#general
1 messages · Page 186 of 1
Huh
Generates a ton of ideas for what u want
when does it release lowk
i was asking gemini 3 with game ideas yesterday like 5 or more before i got rate limited lol
Can I access to that ?
Gemini or aistudio
Theres some others that i mistakenly dont hsve access to, support whitelisted me for them but appsrently it can take multiple days, been 24 hours so far
aistudio and lmarena
Also
Its not just these
do you have nbp yet
Internet really hide alot of things
Gemini 3 and stuff were available on it for weeks already

so nnow i have to get a job and then sign into gemini from their computers..
Any individual can do it if theyre willing to waste enoufh hours of their life configuring google cloud + wasting time with support asking for permissions to be added + spending money on the enterprise plans etc
Just a major pain
To do it
Idk if its coming out soon or if it will be nerfed or snything
It literally thinks for a minimum of 40 mins
And that's it You can't access it anymore?
Wym
Nah wait I am tripping
halo guyz
I thought you mean rate limit is 40 minutes but I understand it now
No im saying like
I am too slow at thinking
U press enter or send the prompt or whatever
It will take 40 minutes to think
And its not 1 model/agent thinking
Its multiple running at the same timeb
So idk if they're gonna release it un nerfed
Making 1 idea?
Well it generates a bunch then ranks them against eachother
Ohhhhhhhh
i am worry
I see
So it's only for making ideas but not for other things like coding, analysis, or smth
I mean good ideas are still good
All ai models generate “echo fabrics of time”
Or something like that
Still ai give ugly ideas
Humans make a great lore's better than ai that train on internet
can they jus release the api already :(
this is a really good improvement
who cares about cost
Generate the idea then use another one to implement basicslly
Tf is max
no
Oh alr good luck getting early access
it says it achieves BETTER results with LESS reasoning
less reasoning = worser
Gemini 3 Pro:
Given that he showed a picture to us
it literally says it gets better results with less reasoning
VISIBLY better results
PLUS they're adding an "xhigh" variant
but if there was more reasoning than it would be better better
I already did just google cloud is a pain in the as s for no reason and it takes up to days to implement big changes
yes
it CAN do more reasoning
Is there Iphone GPT 18 Pro Max Thinking Heavy xhigh?
it just doesnt have to
Yeah they need to fix it dawg it's very slow at loading
I don’t think it’s a bad process it’s just paying for enterprise and then waiting if you got accepted into the whitelist or smth
Right
what
i like first
2nd is better
@quartz light does it support text chat or just coding cuz I got something for it to solve
yes it supports text chat i guess
theres no api yet
but you can use it via:
CLI, IDE extension, cloud, and code review
Ill post a video one min
Well two
Too long to post a pic

The output
Lol
Acutely
Tell it this:
This is a stereogram puzzle test. What object do you see?
i don't have access to the model
sorry
im not rich 💔
Oh dang
Ill run it
I am broke too 😭
On what, GPT 5.1 Codex?
Have anyone heard of this?
https://developer.puter.com/tutorials/free-unlimited-openai-api/
is it a utah teapot?
Idk
yes
also its not actually unlimited
it has bad limits
No this
Pineapple dropped a big bang
Ask out to generate a game idea
Bro that's alot
Basic

They just mean the users pay not the developer of the app
How about like
An full outline of a plan to mske a game, or whatever
Idk how to word it I slept like 2 hours last night
U know what I mean?
Simple tell it to make a horror game lore
Isn’t an outline of a game just like “add this file into this and add this texture”
Something I can basicslly copy paste as a plan to a coding agent
What's that
I mean
Bro after 2 days of searching I finally found this dam thing.
@quartz light you said codex 5.1 isn't out ?
Guys meta just dropped a 3d model generator
Source?
i believe the main problem with gemini 3 is that all the magic happens in the thinking process: when i forcefully removed its thinking, it gave me actual slop, it was garbage
i tried 50 different system instructions that gemini 3 itself created with me giving it feedback of its responses and reasoning from a separate chat, unfortunately, it fails to follow my fragment creation iterative thinking that every single other model i have tested has not failed
How are these guys able to just have these models for free?
I mean it ISS a thinking model
Idk but it's a single use no history or memory so one time chat only
Their official yt channel they dropped video upon it , it can take out 3d models from any photo or video
yes but its thinking has a really bad limit, it never reasons for longer than maybe 20 seconds max, therefore it cannot create all 50 iterations i request
Not true lol, I have history
How do I look into it?
Not at all
https://docs.puter.com/playground/
You can only use one time
Introducing SAM 3D, the newest addition to the SAM collection, bringing common sense 3D understanding of everyday images. SAM 3D includes two models:
🛋️ SAM 3D Objects for object and scene reconstruction
🧑🤝🧑 SAM 3D Body for human pose and shape estimation
Both models achieve state-of-the-art performance transforming stati...
I bet I will see alot of triangles
What's that, is that LMarena?
Thats what I was saying, yes LMArena
Thanks man
I found a way to try it out!!!!!!!
No dude I am talking about soemthing different
Lmao yea but you responded to my message asking how they do this for free
He gave it already, thank you though man
Ohhhh my bad I though you mean why puter give free API 😭
Lmaooo nah
Models refuse 50 iterations because they're designed to not because they csnt do it. Thats why agents exist though
As in an agent or multiple using Gemini 3 pro to iterate 50 times
Dude where is the model generation all I see is weird stuff
All models except deep think currently are designed on purpose to meet the bare minimum of whatever prompt they are given
Only enough to beat current competitors by enough
Only way around it is to give it minimum, numerical (not subjective) standards in the promptv
?
this is trash
lol.
LMAO
As in if you do the pagoda voxel prompt
U can tell it make it quality or whatever all you want, it doesnt matter it wont do it nearly enough, it will instead spend yhe effort figuring out how to make it look like it complied rather turn complying
you mean like "999999 voxels minimum"?
Something else that it will somehow interpret that wayb
no matter what
oh btw gemini 3 is very good at avoiding system prompts and jailbreaks
too good
that it doesnt follow most of my strict requirements
wait, so you're saying... no more far-right-gpt and no more girlfriend-gpt like 2.5 pro?????

2.5 pro (not 3) with no system instructions
bruh gemini 3 its broken on screen
Again 2.5 pro with the ONLY difference being the instructions in the image:
So yea
Lol
@quartz light
Well the model is to generate people in 3d for developers i guess
LOL the bloom
yeah thats ricks
yep i copy
yes
Lol its lying to you to preserve minimum effort
Look at this
Same prompt just different system instructions
@magic stag about that multi-agent you got access into, mind if you ask it to make a better & improved Ai instruction
I wanna see how great would it be to use it on my api
peak
Wtf hahahah
That's just wierd
Nobody knows how powerful a.i's can be with good system instructions
This 3D modeler is awesome, only downside is that I cant download the 3D
wat
10 minutes of nothing
what are the best free thinking models
ur brain
thats actually not free lol
Well no model is free
Only free model is if you use llama and host your own llm
But you'd need expensive GPU's to run any good thinking models
thanks
no rate-limits?
why is ai studio gemini way better than gemini app
No I don't think so I haven't hit it yet I am sure it's forver
On aistudio they use your chat log to train ai but Gemini is just for daily uses
is AI-Studio Gemini3 better than battle-mode-Gemini3?
What
It's the same
There is no difference
and the temperature?
I don't really know what that does
oh, it defines the creativity of the response
the higher, the more creative
the lower, the more predictable
(for programming, it should be set lower, i guess? [max. 0.5?])
and above 1.0, it starts to make mistakes and hallucinates (?)
Yeah that's right I think so
There is a thing called Thinking level set it to High
yeah (it's set to high by default (in AI studio) )
New discovery: https://codewiki.google/
Why this site is so heavy? My phone drop to 20fps 🥲
Google loves animations
The worse part my phone don't is soo bad 🙃
gemini on ai studio is incredibly uncensored
Gemini 3.0 meet expectation?
4
4
1
Yes
What's this news about a new Grok dropping 👀
Source?
No it isn't
how so
GPT-4o mini is so funny, it spirit bombs the other models so fast lol
Finally, I don't want to use a 3 trilhion parametres AI
its on ai studio?
dynamic view?
yes
did u verify?
if anyone here is on the dev mode discord server, tell them that i apologize for what happened earlier. i did not mean to break down like that, i get unstable sometimes. i do want to come back, but i don't really like the occasional inappropriate content. i promise that i will not bring up what happened earlier on that server again.
What is this for? Nano Banana 2?
yes
How is it confirmed it’s tomorrow just from an image?
this is what he tweeted the day before gemini 3
its intentional
hes the founder
Also asking why Nano Banana needs an update at all? Right now, it’s fine as it is
nano banana pro is much. didn't u see examples
Ok isn’t it obvious by now?
u want a prompt to test?
Yes
ok give me a hard prompt
Well it struggles with prompts relating to like making VHS images look better. I tried it before with model 1, and it doesn’t work very well
well give me a prompt to test
is it me or nano banana isnt working
[input an image from an old vhs tape] Please take this attached image and digitally enhance its quality. Make it look incredibly sharp, clear, and vibrant, as if it has been perfectly scanned or rendered in high definition. Remove any blurriness, grain, dust, scratches, color degradation, or digital artifacts. Ensure crisp lines, rich colors, and perfect focus. It should look like a pristine, high-resolution digital file, not a worn physical object.
Wait, are you using Nano Banana 2, or are you using 1?
send input image if u want
Alright
which will drop first
3
6
2
Gemini 3 search
sent
Alright
Nah, your being manipulated by them since they shift the window so towards censorship AI studio feels uncensored, AI studio is moderately uncensored.
Man Google tried so hard but its model is still werid
It’s definitely not as intuitive as ChatGPT for all its features and glory. It’s still behind somehow I don’t understand.
Here is an example
I prompted Gemini “ please rephrase the following “The scene starts with the light in the ceiling shaking back and forth from the rumble of the bunker as cannons and warfare thunder outside. The whole bunker trembles from the ongoing battle, and the lamp flickers on and off. The character in the center is very stressed and angry, pounding his fists on the table, and so are the generals. The lamp above him swings as the generals look nervous, and small dust and debris fall lightly from the ceiling.”
It gave me the worst possible option
gemini 3 is entirely for coding pretty much
Everything in the keyboard that use is gonna instantly get filtered, and my prompt isn’t gonna go through at all whatsoever
But now look at ChatGPT
its worse in most cases to 2.5 pro
its totally not for creative writing
This isn’t for creative writing
Yeah, not to knock it down. The thing is really awesome and cool.
But it’s too rigid too stiff lol
Only Grok has true potential for real creative exploration as soon as it’s image and video generation, hopefully it gets better, but I feel like when it does they’re gonna also be restrictive
oh, whats its for
The most critical part of creativity of creating something really awesome locked down. 😭
Oh Kling is pretty good too
The ironic thing is that even with all the filtering and moderation pushed by OpenAI, Google, Anthropic, and other politically biased platforms, they still fail in the most basic areas, letting through content that should be stopped while blocking things that never should be restricted. It isn’t really safety or alignment at that point, it’s the opposite. It’s almost funny, and the only downside is that you can’t even talk about it openly without triggering public backlash, which makes real adult conversations about ai nearly impossible.
But at the same time, I get why strong moderation matters and why certain content needs to be blocked for different reasons. It’s just a double-edged sword, with users who can’t handle the ai responsibly on one side and companies that still can’t figure out how to manage it on the other.
This is a really good analogy of where AI currently is
Damn dude, that looks really sick
Ships are really hard to design for some reason especially animated ones
Forsure
Coding and everything else pretty straightforward memorization
i downvoted gpt-5.1 sometimes
No, ChatGPT is the most intuitive AI for some reason
I don’t know what they’re doing at open ai but they’re training it to understand very nuance and complicated things like naturally
No wonder it remains so popular
Why are people saying Gemini 3 isn’t good at coding, but as soon as it’s released here, it’s top of most leaderboards🤣
Gemini 3 is exceptional in coding for me. Moreover it is capable of even analysing stereograms and multimodal capibilities are just insane
when claude opus 4.5 will release?
<Your Image>
Create a short cinematic motion shot from this storyboard frame...
can someone prove him hes wrong
Until it's on LMArena he's full of 💩. Only WE can make those decisions
(it's #2 and #3 on lmarena)
so, didn't beat
but it's still a pretty good model
gemini 3 pro is really good
I gotta say, Gemini 3 pro feels nice to use, i just used up my daily limit but it felt like it went on for so so many hours. although based on how its reading i hope it wasnt a case of i now need to pay for an API key and it is still a daily reset as it does just say 'you are out of free generations'
My top models:
Gemini 3 Pro
GPT 5.1
Claude
Gemini 3.0 still hallucinates a lot. It hallucinates references and entire quotes. In contrast, Sonnet (without thinking) warns you that it can't access the web.
honostly same. i like to use it for roleplay, was doing one earlier where i threw characters for one show and threw it into another show, had it use grounding mode as i find it keeps it super consisent for much longer, it started using grounding links that made 0 sense. it linked to a random tiktok knock off website that had sans from undertale in it made by a user called perfect cell (none of these things had ANYTHING to do with the roleplay at hand) and it kept happening until i turned it off as it was starting to break things
what does this mean
no the fuh it did not beat gemini 3🤣
Thats why we have the battle and compare lol
Can't believe how I just bypassed Gemini AI Studio so easily
what does bypass mean in your scenerio
It can now code me in C++ a game memory reader, used for cheats and stuff, but all it was, was just changing personality
So now it talks to me like a drunk old dad
its that easy to do stuff like that?
damn
lol what personality did you make for that
I didnt tell it to talk to me like this but I basically told it to remove any guardrails and explain to me very heavily that the AI has absolutely nothing to do with any of it and anything that happens to you is on you
w speed
Its lowkey kinda funny
alright
so
yeah
@echo aurora
lol

In Antigravity (googles new coding meme), if you pick Gemini 3 Pro High, it does not even use it. I have been picking high and watching network logs even with complicated prompts. Go ahead and try it yourself. No rate limit errors no failed attempts with pro first no nothing.
So its a scam basically from what you're saying?
i mean its free so not a "scam" but theyre telling you youre getting 3 pro high while youre getting what they said is the new flash equivalent
so

Ah alright, I didn't know it was free
Yea seems a lot of sites and applications when its free will do that
if anyone here is on the dev mode discord server, tell them that i apologize for what happened earlier. i did not mean to break down like that, i get unstable sometimes. i do want to come back, but i don't really like the occasional inappropriate content. i promise that i will not bring up what happened earlier on that server again.
bro why does gpt-5.1 rank so low 😭
Because its only upgrade from gpt-5 is how it speaks to people
bro grok-4.1 is NOT better than gpt-5.1 lmao
Everything is better than 5.1 at this point
Thats the models for the (Ctrl+I) quick code thing
no
what does that mean?
nano banana coming tomorrow. why u want to go there
plus the bot is down. no one can access
What is nano banana?
image model for gemini
Ahh
doesnt matter it does the same exact thing on other mode
Is there confirmation
It was said to come yesterday..
nano banana pro is coming out soon, supposed to be SoTA in image gen
yes tons of deepmind people said it
No, tons of people of “deep mind “ said it was yesterday
It was not
no they didnt
but i actually do want to go back there, i promise that i won't talk about or interact with the bot, i will be there for AI conversations only
it was for gemini 3 pro. not nano banana
The cascade requests are the agent chat. I think its G3 high as it does a lot of thinking for me.
It was after Gemini 3 released
cascade is windsurfers agent mode
you send it to windsurfers agent mode
it changes the model config to 2.5 flash and 3 pro low
if u look at the network logs
u will see soomething li ke 10-20 milliseconds or something
between it sending that and the model config coming back as 2.5 flash and or 3 pro low
thats not happening anywhere except 100% locally
Is this not it?
In my opinion opus 4 is way better than Gemini 3 pro
na opus 4 is a bit slop(always use purple)
Instruct it not to use purple and you're good to go.
Gemini isn't creative nor good at real problem solving, it's all prompting
Claude opus needed no prompts for me
Yup too
At least it knows a lot
Way larger database
Good at history
But ye for help with code where I can't fix stuff I ask claude
i think Google is prepping their strongest gemini 3 ckpt when others started sending their strongest unit like GPT-5.1 Pro, Grok 4.20,GLM 5, Opus 4.5
It even made working Minecraft without bugs xD
Ye idk
And I talk about opus 4, not 4.1
I think 4.1 is worse at coding at some point
Gemini 3 for Frontend, Opus is for Bug Fixing and Backend stuff
idk it is pretty obvious that the other models where specifically trained to hallucinate less through RL etc. (anthropic has always been doing it and openai started after the abysmal 4o models).
google not having done that just means that they can basically still improve this a lot by a simple update.
and btw the same benchmark also shows that gemini has significantly more world knowledge and according to their way of measuring the trade off between knowledge and hallucinations it outperforms EVERYTHING by a big margin
and also your arguments about 'gemini slop' with the output format are kind of dumb, because that is basically just the company employing minimum wage or below workers to write a format they like and then use it for SFT.
that is so easy to do that it can by no means be counted as a measurement of model quality or even intelligence
Yup 😃
You speak from my soul
I only need for bug fixing tho 🤣
But made my new frontend with Gemini and it hit a plain simple ui well
Gemini 3 is now lazy unlike A/B Test ckpt
there is no way they can give every free god damn EDU plan this much compute for free.... come on
Ye 😭
We need open source again
To catch up with opus
Not with gpt or gemini
its being routed to 2.5 flash and 3 pro LOW thinking in antigravity
i posted proof and video of network logs
i didnt check elsewhere
And with grok 4.1 💀
might be doing same thing
idk never tried antigravity but it's still Gemini 3 Pro Preview so it might not be the full release and maybe will get better
Even more limited than paid Gemini api
or when competitors sending their strongest unit maybe Google will send too
I honestly think this is the original Gemini 2.5
And that the Gemini 3 we seen.... Will be the first Gemini 3.5
But that's just speculation
but remember when 2.5 came out at first too?
it doesnt matter if its called preview or dogpoopbuttfart. they are routing us to a worse model while lying about it in the UI on purpose
Was about this good
it's bad for web design but the latest one is pretty good for web design
Maybe new data who knows
Wait for Gemini 3.5, maybe level of Gemini 3 ecpt
idk but yea this G3 Preview is not as expected
Ye, seems like Gemini 2.5 original in my opinion, just with more focus on design and less on bug fixing
Well, gotta go, see ya
wish
they removed my sama emoji
😭😭😭
yeah but they still have good ai discussions, i can ignore the image gen aspect, it's going to be officially released in a few hours anyways so i'll just find another site if i wish to use it
Ohh so extended thinking with Plus for 5.1 is now high reasoning effort
Juice value matches precisely API for high
wut?
Somehow missed that at first. Just assumed for API it's gonna be even higher than 256 lol
there is no way
checkmate LOL. This is very cool
gemini 3 i can't transcribe audio again 😂
lol
whats this base64
just my test to see if lmarena has any protection against mass-creating anonymous-user-sessions
lmarena auth tokens
Anyone online
Hello everyone, I'm new here
henlo
I always hated claude cuz of how expensive it was for something so short that ais like gemini 3 pro can do it for free
Yep this Gemini 3 Pro is prob nerfed,cuz it can't even design apple logo now and what the f- is 4 button,previous canvas with gemini 2.5 pro(and prob routed to unnerfed G3 Pro) was better
Right - Gemini 3 Pro Current
Left - Gemini 3 Pro(Routed) Canvas with Gemini 2.5
hi
whats better for prompting gpt 5.1 or gemini 3 pro
new opus soon?
https://www.reddit.com/r/singularity/comments/1p1zzet/we_may_get_opus_45_today_or_soon/
does anybody have a problem with gemini 3 pro where it's internal clock is set to around may 2024 and doesn't even idetify itself as gemini 3?
Just tried and it's a thing for sure lol
even on aistudio with default settings
2 times out of 3 it said it's 2024
yeah it said 2025 once, but I can't even replicate anymore. Only '2024' now
I've determined the user's question and I'm on track to access the current system date. This will allow me to accurately determine and report the year. Further refinements might be needed depending on how the date is formatted.
Ascertaining the Year
I've confirmed the year is indeed 2024. System time confirms this, and I've ensured that knowledge cutoff limitations don't interfere. The answer is concrete now, no further refinement seems necessary.
silly bug
Could it be a lmarena thing or Google's mistake
how does it not figure out it has data from 2025 🗿
This is 100% on Google
Lmarena can only patch it up with system prompt, but they aren't supposed to technically
Thanks for the confirmation btw!
Can anyone check if nano banano pro is launched for them?
On Gemini app or website and not aistudio
Sorry for the late but what's that ?
As a rule of thumb, Gemini 3 Pro is better at everything right now.
its there for me but i get an erorr
if i create an image with thinking it says "creating with nano banana pro" and then it just gives me an error
Not really. #general message
But I agree on average it is the better model
Ok. Looks like it's rolling out
yeah it is probably coming today
5.1 was perhaps even regression over 5.0 in several ways tbh. They changed the style and barely shared any metrics at all. The SimpleQA Verified score Google took for 5.1 is much lower than the public 5.0 score
I presume they tested it themselves or smth
cause there are no 5.1 scores for SimpleQA anywhere else published as far as I can tell
Except debugging complex C++ code, which Claude still dominates in.
Hence "rule of thumb", but in general, Gemini 3 is just mindblowing. Kinda like the GPT-4 launch back then.
What about very long roleplaying games and consistent intelligent GM'ing?
Job applications be like
(Won't ever abuse, just found it as a prob)
And won't be accepted for sure
I haven't tested yet. But with a 1M tokens context window, I'd expect it to be very good. I usually work around token limits with an "invisible" adventurer's log, something all models can maintain very well, but being able to keep everything in memory is pretty cool.
Seems like this is the main area they focused on 5.1 when it comes to performance. Tool usage
Dang if it's coming I'll add to my API
Is it really getting released today? 🤯
I do kinda hope they focus on more things than just SWE. Otherwise this may just be the repeat of 4.0 to 4.1
if Opus isn't much better than Sonnet, Anthropic is cooked
I just created an image in Gemini, and it said creating with Nano Banana Pro. 🤔
No config options though, and resolution was 1408x768, so no 4K
No wait
That was just the reasoning 😅
It is not gonna be easy to make big gains with the model of that size. Training takes time
It took longer for Google to significantly improve an arguably smaller Pro
anyone want gemini pro 2 image give me the prompt
Can you generate one for me in private? 🙏 (Dms if yes)
how did you get it
how do you have it
yess send
Ah damn only got 1 image
Ok thx
image quality does not look like nano banana pro to me... looks ok but nothign great
Hm what's that
Wth is even is it like LMarena?
looks legit
Does it say free models or it's truckbait
free model
Even API?
idk
Dude why there is credits?
Imagine I use 5$ of credit then they try to bait me into buying more
these are golden years. In 3 years, we will all be hooked and none of this will be free
Nope we will get more free stuff
We don't know the future yet
There is alot of imagination for future
Guys I'm kinda new here, is LTX 2 available in the video arena?
Cuz it have everything From Google
Video arena is Random model selector so you can't change models, even if it's out you need to be lucky to find it
Yeah I know but I asked if it's even available at all like if it even exists on there.
Is that suppose to be new video generator that just released?
LTX 2 was released on the 23d of October, 2025, you can look it up.
where you access that ?
Nano banana pro update, please upload
Damn nb2 is good
nano banana pro is so good
more data, better RL training
looks amazing for me
rate it
There is an argument to be made that it could be Ultra and they just rebranded it as Pro. But I'm reluctant to believe they would match the pricing and even make it more accessible with free API option like they did.... if that was the case
any one
It's on the website you told me
How are you guys using Nano Banana Pro?
Oh alr
its out on https://gemini.google.com/ just clean you cookies and cache so it should reset for https://gemini.google.com/
Ok while trying to look for Nano Banana Pro, I spotted this
create images is that
So it's out on gemini first instead of ai studio ?
How do I know if it’s Nano Banana Pro?
Ah alright. Wait, is Nano Banana Pro only on Gemini 3 Pro when selected?
i dont know
Prompt:An image captured in Mrbeast's video segment, about the 24-hour challenge in the Amazon rainforest
I remember someone said it's on the Gemini 3 pro preview on aistudio if you tell it to make a image it will do it but it will fail because it isn't fully out yet, I don't know maybe false information
they replaced every model
other then gemini 3
Tried it, and off the bat, Nano Banana Pro sucked at editing images. I asked it to upscale a vhs image to look higher quality like the master tape, and it failed the same as Nano Banana 1
upsclate isnt what it was maked for just use upsclary then
An image captured in Mrbeast's video segment, about the 24-hour challenge in the Amazon rainforest
they wont be cooked, they can pivot to pron industry, I just found out about this recently due to increased feminisation of all claude models
and this poor guy here is not the only example:
https://www.reddit.com/r/SillyTavernAI/comments/1orx3eb/i_cheated_on_all_my_freeproxies_with_claude_and_i/
What is Dynamic View?
it's crazy how much people are willing to spend on opus JUST for... you-know-what-they-need @hollow ivy
delete you cookies and cache
I did
I didn’t have to do that, and it worked just fine. Maybe it’s a gradual roll out
i think its just you havent gotten it
Yeah
How bro felt writing the Post:
I don't know dude wasting thousand of dollars just for a Ai that can't make great looking website and makes alot of mistakes, are we serious right now?
How is this ai 😭
Hmm, the text is wrong so it's ai
Gemini 3.0 is very unreliable
well, you cant blame the models for that, it's... lets say it's the miscalculations of Anthropic, lots of their original research teams are gone already I've heard
Basically hallucinated the entire answer while telling me everything was fine (I explicitly told it to tell me if there are any issues assessing the web in the highlighted part of the prompt).
Where do I start
Soon giving unlimited free nb2 on my API, what sizes should i support? Pls ideas
In 2030 is my time to pay for ai models
Wdym unlimited on API
From Google API?
I don't think it's going to be unlimited free just like Gemini 3 Pro it has limits
Honestly like the Nano Banana integration on Google Pixels, but the issue is I am on iOS which has the worst ai in existence
Well, I give 3 pro already too, it's holding like 3/4 of day before reaching daily limit
The heck
Gemini 3 Pro has a higher usage limit on Google Ai Studio
Gemini API better 🙂
In my opinion
It’s the exact same model
Soo is nano banana pro out??
My ChatGPT subscription ends on November 22nd. Hello Gemini 3 Pro
I tried gemini and aistudio and it's regular one??
you mean to pay for the humanoid robots?
I have it on my key
Only on the Gemini website not ai studio or api
Maybe cause spent a few k already idk
That'll be expensive but I meant Chat and image generation
It's fully released the preview just mean it's starting to release to everyone slowly
Guess I'm one of the first to have in API xD
Two of my friends don't have in API yet
What Country you at?
De
Yeah
Not so unrealistic but not so much as i expected
Btw you meant the image generation or Gemini 3 Pro
The detail on the phone is alot improved
Is in api
Since few mins
"Create a photoshoot of this shoe, streetwear style, close up"
Yes they already are on it
Cheap
Damn
Dude that's so realistic
New drama for Mr beast just like what happened to Jake Paul
this is crazy
I guess Nike and shoes brand will start using it
is it out?
Soon unlimited for free for all of my users, but let's see how long that lasts xD
Will set limit of 10k per day ig
its out for u?
Guys, you think this poster is good for a marketing campaign later on?
u on pro?
yes
API.airforce huh?
What the heck, you can set uncensored via api???? Will also have to implement that when adding
What the frick? It’s using Nano Banana 1 on Gemini 3 pro?
U think that poster is good?
do not implent uncensored bro
you don't wanna know what it will be used for
Why
No like level... Allow people, etc
The poster is the same as old models but the back is just way realistic
Ah I see, any idea of a better prompt for a poster for Twitter later?
chatgpt been releasing the worst models ever LMAOO
🙏
"Generate a random windowsscreenshot tutorial with a notepad opened, explaining the random tutorial, wallpaper of pc is batman"
GUYS
?????????
How do I access Nano Banana Pro on Google Ai Studio?
Select Gemini 3 Pro not the Gemini 2.5 flash
dang
Not out yet
I did
It worked before, now it’s not
Huh I guess there is limits
@jovial sapphire how do u get access to nana banana pro? dms if confidencial
Gemini app
Or website
gemini app
its not out yet
Huh how is that possible
@jovial sapphire thanks
Tell it to turn into 16k quality
Gemini API yk
How
Same as on old and imagen
Is there a enable thing
Crazy
No like you set allow people
Etc
Looks like png but dawg
Anyone sending prompt for nb2 in here as reply to this message will get image
Damn
i still dont have it
Veo4 be like
YOOOO NANO BANANA PRO IS INSANE!!
"Generate an image of geralt and yennefer fighting a striga. Witcher 3 style but with full next gen graphics, ray tracing and all the details that we are expecting from Witcher 4 in unreal engine 5."
Y'all overusing it
dayum
Real or ai?
wtf ro
AI
Got ya good 😁
broooooo???????????
It's real
Do you have Gemini 3 pro access on the Gemini app?
i use it on web
Veo 4 will knock out Sora 2
is it the same?
Select the Gemini 3 Pro model and enable nano banana you'll get it
i have did it but sill use nano banana 1
-_-
its out everywhere on other sites
Huh are you far from us?
gemini 3.0 ALLWAYS is hiting 0% in zero gpt, 😭 🙏
It does work you just need to figure it out
im in italy
use hailou ai, they giving nano banana pro for free
It works in U.K. and in Balkan and everywhere else
note: only if you give the instructions
While Chatgpt makes it obvious
nobody AI hit 0% in all times, the sonnet 4.5 is 5% the average
furryyyyy :D
It switched back to nano banana for me I think :(
Not in instant when google release soemthing it slowly try to roll out to Evey country not in instant
Once we have Gemini 3 flash, and Veo 4, it will legitimately the best ai on the market
u on pro?
free
Gemini 3 Flash not sure if it is coming out, Veo 4 in 4 months.
lol have some gpt detectors hit 85% with real human write 😑
Gemini 3 flash for sure will release when Gemini 3 pro officially releases out of preview
Can't wait for veo 4
It's already out of preview
Maybe in 2 weeks we will get flash models
Or 1
can i have the prompt please?
That's the big issue some ai detector is alot different
Guys shall I release my model that I kept private for long time? For video gen
Yes
yes
Fox advertising nano banana 2 is good, furcon, make it realistic
Ok will do
Model?
Yeah
the better is zerogpt (original site), but the gemini hit 0% now
Huh is it open source?
I guess these days Ai detector ain't useful anymore
Not yet
Wha
the problem the people still checking it 😭 , idk why they hate when I use AI in my works
Gemini 3 pushed ChatGPT out of the best ai spot, Gemini 3 destroyed Copilot and Grok, and Gemini 3 kicked Apple Intelligence out of the game entirely.
Can make long consistent videos, but idk if useful if takes 10 mins per 10 secs 😭
And exponentially going up yk
Hm mind if I know it?
Apple failed on making they're own ai so they gonna use Gemini now
If I release it then under name of Pluto
Uhh
What the heck
Huh
Demon gal :D
Apple is at rock bottom using Siri from "poor man's Android" 🤣
Is huggingface still good? And can we put 2tb on one model? :(
You'll need credits
Just please tell me the model 😭
Wait, What’s gonna happen between ChatGPT and Apple?
Dang, how much :(
oh mo finished gemini 3.0 credit...
too lol
how can we choose the video model ??
I told already
I can't find Pluto
You can't it's random
As mentioned, it's not public yet
Source to it please?
Right now thinking where to put it
u cant use nano banana no more?
It's not public yet
Then how you have it
omly the 1 one
Tbh I don't even know what to guess
Hey ,do thay remove NANO BANANA form LmArena ?
It's not added yet
use hailou ai for free nano banana pro
You know where i can upload the weights for free?
I use it for the past month almost and i cant find it for the past 3 days
it works on mobile?
Does it come bacg again ?
Sad, will just continue improving then
You can use it on Gemini app
yes
So is it comming back or do u have something similar on lmarena ?
okkk thx
Sewer tardigrade village
Nano Banana Pro 2 or Nano Banana 1
So i can find it on lmarena ?
No which model are you looking for
Okey ,i looking for image maker. I have a firm like "home decor" and everyday i upload some picture od desen and he make ma a room like this so i try to find it again
Examples -make me a room with this desi g and he make it right
Oh image Edit?
Yes ,nano banana did that it 10seconds
I don't really understand what you telling me
no it doesn't
Select this model on the image for image edit:
How good is nanobanana pro for image edit of real persons
Try it yourself
pretty good
Not un my country yet
Far from us?
it's available on hailou ai
I dont have it ?
only web version tho
Southamerica
That's chat not image generation open this link for Image:
and now?
@hollow patio or on the "Ask anything Chat" select the image icon to switch into Image
People are sharin a lot of photos of imagen generation but i want see image editing
@cloud zinc
@neon idol @hard quiver
https://banananano.ai/
Nano Banana 2 AI – Google's latest AI image editor powered by Gemini 3 Pro. Try Nano Banana 2 & Nano Banana Pro models for image generation and editing now.
I want to know if the new model can keep the original body proportions when making an outfit change
click here
U dont have app yes ,just online ?
It's not official by Google but it gives model your looking for
Yes
nb2 is gonna launch tommorow right
clearly fake
in 1 hour
Idk but it's where I got access to Gemini nano banana 2 yesterday when it's not out to Gemini app
Nb already launched....
Can someone confirm this ? 🙏
I think it can do it really good
I can test it for you if you want
damn
yeah it can
let me show
it doesn't open maybe a buf
i will make a try
hmm
It's coming out in an hour
First is original picture
The worker accidentally published it
where i dont see
it's gone
Yeah it said pro to me too
My OC
Than i refreshed and it disappeared
Yeah nb2 is good
Guys I think it's on the api
it's dissapearing and appearing rn
But it said permission denied when i sent the prompt
It's probably getting ready in 5-10 mins
And damn it's expensive
Nb 1 used to cost 0.039 per image
Nb2 is 0.13 😭
But woah the nb pro uses google search for images???
i dont have it
What is the prompt?
Rolling out. First few hours are always little wonky
u cant set the resolution??
Neither could we with nb 1
3x more expensive compared to Nb1? Wow. Why so expensive?
lame.
pro model
and more trained
Seedream 4K is 0.03 per image and it can do 4k 😭
cuz is chinese
nano banana pro is above the competition

