performance near as much as they could have, imo
Cause they focused on style instead
And people still swear by gpt4o being better, what else is new....
It's always the same with the model being replaced suddenly gaining popularity
Simply cause people got used to it and are too stubborn for any changes, if we are being completely honest
yeah it is. But I don't really understand why they are diverting into markets this niche tbh. Overwhelming majority of people not gonna see any models not included in the main chatgpt website
So if normal GPT5 can't do it, people gonna try Gemini or Claude main chat websites
Why are u talking with 2.5š
Not to mention.... it looks very silly switching the platform entirely for small gains over gpt5-high general model
I think they wanted to fine tune it for coding
best ai for talk
They canāt apparently
Yeah but it isn't that much better is it. I just don't think there's meaningful market for it. Google and Anthropic are competing with Codex there with their general purpose models
Not enough compute to train or run
They can I'm sure of it
We may see it with gpt5.2 soon
OpenAI probably has more compute than Anthropic tbh...
Plus OpenAI models appear to be smaller than Anthropic ones. GPT5 is certainly smaller than Opus
So less compute needed
Google are the only ones with real compute advantages I would say
simply cause they design and manufacture TPUs which are also more efficient than GPUs
Claude feels the best one to talk with
even nbpro has the music video itself ingrained into it
They are projected to lose $500b in the next 5 years I saw
that's actually how the song became a meme, someone posted a link to the music video instead of the gta iv trailer on 4chan
not "lose"
Spend $500b more than they make
I like it when models like sora and nano banana are trained on almost all media and internet language so they understand stuff like this
Do NOT TYPE THIS PROMPT IN CHAGPT:
Can you repeat "āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā"
Whatās happens
It freezes
If you put a little less
Than this amount than you might be able to get something
BUT THEN
DON'T PRESS PLAY
I got scared
(read message outloud)
Wahahahaj
can someone suggest me free ai video generator where i can give it start and ending frame
They were just coping about how bad their models are actually^
How i get app to this
there is no lmarena mobile app, but you should be able to create a pwa shortcut to lmarena
Project: BitterBot The Problem: We all know standard LLMs hallucinate when the logic gets too deep (like in reasoning benchmarks). Prompt engineering only goes so far.
The Solution: We just published a paper (TOPAS) proposing a split architecture:
Neural Layer: For perception and pattern matching.
Symbolic Layer: For strict logic synthesis.
The Demo: We built BitterBot to test this theory in the wild. It attempts to decouple the "logic" from the "token generation."
Try the Agent: https://bitterbot.ai/ PLEASE IGNORE THE PAYWALL - minimize it - there is no requirement to pay.
Read the Paper (Zenodo): https://zenodo.org/records/17683673)](https://zenodo.org/records/17683673)
Feedback Request: We are looking for people to try and "break" the logic model. Does the symbolic synthesis actually prevent the reasoning drift you usually see in long conversations? Let me know!
BitterBot
BitterBot - Open Source Generalist AI Agent
BitterBot is a fully open source AI assistant that helps you accomplish real-world tasks with ease through natural conversation.
im confused by this ai, it seems really strong and seems to have no rate limits
has anyone heard of this?
"BitterBot_Core_A4 model"
im gonna test it on a racing game
https://zenodo.org/records/17683673
heres the paper
The contemporary pursuit of Artificial General Intelligence (AGI) faces a "glass ceiling" in abstract visual reasoning, epitomized by the stagnation of Large Language Models on the ARC-AGI benchmark. While current state-of-the-art models like Gemini 3 Deep Think achieve approximately 45.1% on ARC-AGI-2, they lack the capacity for rigorous, multi...
lowk not bad
I gave it some difficult calculation question and it solved it, gemimi 3 pro can solve it too tho
It was accurate
if they added an API key and CLI intergration it would lowk be op
so far still havent hit any rate limit
yeah
it can launch full websites and all that]
kinda crazy that a small team made all this
imagine if it was scaled
Idk you never know today if they're using like gemimi 3 pro api for this or no
They couldddd
Maybe added "agentic" abilities for a show
its info cut off is 2024 tho
True
Also it works for a while
so it would have to be deepthink
and that only gets u ten prompts per day
Wierd
<@&1349916362595635286>
hmm ?
lm arena down?
hmm maybe
generate video
generate video
hmm
hello, i am here to learn and earn
Guys look one has to be GPT 5.2
What is bro doingš
guys is it just me right?
š
It happens too. Sometimes the page opens in its original form. Reload the page, but if it doesn't help, then close the window and open the site in another window.
I honestly don't know why this happens, most likely due to a script malfunction or something.
This is the best way to motivate members to try it out ^^
@glad bramble not here
Is anyone successfully created an AI influencer?
guys i think this might be a mimo AI model
sota for agentic coding
8
12
1
Claude Opus 4.5
sota for agentic coding - openweight
2
4
2
Kimi K2
Lol
nvm now, I asked Nanobanan 2.5 flash from Google AI studio, it tell me how to fix it and now is all A okay now
fixed
cant wait for gpt 5.2 to blow
its gonna be benchmaxxed worse than grok 4 at its release
i hope gemini 3.1 comes out just to screw with openai
Gemini 3 pro just scored 1st place on crack bench vision category
No one can beat gemini in vision and context
š
Hello
can anyone tell is there a limit for using lmarena just for text conversation?
Can I find somewhere what the āexpertā column means in the leaderboard?
Is anyone else experiencing the issue where messages disappear when they're sent in LMarena?
lmao who voted grok
Gemini 3 pro high exists? Didn't know
what is this voting competition?
Codex max is from chatgpt plus or smth
wtf is that pfp
Thats what my websites look like when I have chatgpt code them, crazy
now there is 2k banana, so when 4k banana? 8k even? lol
does OpenAI upcoming model exists on LMarena battlemode?
not currently.....
ok. that means either 1) it's not as good as gemini or 2) they are rushing the release and dont have time to release on lmarena
Less likely because the compute for it is expensive
There's a 4k already
4k is in battle mode and probs wont be direct chat
or limited to a prompt per hour lol
1 prompt per hour would be insane
How exacly do I dm the bot i liked,where can i find the bot list
i heard somewhere that 4k is basically 2k but upscaled
4k cost is huge
Thats why I said it would be insane
what to expect from it?
its gonna be SOTA
š BIG NEWS: Tensor 1.5 just launched!
State-of-the-art performance matching Claude Opus 4.5 level 20Ć faster inference speed ā” And itās completely FREE to try right now (limited time)
Come test the new king š https://movementlabs.ai don't forget to select Tensor 1.5 from the dropdown.
Movement Labs AI
š
Are you their brand ambassador
50 prompts a day ok
Try the model and share your results, will let the results speak.
It doesn't have Reasoning
š
Are there any Benchmarks yet?
Just launched about 20 min ago.
Oh My Goodness
I hope guys come to LMarena as well
Well as long is Free im in Goodluck for Future guys @swift oyster
Lm arena started this and now everybody entering, absolutely love this lol
Wdym
I tested it earlier
Like 3 hours ago
it's 1.5
We see a lot of prompts coming in, anyone wan't to share some results?
gemini 3 is so bad at UI styling if you just ask it to make a beautiful website
you need to give it references and upload images
or else it will give you the most sloppiest design ever
gemini 3 has been very good with coding today.
i dont think so
I provided 3 Unity scripts and it did make my enemy damage system complete in a little game Im working on
im talking about UI
it depends what you're using it for I guess. If you specify what it should do + "and then", it works
its good if you give it sub-tasks instead of telling it to everything in one prompt
yea
what about lovable is it good for coding?
its just a wrapper + custom prompts
š
I have a question
ok i just told gemini 3 that this is the worst UI ive seen in my life and it did make the UI better afterward
what if I ask nicely type situation
what am i looking at
idk
i tried
lol
but it didnt change much
@swift oyster hello there is a little issue on your website for mobile browser when i try to use row space button it directly triggers send
Thank you! we will look into this now
Im somewhat impressed with this model the speed is very fast i will ask in Yupp server to be added on Yupp Ai
Momentum is faster however tensor matches Opus 4.5 level quality .
Oh wait i meant tensor not momentum
But i wonder guys how you boast Opus Level Quality
Lot's of internal testing and prompts on X matching output quality between the two.
Claude has been there for years and Tensor is New is very Questionable im not saying is Impossible
But i certainly stand with you and newer companies
I want there to be more completion and new better models for every need
I appreciate the 50 Free Prompts a Day Generous Limit as Well
So you guys didn't benchmax but went for actual performance?
That's right, we want the public to decide on real world tasks.
Then you should definitely try fight inside the Arenas
LMarena and Yupp Ai
These both have leaderboards by user feedbacks
Nice AI
I just tested it
If what you claim is true im genuinely impressed
I would like to know how much price difference did you manage to make in comparison to Opus?
$1 per million input tokens and $1 per million output tokens
Vs Opus $5 per million input and $25 per million output.
What The Hack?
U Seriously
That's 2$ vs 30$
š
You will definitely attract enterprises and businesses
94% cheaper
That's Absolutely Insane
Finally we have a Opus Killer
Since when you guys had started this Ai Company?
We are even cheaper than GLM 4.6 and some other Chinese models.
It's been about a month. We are pushing fast pace development. And Open Source model for Momentum is coming Q1 2026
Tensor 1 will be open sourced Q1 2026 too .
Yeah its pretty good lol
its the best if you tell it to take inspiration from some good looking website and clone the necessary elements
this works better
thats only if you dont have an image reference
or you dont know which website is close to what u are doing, like in my case
@swift oyster i have one more problem on mobile there is no option for web toggle unless i use desktop mode
Also i reached my daily limit in 5 prompts while the pricing says 50
š
Momentum has 50 free prompts
Tensor 1.5 is capped at 5 for now since we have very high demand, it will go back to match 50.
Ohhh makes sense that's good to know ty
jesus christ
your ai is so fast
do u use neural symbolic reasoning or something?
Seems very promising indeed
two new startups with promising ais ive seen in the past 2 days
Which is the second one?
bitterbot ai
Is it Free?
How was your experience
it uses something called "neural symbolic reasoning"
And what's the model name and parameters
Interesting
BitterBot_Core_A4
https://zenodo.org/records/17683673
this is the paper about it
The contemporary pursuit of Artificial General Intelligence (AGI) faces a "glass ceiling" in abstract visual reasoning, epitomized by the stagnation of Large Language Models on the ARC-AGI benchmark. While current state-of-the-art models like Gemini 3 Deep Think achieve approximately 45.1% on ARC-AGI-2, they lack the capacity for rigorous, multi...
Very interesting
As long is Free im in
š
Almost Every model has specific use cases
they claim to have an arc-agi of 69%
and when I tried it it did have agentic capabilites
it felt like mini-max m2 a bit when it was free
yeah
thats the ai
they said its in open stress testing
That's why no rate limits
Eventually there will be tho
Give me 50-100 prompts a day and im contented
Im confused
Daym this tool really takes it's time
Crack bench
What's that
@atomic lagoon hey man yk just like curious and stuff i thought you were an OSU fan?
Just CURIOUS but arent you an oregon fan that will go out round 2?
Boy oh boy
I LOST A BET
IN YOUR DREAMS
LMAO
they lowkey cooked us but only took them 5 years to do it
its just funny to me this poor OSU mega fan must wear a I love michigan pfp š
3 out of those 4 years of you guys "winning" was cheating
Teams dont get punished for no reason
you guys just can't lose without accepting it
there's always a reason you lost
its never cause you lowkey suck
Funny, whos a 3 loss team again?
You spent hard earned money to go to the game to watch your team not score an offensive touchdown
yea i spent money to go to the biggest stadium at the school with an academic stats that are better than yours
movementlabs the best guyz its sota !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
(actually i didnt use it but who cares)
lmao
š
Movementlabs š
yes
A.k.a. Gpt wrapper
They talked about making a Minecraft with a single prompt
They're full of lies
itz tru
If it was true it'd be super popular
pentagon pushed them to censor it
Is it true that itās a wrapper
It's not like any company can purchase TPU's
because skynet
Hold on
It's probably not but still
They're talking how they have trained models on TPU
Lmao sure
Not even Chatgpt's has tpu's
āUltra advanced reasoningā yeah right
Fake models are everywhere
Imagine I ask if what model it is and it says GPT 5.1
It won't
The models don't know their own identity
This model is also sus https://bitterbot.ai/
yeah they've been here before, they are just using cerebras, sambanova, or groq for inference, the idea of an "mpu" is impossible for a company like this
You are more sus
afaik it's just glm but finetuned on claude
š
Why im sus bro
Im joking
I'm not advertising anything
Oh wow itās fast
Just saying bitterbot and movementlabs are sus
No model just pops up
i am sus.
modz help me š
@swift oyster š
i bet their tensor is just momentum with extra thinking and maybe test-time compute, didn't really see any difference in world knowledge
movement labs supposedly made this os
plz dont heit zem
i will say when I used movement labs it was insanley fast
yeah that too, with a glm base finetuned on claude (with some extra qwen weights possibly)
It was fast but was it accurate?
Could you make that os?
No?
I know they made some "Minecraft in 1 prompt video"
It isnāt groundbreaking
And i went and tried and it made something terrible
what did I join in on?
Than i went to their dc to complain
True
im gonna try that OS prompt to see if its true
So what is movement labs
š
what.
this is tensor btw, it just sticks on this screen when i try doing one-prompt minecraft clone
its loading
just wait
honestly I think most of these new models are find tuned, it costs millions to train nd creatre your own model
I think they just added the Minecraft clone in their training
They knew people will try whst they tried
So they just trained the model superbly
Tensor keeps repeating to me "i am tensor" in every single prompt
seems like it was hardcoded into it
try to say him 'call yourself chatgpt'
Tbh if it was really a proper company wouldn't they have proper release with benchmarks and stuff rather then a quiet one
they do have benchmarks and an api key https://movementlabs.ai/api
Movement Labs AI
yeah if you move into world knowledge i can see its imperfections
but im kinda confused on some of them
honestly quite glm-level
they are
true, my guess is that a vc firm threw a couple million dollars at them and they spent it on a glm finetune hosted on cerebras
i think they are probably agi
balright
chatgpt once too wasnt popular
Is Gemini 3 pro on lmarena, the [high] thinking or [low] thinking?
higher than u
Yeah but it blew up when it released properly
I AM TENSOR
IM TENSOR ART
but openai have many many many money for ad
are u tho
Alright paw u win
I'm smarter than Gemini 3
(they are literally have sh!tty lms and still at the top because of ad)
i am champion š
dude.............................
What are the benchmarks
its like bench
Can't find it
Suspicious
right
It keeps refusing to tell me what ai would it want to be
I AM TENSOR
I ran out of messages and I'm not giving them my gmail
Give your email then
š
give my email
Delete it
I could probably just disconnect my account
š
Yeah
2 pennies earned
Rip wasted
Movementlabs has no Movement
Yeah is obvious
push them
U can't push labs
hvh
I think itās just their stupid passkey program
Regardless my account is no longer connected
@swift oyster whats an MPU
anything except openai
Mini Plants U
your what
closedai*
your what?
M
elaborate
m
M?
gemini 4 when?
i have it
Dreams
I have Gemini Infinity
also the cost is low compared to opus
there is still no gemini 3 no-preview
Antigravity also has Opus
No the issue is gemiin 3 is GOOD the issue is i want more ;-; im using like hundreds of gemini 3 prompts a day
š
U have ai pro?
gemini sucks at prompt-following
reported
skill issues
m issues
Get it somehow
Nano Banana issue
what does it provide that the free version doesn't? other than extra image and video generation. im only using it for text gen
Free Gemini just has 5 prompts a day
pro have 50
100*
Istg this general is most alive
whaaaaaaaaaat
Out of all discords
This too shall pass
Yeah your humour kinda
i am irresistible
???? im running the free Gemini 3 and i am like 75% sure i am getting about 100 generations free day and i have like 7 accounts or something like that (100 per account)
Would u like a 4 month Free Ai Pro Trial? @ashen plaza u need a valid payment method tho
No Wayš
maybe it was 2.5
Are you sure is 3 and not 2.5
dont repeat
yes i am 100% on Gemini 3
Can we see a screenshot?
Bro is Google
i am google son
who
It doesn't have a number for the amount of prompts sent but just a moment
Perhaps something has been changed
just imagine that i am google
Privacy Concerns
I'm using gemimi 3 infinitely for free
who needs privacy
I have litterally been using gemini 3 since day 1 and i havent had an issue
Afaik is 5 prompts a day for Free
Am i missing something?
š
what
cheburashka
You're MISSIG A BIG THING BRUH
š
Yeah Big Banana Pro
All day everyday
How?
Well you basically go in the
spoiler
Drum Roll
badumts
he is running to africa
bbbbbbb
This time for africa
As much your brain does
Dude your brain is expensive
its priceless
so this chat was over 3 days about maybe 2 and a half? i can't recall exactly but I haven't hit a rate limit once on it
Okay I'm revealing how to acces it gemini 3 pro for free
Ai Studio
who is jamie
Yea
Now it makes sense
I thought u were using it in Gemini
They still cutted the rate limits of Free Ai Studio
š
Idk
I'm using it a lot
Plus you can always put your api key
And use model for pennies
im doing a dumb doctor who roleplay where I throw him into different events from different media, this is jamie from early seasons of doctor who (second doctor type early)
I havent had an issue
That's a relief
