general

[limber crag]

Nope

[keen beacon]

Hmm

[limber crag]

Rate limits don't last for hours

[viscid cloak]

Just one quick tech question if gemini3 image preview’s server undergoing some troubles, is Battle mode gonna less likely to call its api?

[keen beacon]

I only use battle when I need to use lm arena

[echo aurora]

gemini3 image preview’s server undergoing some troubles, is Battle mode gonna less likely to call its api?
I'm not sure about how the specifics work, but in Battle mode in theory you shouldn't be running into the Something went wrong generic error message. It should sample a new model if one it initially tried didn't work.

[viscid cloak]

Gotcha, thanks boss🫡

[obsidian cargo]

Zoot Suit Samus

[spark python]

i have the same problem

[keen beacon]

Nawh Soras got more smooth animation like they do movies and stuff

But surprisingly grok is really good at animation

[echo aurora]

Really sorry about that! Our team is looking into a fix. cc @dark sage

[viscid cloak]

What about hailuo? Are those popular anime to real human films usually made by hailuo?

[echo aurora]

This is hella cool. What model?

[obsidian cargo]

nano banana pro, one of my few generations that succeeded :P

prompt was literally just "zoot suit samus" too

[spark python]

Its been happening for like 7 hours already, I understand the traffic is high but its an inconvenience 😭😭😭

[obsidian cargo]

oh hey got another success

metroid bread

[dark sage]

Thank you, I'm waiting, otherwise I was already scared that this was censorship.

[keen beacon]

Would you speak at the devil look what I just saw on YouTube

https://youtu.be/FlClOCpiztI?si=Z0WW4bBP90Td3fIP

Hehe I want to go check this out real quick I’ll be right back guys

[spark python]

Banana pro is like sora 2 first day

[pulsar agate]

GUYS

[spark python]

no guardrails

[pulsar agate]

GUYS

[spark python]

What

[pulsar agate]

WHATS THE DIFFERENCE BETWEN VIDEO ARENA 1 2 3

[spark python]

Yall know another site I can use nano banana pro for free?

[spark python]

No difference

[pulsar agate]

KK

[keen beacon]

Ya it’s gunna get nerfed soon lol

Disney is about to launch their own and they’re about to start charging royalties

[spark python]

I dont think so they never censored it

open ai is censoring their best model

and lowering the quality

just to get money

[keen beacon]

https://www.cartoonbrew.com/tech/disney-plus-artificial-intelligence-tools-bob-iger-256752.html

They just need to get these court cases over with dammit

They’re going on too long just make a decision if it’s OK or if it’s not OK

[spark python]

they want to use it themselves

so they make their own money from ai

[keen beacon]

Yeah, I know but damn somebody needs to step in the middle and just decide once and for all if it’s fair use or if it’s not

[obsidian cargo]

they need to dismantle intellectual property laws, make all fictional characters creative commons

[spark python]

only fair if u pay them

[keen beacon]

Man, the law is always so behind

I can’t imagine the state like California was gonna stand still

[hushed gyro]

@echo aurora Am I allowed to repost my feedback in different posts because I dont have any threads rn

like each issue to each post

[keen beacon]

I know the game plan

https://apnews.com/article/udio-suno-ai-music-universal-b90f9f5f968101ef617e41c5369da02a

[echo aurora]

Yes, ofc. Adding your feedback onto already existing feedback posts about the same feedback is perfectly fine.

[keen beacon]

They’re gonna do the same thing they’re gonna partnership

And they’re gonna do it as soon as the technology starts getting good enough lol

They can’t put the genie back in the bottle and they know that so the next best thing is just partnerships

[hushed gyro]

Oh I mean like new posts, I don't think there's any existing ones (or not related to my feedback)

[spark python]

udio is so bad

model wise nothing is new

they have 3 models

[keen beacon]

So what does all this mean, that you’re gonna be paying for ai tools to generate content from intellectual property brands and then you’re gonna be paying these companies and you’re not gonna own the content you generate

[keen beacon]

They’re huge dude they might be even bigger than Suno

[spark python]

pineaple

[echo aurora]

If there is a specific piece of feedback that doesn't already have an existing post, then you can make a new post. But don't create a new single post with all issues you'd like to bring up.

[spark python]

I got a working image

[echo aurora]

[keen beacon]

Speaking of Suno

Dude, that company generates so much money. It’s insane. I would’ve never guessed.

https://techcrunch.com/2025/11/19/legally-embattled-ai-music-startup-suno-raises-at-2-45b-valuation-on-200m-revenue/

Yeah, look you guys you can’t make this up

https://www.inc.com/ben-sherry/suno-ai-startup-just-hit-a-2-45-billion-valuation/91267866

🤣🤣🤣🤣🤣🤣

Despite being sued by nearly every record label, that’s hilarious

They’re the last biggest target

[vast fern]

I LOVE SUNO

[keen beacon]

Yeah, they’re the big target now for the music industry now that udio fell

[vast fern]

lmareana music ai leaderboard when ?

[keen beacon]

https://youtu.be/9MB0dm0vv2U?si=4mt_kjDEQbcIAvob

[vast fern]

yeah i wanna listen two different music and pick best one

fu kk labels

[keen beacon]

Well, it’s super messed up how they settled it

You gotta look into it and I’m sick. I sound like a broken record always repeating myself about this story.

[cloud zinc]

ai voice

[misty vault]

ai sydney

[neon idol]

Ai pizza

[wicked mason]

ai sponge

[keen beacon]

All I’m saying is it’s super messed up what they ended up doing

The way they went about it

[misty vault]

shaking and crying rn

[keen beacon]

It just shows that this whole industry is about money and not the users

[misty vault]

https://tenor.com/view/of-course-gif-21373815

[neon idol]

most honest reaction lmao

[keen beacon]

https://tenor.com/view/simpsons-ned-okily-dokily-gif-4289162988260002514

[high ginkgo]

So Real

[keen beacon]

All right well we’ll see where that energy is at if God forbid this comes to fruition

[neon idol]

https://tenor.com/mlcphuqeCXO.gif

[keen beacon]

If you got people crying about that retry or resubmit button, I can only imagine what would happen and the outrage if God forbid this would occur to other parts of AI that we all enjoy.

[hushed gyro]

the worst example (of all AI companies) is probaly openai, their models are full of censorship

literally the opposite of a non profit

[keen beacon]

Yeah, but you know the users also have some blame

You need two to tango

You got die hards out here who still think chatgpt5 was an upgrade 😂

[dark sage]

hooray, it looks like the images are being created

[keen beacon]

https://youtu.be/YWzty3ADuLY?si=1RiFiq-txnOApI_B

[echo aurora]

Glad to hear it!

[keen beacon]

This video made me laugh so hard

I can’t believe that guy freaked out like that and he’s a politician lol

[knotty fable]

Hirr hirr.

[river minnow]

@echo aurora I know you probably got asked this a lot by now, but is your team looking into the problem of the image generation? It always says "Something went wrong [...]". I tried it today in the morning and since then it doesn't work. Its the same for Nano-Banana 2, Seedream 4 etc.

[knotty fable]

@echo aurora Voting appear to be somewhat on the blink - not entirely but definitely glitchy.

[dark sage]

Will there ever be any censorship added for character creation of IP‑protected characters (such as real people or famous characters) in Nano Banana Pro on LMArena?

[sullen sand]

today I would like to eat some pinapple

[spark python]

Pineaple WE ARE GOING TO DIE

without the site working

our lives depend on it

I only got 1 image working

[fickle venture]

That's just the bot failed but try again

[spark python]

nvm pineaple im still alive I got another image working

[fickle venture]

That message for Google Gemini Not LMarena

[fickle venture]

If image isn't getting created it just means the model is over loaded by alot of users so just try again later

[tall tulip]

Is it?

[fleet lintel]

I am able to generate images without issues (tried few just now)

[abstract orchid]

thats dope LMArena seems to have good catch with all llms

[echo aurora]

https://tenor.com/view/kedinehir-nehirkedi-mitcikedi-gif-9396221664717142950

[echo aurora]

[echo aurora]

Bit hard to answer, but overall we (and the model providers) have guardrails for what their model responses can/can't be. What that looks like specifically for IP‑protected characters I couldn't tell you.

[echo aurora]

Hmm looking into.

[echo aurora]

I was able to vote, but I am seeing this which is a different problem

[spark python]

I know yall got nsfw filters but I got past them

[knotty fable]

Not all got those. 😉

[cloud zinc]

how

[spark python]

all what?

[knotty fable]

So no need to hack, just use those Ai's who allow such.

[spark python]

ahh the undress ones

[knotty fable]

[Talking about general NFSW now, not actual ic pron.]

[spark python]

what is ic pron?

[sullen sand]

yes

[spark python]

ik pron but IC

like ai?

[knotty fable]

When they hump each other etc.

[echo aurora]

Hey how about no.

[spark python]

😭

[sullen sand]

bruh

[spark python]

I couldnt generate them anyway cus image gen doesnt work

🙄

[echo aurora]

But if you do have a reliable way to get past filters I'd love to hear about it so we can patch that.

[abstract orchid]

how is this possible

[spark python]

why would I do that

[echo aurora]

Because it's the right thing to do.

[spark python]

plus dont worry its nothing THAT crazy

[sullen sand]

he can ban you I think

[spark python]

its soft stuff

FOR what

I didnt send anything

I was just saying its possible

[sullen sand]

I THINK idk

[cloud zinc]

dont do it, its a trap

[spark python]

I know

but for pineaple

dont worry its soft stuff and niche

[sullen sand]

pinapple its a good person

and tasty..

[cloud zinc]

hopefully its hard stuff

[queen veldt]

https://tenor.com/view/dexter-james-doakes-surprise-mf-can't-prove-it-gif-7408854170742055240

[spark python]

this makes me smash my desk

[tall tulip]

Same

[sullen sand]

yeah

[neon idol]

i feel u

[spark python]

gemini when I tell it to put a picture on a wall:I can help with editing images of people, but I can't edit some public figures. Is there anyone else you'd like to try?

[limber crag]

Hey have the error rates in nano been fixed?

[limber crag]

Ig not

[echo aurora]

It's still looking like we're working on a fix.

[knotty fable]

Send images of that smashed desk please!

[hushed gyro]

WTF is this lmao

[fiery gull]

LOL 😭😭😭😭😆

[fiery gull]

The model trained with it will be ahead that gemini 3.0 pro

[hushed gyro]

Chat is Gemini 3 pro superior to gpt 5.1?

[cloud zinc]

its good

[knotty fable]

Hugging face guys are a riot at times. 😺

[fiery gull]

3.0 pro > sonnet 4.5 > opus 4.1 >>>> Gpt 5.1

[hushed gyro]

WTF Openai is so behind 😔

[cloud zinc]

grok 4.1 >>> 3.0 pro

[sullen quest]

no

[hushed gyro]

Doubt

Grok models were never the best

[spark python]

pineaple fix this mess or I will do something bad to my body

[hushed gyro]

Guys I think kimi k2 thinking is underrated

[cloud zinc]

u wrong. grok is most uncensored and is a beast model

[hushed gyro]

Uhh lemme guess

🤔

[fiery gull]

3/10 ragebait

[hushed gyro]

Well I can't generate uncensored content with the guardrail

(pls make me switch it off)

[fiery gull]

Calm down 😭

[cloud zinc]

u are hater of grok

[limber crag]

??

[waxen fern]

Lmarena is now using turnstile again. God help us all.

[hushed gyro]

I'm not lmao

It's just not the best model I've used

It's either gpt 5 or kimi k2

[hushed gyro]

Wdym

[fiery gull]

Turnstyle? What is this? Is bad? 😭

[cloud zinc]

why u using chinese kimi k2?

[hushed gyro]

It's excellent lmao

Guys do you find the guardrail annoying

[cloud zinc]

u work for china?

[fiery gull]

The kimi k2 thinking turbo >>>>> grok 4.1

[hushed gyro]

Holy ragebait

[hushed gyro]

Thank you

I think it can compete with gpt 5, no joke

[cloud zinc]

bro is glazing kimi k2

[sullen quest]

cloudflare turnstile

[spark python]

im just joking

[fiery gull]

Personally, I hate GPT 5.0, so I think I actually prefer it

[hushed gyro]

Yes because it deserves the glazed donut treatment

[fiery gull]

Kimi k2 thinking is really good now, it stop hallucinate

[hushed gyro]

K2 never did since I started using it in June???

[fiery gull]

When I started to use the kimi k2 thinking (PREVIEW) it hallucinate some times

But the stable version no

[waxen fern]

Cloudflare verification

[hushed gyro]

Guys is qwen 3 max good or no

Cuz I have been using it for a month and I think its quite good

[fiery gull]

Yep, is soo annoying

[native yarrow]

they've been using it?

[hushed gyro]

It pops up every login, and every 5 messages

LMArena lemme grind nano banana Pro 😭

[fiery gull]

Hell no, I fell it is worse that qwen 3 235b

[hushed gyro]

Should I use 235b 2507 or VL?

[fiery gull]

Same thing, but I stoped to use, I just tested the vl, (I'M NOT A CODER) 😆

[magic ravine]

Welp, back to the errors for Nano Banana Pro 🫠

[obsidian cargo]

@echo aurora it's busted again 😭

Or still busted? Idk

[jade egret]

I wil say yes

[echo aurora]

Hey let's not say things like this, even if it's in a joking matter.

[jade egret]

.

[echo aurora]

Yup, since reporting this morning it's continuing to be a WIP.

[native yarrow]

brah 3:

[echo aurora]

Currently, sometimes it'll work, other times it won't. But we're working on getting this working again.

[magic ravine]

Pain 💀

[native yarrow]

it worked for me just now

i notice it's just slower

[magic ravine]

https://tenor.com/view/you-dropped-this-king-you-dropped-this-queen-you-dropped-this-you-dropped-this-king-meme-you-dropped-this-meme-gif-25206668

[jade egret]

also so is claude 4.5 opus today?

[native yarrow]

the model isn't nerfed by quality it's just slower..,.,

[fiery gull]

I hope yep, but I think the opus 4.5 need more 1 week to be released

[hushed gyro]

We really need an edit button

Seriously I generated the same prompt twice because of a lag 😭

[native yarrow]

yeah nvm it's down now for me too...,,

[hushed gyro]

LMArena is stuck on the verify page!

[jade egret]

bruh

[waxen fern]

@echo aurora we need to remove the cloudflare verification

[cloud zinc]

opus released

[cloud zinc]

check https://platform.claude.com/docs/en/release-notes/overview

[torn mantle]

omteresting

[cloud zinc]

opus released

[native yarrow]

opus won't compete with gemini 3

unless.,,.,

[torn mantle]

gimme more info socky socky

seems like its small model ?

[fiery gull]

hm?

[torn mantle]

i mean smaller

[hushed gyro]

@echo aurora Pls add immediately! 1!1!1!!!1!1

[fiery gull]

hmmm...? serious

[hushed gyro]

😔

[native yarrow]

coding wise

[queen veldt]

That pricing is sus

[cloud zinc]

yes https://platform.claude.com/docs/en/about-claude/models/whats-new-claude-4-5

[native yarrow]

still i don't imagine a model overtaking gemini coding this quick

[queen veldt]

Why is opus 4.5 cheaper than opus 4.1??

[native yarrow]

:P

[cloud zinc]

official opus 4.5 page will be up in 10-20 min

[cloud zinc]

3 times cheaper

[jade egret]

?

[jade egret]

released?

[queen veldt]

Nahh that opus is sus

[cloud zinc]

yes

[native yarrow]

groks only good for uncensored creative writing brah

[fiery gull]

claude app?

[jade egret]

YOO

let see if it betetr than gemini 3

[queen veldt]

It is in coding for sure

But their pricing is sus

Opus isn't cheap!

(shouldn't be)

[jade egret]

:0

[torn mantle]

lmao

consumes usage limits faster

pfffffffft

[native yarrow]

what's the point of using it if you get like 5 messages 💔

[torn mantle]

we will get it for free on lmarena

soon

didnt they like buy tons of gpus from amazon

[native yarrow]

with an insane rate limit 💔

lolz

[cloud zinc]

page is uphttps://www.anthropic.com/news/claude-opus-4-5

[jade egret]

it better than gemini 3?

[cloud zinc]

yes in coding

[jade egret]

:0

[sudden coral]

gemini 3 flash preview

lmao

[cloud zinc]

37.6 in arc agi 2 vs 31.1 percent for gemini 3 pro

[native yarrow]

hmmm

promising

eager for it to drop on lm

[torn mantle]

looks like it but its expensive compared to it

[cloud zinc]

are u gonna pay for it?

[torn mantle]

im gonna use it on lmarena for free

🙂

[fleet lintel]

great work with Claude.

[torn mantle]

i was never fan of anthropic models, seems lazy in general usage

[fiery gull]

I hope it, I'm tried of gemini 3.0 pro

[cloud zinc]

then why u complaining about price?

[fiery gull]

deepseek v4 when? I'm tried of Opus 4.5

[fleet lintel]

so basically. coding use anthropic. everything else use gemini 3

OAI should just pack the bags at this point

[cloud zinc]

anthropic was waiting for gemini 3 pro to drop

before releasing

[fiery gull]

hell no

[torn mantle]

how

????????

noi difference between thinking on / off ?

[native yarrow]

dawg gpt sucks ahh

[fiery gull]

1/10 ragebait

[native yarrow]

chatgpt is unironically one of the worst models compared to what is out there now

[fiery gull]

wow the no thinking is soo good

[fleet lintel]

its too obvious at this point. i tend to ignore him

[fleet lintel]

I see hiring freeze soon and some big changes soon in OAI. future is bleak for them

[native yarrow]

id agree :P

[fiery gull]

Today is the OpenAI most irrevelant day?

[fleet lintel]

i was worried that google might fall asleep again. Love to see Anthropic giving good fight!

[cloud zinc]

https://winbuzzer.com/2025/11/21/leaked-memo-sam-altman-admits-to-rough-vibes-and-economic-headwinds-at-openai-xcxwbn/

[native yarrow]

well, now google will need a response is 4.5 is as good as ppl say,.,,.

[torn mantle]

they need to bring old checkpoints back

this preview gemini 3 pro is dumb

[native yarrow]

i want to test it brah but i am not paying anything to anthropic 💔

[torn mantle]

same

im not paying

why would i?

[native yarrow]

exactly you barely get any messages even if you do pay

[fiery gull]

same thing gemini 2.5 preview < opus 4 < (or =. ) gemini 2.5 dont preview

[native yarrow]

i hope 4.5 is actually good at creative writing

unfortunately its going to be HELLA censored

as anthropic is

[fleet lintel]

Google looks satisfied right now.. I dont expect a good response before 4-5 months

[jade egret]

yea

[native yarrow]

not as bad as GPT though

GPT is so awful

like it just makes up things and then it censors it because of the thing it made up when you didnt even mention that

[torn mantle]

https://x.com/cursor_ai/status/1993031841901928829

Cursor (@cursor_ai)

Claude Opus 4.5 is now available in Cursor!

It's 3x cheaper than Opus 4.1 with better performance.

Try it out at Sonnet pricing until December 5th.

X · Nov 24, 2025, 6:59 PM

[native yarrow]

pal chatgpt is nowhere near these two models as of late 💔

[torn mantle]

live on cursor

there is a reason why they said 'vibe coding'

[fleet lintel]

Gemini 3 flash will be launching in Dec. I think new flash model will be better than gpt 5.1 models 😄

[torn mantle]

its only good for vibe coding and not real projects

[native yarrow]

just test it for yourself brah

theres a diff

[jade egret]

no..

[native yarrow]

ts ragebait

💔

[torn mantle]

:/

dumb

[fleet lintel]

really good work by Anthropic! kudos to them!

[cloud zinc]

google already pushing back

https://x.com/_philschmid/status/1993032447291015282

Philipp Schmid (@_philschmid)

Excited to share a System Instructions for Gemini 3 Pro that improved performance on several agentic benchmarks by around 5%. 🚀

We collaborated with the @GoogleDeepMind post-training research team to include some best practices in our docs. 🤝

X · Nov 24, 2025, 7:02 PM

[fiery gull]

hell no, the openai have a respond lol, I love this era

[cloud zinc]

5% improvement in agent benchmark through system prompt

[fleet lintel]

is this 5% over the current system ?

[jade egret]

already in gemini app?

[cloud zinc]

yes, thats what they claim

[cloud zinc]

u have to paste that long system prompt

[fiery gull]

what is this mercury v0? 😭

[torn mantle]

aha

[fleet lintel]

This tells me that base model gemini 3 is very very strong.

[torn mantle]

thats what i said

im always right sigh

gimme my prize

[fleet lintel]

that would be nice!

[native yarrow]

google wants to keep their first place of course

[fleet lintel]

anthropic wont lose their Coding prowess anytime soon. I have high hopes with them

[native yarrow]

google wants gemini 3 to be the first model people think of when they think AI, etc

chatgpt still actually holds that spot

surprisingly

but its because they were the first :P

[magic ravine]

Btw, anyone recommend the best model for creative writing so far? Gemini 3 is still disappointing. (Gotta keep myself busy as we wait for Nano Pro to get fixed) GPT 5.1 is decent so far but still not exactly that good for novelized writing.

[jade egret]

hard to say

[fleet lintel]

that will continue for 1 more year but slowly they will start dying. first it will be slowly and then suddenly.. all funcding will be gone and they will be force to slow down.

[native yarrow]

yes, usually the one to ignite the ai bubble will always be the first model ppl will think of

[fiery gull]

anxious for gemini 3.0 pro not preview

[native yarrow]

claude is the best ive heard but

bad censorship

💔

and gpt also has awful censorship

[magic ravine]

Damn 😭

[native yarrow]

id say best quality is claude if you dont care about censorship but

[random flax]

“Claude Opus 4.5 is out. When is Lmarena going to add it?

[native yarrow]

decent writing quality with no censorship is actually grok, as bad as the model is

[cloud zinc]

openai focused on shopping, while anthropic and google are cooking

[native yarrow]

pineapple and the team i think are working on the nb pro issue rn

what is polaris?

[cloud zinc]

gpt 5.1

[native yarrow]

theres no point in using that if the censorship is off the rails though

[magic ravine]

Only issue for Grok is that LMArena is also pretty censored 😅

[native yarrow]

you can barely do anything with gpt 5.1 like deadahh

ask like everybody on the openai subreddit what they think of censorship in creative writing with gpt 5.1

including myself

[magic ravine]

Aye. I dont mind waiting. I can think of prompts and ideas in the meantime.

[native yarrow]

its pretty bad.,,

[native yarrow]

actually on the grok website ive never hit the rate limit using their like new model or wtv

writing wise

and thats where its uncensored

:P

[vast fern]

dangg itt

[native yarrow]

and mind you i dont pay ofc

[magic ravine]

Ah, but gotta pay 😭

[magic ravine]

😱

[native yarrow]

yeah, i dont pay and ive never hit the rate limit yet

:3

[cloud zinc]

still gemini 3 pro is better for cost

[fleet lintel]

https://assets.anthropic.com/m/64823ba7485345a7/Claude-Opus-4-5-System-Card.pdf

[fiery gull]

I'm using the opus 4.5 now

[cloud zinc]

gemini is much better optimized

[magic ravine]

Well snap! Gonna head there now! Although, looks like I might still be on the low model. Still saying I gotta get supergrok, which I hear has no censorship.

[fickle venture]

Yoo finally opus 4.5

[normal abyss]

claude-opus-4-5 WHOAAAA

[fleet lintel]

Anthropic is all about coding now! huge market for them

[slim gorge]

opus 4.5 gonna destroy every other model ngl

[echo aurora]

[native yarrow]

theres 4.5

added to lmarena

sweet1!1

[echo aurora]

New model hype.

[native yarrow]

nah, supergrok is for the uncensored videos

[torn mantle]

ive tried it

[native yarrow]

i think text is just uncensored straight up

[torn mantle]

anthropic models are lazy af

[fickle venture]

Is it on Claude API now?

[torn mantle]

omg

[native yarrow]

DO Not waste money on supergrok pls

💔

[torn mantle]

'wont use opus 4.5 ever again

and dont bother

[cloud zinc]

where is 4.5 thinking?

when 4.5 thinking coming

[normal abyss]

its always 1 day later

[fickle venture]

It's on opus dude

[magic ravine]

Ah. So like, I can generate anything (Within reason of course) with just the basic model?

[native yarrow]

is this the model?

[magic ravine]

Text only

[cloud zinc]

yes

[echo aurora]

and Code Arena

[magic ravine]

Sweet! Will test that after

[cloud zinc]

why only on text?

[native yarrow]

yeah, pretty sure, i havent tried anything weird and kinda dont want to but id imagine that (MAYBEE) could also be uncensored too?

but

[cloud zinc]

opus model is for code. not for text

[polar niche]

NEW MODEL

[native yarrow]

ive tried things that chatgpt just outright refuses and ofc it works.,.

[inland quest]

thinking when?

[magic ravine]

Thanks. Will give these babies a whirl so long.

[echo aurora]

Which works well because we have a Code Arena contest running right now! #announcements message

[polar niche]

Is sonnet 4.5 coming too?

[vast fern]

give me prompts to test opus

[echo aurora]

[vast fern]

noo they will relase that in sep 2025

[magic ravine]

Ask it to turn your dream into a story

[native yarrow]

ill just ask it to make an asteroids game

[polar niche]

Will it be added to text?

[native yarrow]

similar to how gemini 3 was tested with that

[fiery gull]

the opus 4.5 is fixing a 3.0 pro work, it definitely is more smater that gemini 3.0 in docs !!!!!!!!

🙂

[vast fern]

opus is the og for coding

[echo aurora]

Should be in Text and Code Arena

[round island]

hmm

[magic ravine]

Again with the coding 😭

[vast fern]

what happened

[native yarrow]

apparently its lazy

i havent tried

i dont think its gonna be a bad model, it usually isnt

[echo aurora]

This is my go to Code Arena prompt: #1440135896174170214 message

[native yarrow]

it just depends on if its really better than gemini 3

:P

[hushed gyro]

KIMI YOU ARE AMAZING

https://019ab730-2511-7dbd-8c9b-81b9396ec472.arena.site

[native yarrow]

LOLZ thats a good one

this is

interesting

claude output

oh i just got pieced by an alien ufo

💔

[cloud zinc]

cuz of limited context window

[hushed gyro]

Chat are the hermes models well known

[torn mantle]

btw no thinking is kinda the same as thinking in swe

thanks lmarena for adding opus 4.5 this fast ❤️

[empty stump]

Im guessing its still expensive

[native yarrow]

yeah pineapple you the goat brah :3

[sudden glacier]

It's cheaper than prior Opus models but still expensive

[vast fern]

promt?

[echo aurora]

The full team at LMArena get the credit.

[empty stump]

How many messages do you get on direct chat

[magic ravine]

Just tested Grok and while no pictures (Says moderated), it actually gave me something spicy text-wise. So damn.

[native yarrow]

yeah, so text is uncensored but

[fickle venture]

Bro that's magnifique

[native yarrow]

photos not

[normal abyss]

it didnt cut me off at 5 messages, the rate limit being higher is actually so cool 🤩

[fleet lintel]

but 3x cheaper compared to before.

[native yarrow]

media you have to do supergro

but who cares ab that

[echo aurora]

The rate limit for this model is currently set to 20/hr, but this may change.

[modest prism]

Is opus 4.5 better than Gemini 3 pro in non-coding tasks?

[empty stump]

Not bad

[native yarrow]

20 aint bad :P

[hushed gyro]

Create a fictional URBANSCOOP newspaper snippet set in 2029 revealing a secret advanced PMC / private intelligence agency called DELTACELL.

[fleet lintel]

20 per person is more than enough

[vast fern]

it beats gemini 3 in everything

[fickle venture]

How Claude 4.5 Opus was released:

[magic ravine]

Aye. Got other models for pictures anyway so who needs Grok's image generator 😛

[hushed gyro]

Chat how is the new opus

[modest prism]

Are you serious?

[native yarrow]

the big banana:

[native yarrow]

https://tenor.com/view/why-are-you-gae-pepe-julien-gif-13687553676606303412

[fleet lintel]

looks like Sota in Coding.

[dreamy sparrow]

Hell no dude

[native yarrow]

does 4.5 srsly beat gemini 3

find that hard to believe so quickly

[fickle venture]

In coding

Only

[native yarrow]

ah

[fleet lintel]

in coding yes. everything else , nope

[empty stump]

Gemin8 3.5 better fix

[whole swallow]

What if gemini 3 pro + opus 4.5???

[magic ravine]

But yeah, not that most of what I want written are explicit anyway. Got a thing for just simple fictional stories.

[dreamy sparrow]

that's fake

[native yarrow]

what is this brah

[vast fern]

only multimodel is behind

that too they are not working on images model

[fiery gull]

bro the model has released in 3 segs

[native yarrow]

how do you like the quality?

like prose

its been good to me ngl

creative writing is the only thing ill ever use grok for lolz

[fiery gull]

hmmmm

[torn mantle]

is it working or nah

looks good to me

[native yarrow]

that is kinda cool, it actively deteriorates these weird shiedsl

no it works its just i swear space invader had a different style,.,

[empty stump]

Also for some more "unethical" prompts

[native yarrow]

^

sometimes i have to get shady.,.,

anyway, theres little bonus ufos that sometimes streak across the screen thats kinda cool

[magic ravine]

For grok? It's honestly not too bad. Does paragraphs nicely. And I'm sure with some guiding, could make for some decently written stuff. But it does win points in the uncensored department. So like I could use another model to write up an entire chapter and such, and then ask Grok to give me a spicy scene if needed.

[whole swallow]

I hope they invent some solid way of LLM collaborating, some groupchat typeshi so they can share the history chat

[cloud zinc]

lmarena got limited context window, so opus looks bad here

[native yarrow]

good stuff brah

lets see here

[spark python]

pineapple stuff seem to work

[flint sandal]

I WAS SLSEPING FOR 40 MINUTES. I WAKE UP AND OPUS 4.5 DROPS AND ITS BETTER THAN GEMINI 3? i tried to adjust to gemini 3 and i was almost done and now anthropic kills it..

[spark python]

wait

new model?

from anthropic?

[vast fern]

https://www.anthropic.com/news/claude-opus-4-5

read this

[native yarrow]

huh, thats pretty cool

[vapid elbow]

Is the image generation slow today or is it just me?

[native yarrow]

its an ongoing issue

[native yarrow]

planet focus works

[spark python]

holy

[torn mantle]

like we need some complex simulations come on

dont give it easy tasks

[vast fern]

danggg

[native yarrow]

even gives info on each planet

[native yarrow]

idk what to think offf 3:

[torn mantle]

big bang simulation with sophisticated details

and different phases

[native yarrow]

everything in this UI works

alright

[jade egret]

is polymarket gonna react

[torn mantle]

yea as i said anthropic models are more reliable

[torn mantle]

so you dont have the headache of double checking and re-prompting etc...

[vast fern]

give your top 5 prompts and we will do side by side google 3

[native yarrow]

if this actually works

[torn mantle]

yea

[native yarrow]

i'll be kinda stunned

[cloud zinc]

its based on

[torn mantle]

oh wow

first coding prompt

better than gemini 3 pro

[jade egret]

oh

[vast fern]

can anyone try these

1. Ant Colony Optimization Simulation
   Simulate a colony of 500 ants foraging for food sources scattered across a 100x100 grid terrain with obstacles. Implement pheromone trail laying, evaporation over time, and collective decision-making for finding optimal paths. Include ant roles (scouts, workers, soldiers) and simulate resource depletion affecting colony behavior.

2. Urban Traffic Flow Ecosystem
   Create a city traffic simulation with 10,000 vehicles across 50 intersections with adaptive traffic lights. Include different vehicle types (cars, buses, emergency vehicles), varying driver behaviors (aggressive, cautious), real-time congestion patterns, and simulate how a single accident cascades through the entire network over 24 hours.

3. Pandemic Spread with Human Behavior
   Model a disease outbreak in a population of 100,000 agents across neighborhoods with varying density, hospitals, and public spaces. Agents should have individual immunity levels, social compliance rates, daily routines, and decision-making about masking/isolation. Simulate how misinformation clusters affect regional spread differently.

4. Stock Market Multi-Agent Economy
   Simulate a financial market with 1,000 AI traders using different strategies (momentum, value, random, insider), market makers, and regulatory bodies. Include news events that trigger emotional trading, bubble formation, crash cascades, and emergent manipulation patterns. Track wealth distribution evolution over 10 simulated years.

5. Wolf-Deer-Forest Ecosystem Balance
   Build a predator-prey ecosystem with wolves, deer, and vegetation on seasonal terrain with rivers and mountains. Include animal aging, reproduction cycles, genetic trait inheritance, pack behavior for wolves, herd dynamics for deer, and forest regrowth rates. Determine if the system reaches equilibrium or extinction spirals.

[cloud zinc]

for antrophic to win, it needs to beat 1490

[slim gorge]

lmaoo polymarket is forbidden in romania

[keen beacon]

https://youtu.be/UI-KBieN6jo?si=dQnsUvFoX6z5lBqU

🤣🤣🤣

[fickle venture]

It's free on lmarena try it bruh

[keen beacon]

Ai is so dumb lmao

[dreamy sparrow]

yeah dude... Claude 4.5 isn't better than gemini 3 at all

it got it wrong so hard

[native yarrow]

good ui, design, no buttons work unforutnately

[keen beacon]

LMAO

[native yarrow]

and lets test the same prompt with gemini 3

[fickle venture]

Claude only for coding

[cloud zinc]

ok

[dreamy sparrow]

some kid said it's better than every model

[whole swallow]

Yeah but can compare pricing..

[dreamy sparrow]

in every way

[torn mantle]

show me

send code

[slim gorge]

give me the prompt ill prove that it can do it lmao... models cant do math problems without reasoning unless that specific math problem is in the training data

[dreamy sparrow]

dude claude 4.5

can

think

[vast fern]

its not for you man, you should use gemini

[dreamy sparrow]

: Find the GCD of this series set {n^99(n^60-1): n>1}

: Find the GCD of this series set {n^99(n^60-1): n>1}

[dreamy sparrow]

Sent it twice

[pure pasture]

dude im gonna go mental this is what ive been seening every 2 mins

[fickle venture]

Put him in a locker

[dreamy sparrow]

accident

[vast fern]

its cheaper then opus 4.1 and 4

[dreamy sparrow]

dude just say the truth

gemini 3

is better

[keen beacon]

The price will say it all

If it’s cheaper it’s shet

[whole swallow]

How much bro share please

[torn mantle]

only 2 phases works, i think this is so complicated

gemini 3 fails as well

it only creates like 2 simple animations

[dreamy sparrow]

me when a claude sucker can't see

[pure pasture]

@quick jackal what is this is been like this for 3 hours

[cloud zinc]

gemini 3 still better

[keen beacon]

You know the same you get what you pay for

[vast fern]

refresh

[dreamy sparrow]

wait I've seen you before

wait I've seen you before

[keen beacon]

If the model is cheaper, how is it gonna be better? lol

[obsidian shell]

Why can't anthropic just go through a normal release?

[dreamy sparrow]

ur the Elon Musk glazer

hah

[pure pasture]

wont work

[keen beacon]

Out here we go out the benchmarks

[slim gorge]

what's not normal?

[dreamy sparrow]

I still rememeber you

[cloud zinc]

u want vibe test?

[obsidian shell]

Rate limits.

[native yarrow]

if gemini 3 can do this big bang simulation claude 4.5 isnt better

[slim gorge]

right now the whole globe is using it cuz it just released lmao

ofc they gonna get overwhelmed

[native yarrow]

pretty simple a model can do something the other model can't its usually better

[obsidian shell]

Erm... You need to reevaluate your methods.

[whole swallow]

Wow that’s really cheap. But makes me wonder how much less powerful it is than it could be at full power..

[obsidian shell]

Correct.

[torn mantle]

nah

thgey ciooked with this model

at coding

[native yarrow]

gemini 3 just gave me a great big bang simulation

[keen beacon]

It’s the end of the year the end of the fourth quarter

[vast fern]

opus is made for coding only why are people comparing with other stuff

[keen beacon]

They all have to release something even if it’s nothing meaningful

[cloud zinc]

openai slacking

[obsidian shell]

Incorrect.

[torn mantle]

lol

[obsidian shell]

Ouch.

[dreamy sparrow]

SO U AGREE

[keen beacon]

I’m gonna make negative benchmarks

Were AI fails on simple tasks?

[dreamy sparrow]

wrong too

[spark python]

4 minutes btw

[slim gorge]

gg

[dreamy sparrow]

gemini in coding is ass

[native yarrow]

gemini 3

[dreamy sparrow]

everything else?

no

[vast fern]

what is the answer

[torn mantle]

holy f sht

[obsidian shell]

Its not as good as people expected.

[keen beacon]

I’ve heard a lot of people say that about Gemini

[torn mantle]

send @native yarrow

send file now

[dreamy sparrow]

lemme get it

[torn mantle]

yes

[keen beacon]

I heard it’s very generic from people on X

[torn mantle]

whats generic

[native yarrow]

see it for yourself

[torn mantle]

opus?

[obsidian shell]

Welcome to X.

[native yarrow]

you have to refresh every time you wanna redo the simulation but

[keen beacon]

I don’t know same with Reddit

[cloud zinc]

yes x is full of experts

[keen beacon]

There seems to be a mixed reaction about Gemini

[keen beacon]

Just like here

[cloud zinc]

just like u are an expert

[vast fern]

tell the answer

[torn mantle]

just little tweaks and its perfect

[keen beacon]

Just as much as u buddy

Maybe lil more

[cloud zinc]

ty

[native yarrow]

yep :P

[keen beacon]

Jk

[native yarrow]

i still feel gemini 3 is better at coding so far but thats js me

[native yarrow]

ima do a side-by-side

[dreamy sparrow]

the answer

[keen beacon]

Are there like any real developers here that like actually?

Know how to code and stuff

[dreamy sparrow]

lemme test gemini WITHOUT tools

[keen beacon]

We need professional opinions

[dreamy sparrow]

In Ai studio

[vast fern]

I thought we all

[torn mantle]

https://x.com/alexalbert__/status/1993033924805902491

Alex Albert (@alexalbert__)

Opus 4.5 "seems to be able to vibe code forever"

I've found this to be very true. Much more to come here but basically you can set-and-forget this model as it works on coding task for you in the background.

Feels like we hit a step change.

X · Nov 24, 2025, 7:08 PM

lmao google

[jade egret]

gemini? Did it get it correct

[keen beacon]

No, we just add to the speculation

We muddy the waters

[vast fern]

i use it for my everyday coding projects

[dreamy sparrow]

ofc

[vast fern]

sonnet 4.5

[drifting crow]

AI is superior at coding than any human

[dreamy sparrow]

lemme show u

[jade egret]

did claude 4.5 opus get it correct?

[keen beacon]

So why doesn’t it fix its own code?

[native yarrow]

naw

[dreamy sparrow]

nope

[jade egret]

oo

[dreamy sparrow]

completely wrong

[drifting crow]

It wants the human to learn

[dreamy sparrow]

lmarena doesn't use tools?

[keen beacon]

I’ll take it

lol

[vast fern]

non thinking vs thinking

[hushed gyro]

https://019ab766-6a03-7b19-89b0-1a7cc5a52b40.arena.site

Kimi k2 is amazing.

1984 inspired newspaper

[dreamy sparrow]

yeah that's correct

[drifting crow]

I’m a non thinking human

[strong cipher]

Opus 4.5 or Gemini 3 pro for coding? I dont use opus 4.5 for now

[dreamy sparrow]

opus

[keen beacon]

You wanna c big brother

[vast fern]

share your benchmarks

[keen beacon]

https://youtu.be/gNvvBWCnw8Y?si=2wq8xJKdOLHD9rTp

That’s in some central Asian country pretty crazy technology

[normal abyss]

when you type to any AI model too fast, it corrups your entire chat (for some people), be sure to type slower between responses

[drifting crow]

Shitposting: 10000000%
Productivity: 0.000000001% (margin of error 000000001)

[keen beacon]

🤣🤣🤣🤣

[cloud zinc]

Kazakhstan

[keen beacon]

Yes

And the program only started in 2018

I guess that’s a while now

[obsidian shell]

For me they found different ones.

[native yarrow]

claude

[keen beacon]

What the hell is that supposed to be Google maps or something?

[inland violet]

opus 4.5 or gemini 3 pro?

[keen beacon]

Oh traffic

[native yarrow]

gemini

[keen beacon]

Nvm

[obsidian shell]

Depends on the bloody task...

[keen beacon]

Damn

[vast fern]

both combined

[keen beacon]

Tf is that

[inland violet]

hide 70kg chicken

[drifting crow]

I mainly just use gpt5 mini for coding stuff

[keen beacon]

Which one made that

[obsidian shell]

None.

[drifting crow]

Sometimes it overengineers so I use gpt4o

[inland violet]

why

[vast fern]

a 3x3 city block

opus

[keen beacon]

Damn that’s nice

[vapid elbow]

image generation is unusable today...

[vast fern]

fr

[keen beacon]

There’s a little bench in a rock there in a little looks like a little Mario pipe or something

[obsidian shell]

Well...

[inland violet]

ok opus 4.5 or sonnet 4.5

[keen beacon]

All right, I’m gonna run my own test

[obsidian shell]

Task?

[inland violet]

hide 70kg chicken

[drifting crow]

I don’t let ai code everything tho normally I just want it to make simple functions Icba to make so I’m prob not the best benchmark as I want it to do the minimum

[dreamy sparrow]

GEMINI WITHOUT TOOLS SOLVED IT

now what

[inland violet]

gemini the best

just like always

[fleet lintel]

i am too stupid to know the real answer. what is the right answer?

[dreamy sparrow]

ofc

[obsidian shell]

Me too, I think its 30.

[inland violet]

its 67

[drifting crow]

I think google is in the best position, they have the researchers from Deepmind and they have their own hardware

[dreamy sparrow]

damn that's OFF

[dreamy sparrow]

really far from the answer

[inland violet]

waiting for opus 4.5 thinking

[fleet lintel]

i am not joking.. i have no clue. what is the right answer?

[dreamy sparrow]

ofc u dont know dude nobody knows it

lel

[fleet lintel]

ohk.. not surprised that gemini 3 is better here

[dreamy sparrow]

gemini and grok are the ONLY

models

that got it

[fiery gull]

I'm using it now, is insane to me is smater that gemini 3.0 pro preview

[dreamy sparrow]

it's not

[fiery gull]

without tools... insane

[dreamy sparrow]

BTW it's preview

[fiery gull]

to me is going better

[dreamy sparrow]

not the actual model yet

[fiery gull]

YEAHHHHHH HAHAHAHA SOO ANXIOUS

will jump 20% of inteligencie

[inland violet]

or just more censorship

[fiery gull]

bruh, I need to be optimistic

[dreamy sparrow]

grok told me how to make drugs

[inland violet]

how

[obsidian shell]

No model is truly that censored anymore.

You just have to ask the right questions.

[dreamy sparrow]

it's easily bypassed

[cloud zinc]

actual model will not be better

[inland violet]

no

how to make

[dreamy sparrow]

oh..

[inland violet]

say to me

pls

[dreamy sparrow]

yeah uhm

a hint is

[hushed gyro]

WTF is this lmao https://019ab76b-1909-78ab-9c9c-e6fdd7954750.arena.site

[inland violet]

its phishing

[dreamy sparrow]

u need some kind of very very rare stone

[inland violet]

like emerald from minecraft

[fiery gull]

don't show in twitter pls

[dreamy sparrow]

or just buy it

[native yarrow]

claude won again i think

:P interesting

[hushed gyro]

I don't even have an account 😭

[native yarrow]

it couldnt do a big bang sim but beats gemini two straight

[dreamy sparrow]

I guess yeah

[hushed gyro]

What is the prompt

[inland violet]

'make dogshit'

[native yarrow]

Model a disease outbreak in a population of 100,000 agents across neighborhoods with varying density, hospitals, and public spaces. Agents should have individual immunity levels, social compliance rates, daily routines, and decision-making about masking/isolation. Simulate how misinformation clusters affect regional spread differently.

[dreamy sparrow]

dogshit

[hushed gyro]

🤔

[dreamy sparrow]

yay

[native yarrow]

its better than whatever this is

the sliders dont even work brah 💔

[hushed gyro]

☠️

[dreamy sparrow]

yes it will

[cloud zinc]

preview one will be general release

this is not experimental model

[dreamy sparrow]

remember gemini 2.5 pro previews

it was removed and replaced with a new no preview model

it was significantly smarter

and better

[cloud zinc]

nah, cuz they gonna release in december

[dreamy sparrow]

yeah

prob they release it early because

they were training it

[cloud zinc]

general release is just a name swap from preview to general release and nothing else.

[dreamy sparrow]

nuh uh

nuh uh

[cloud zinc]

cuz gemini 2.5 pro was behind. google feels confident in gemini 3 pro preview

[dreamy sparrow]

wait lemme ask gemini

[cloud zinc]

lmao

[inland violet]

gta its grand theft auto lol

stupid llms

[keen beacon]

Both failed my safety test

[inland violet]

send me prompt

(for tests)

((really))

[keen beacon]

Hold on let me test it out real quick

[dreamy sparrow]

send the prompt babe 🤑

[balmy mist]

is opus 4.5 better than gemini 3?

[dreamy sparrow]

no

NO

NO

NO

[waxen fern]

Does lmarena still have the verification

[vast fern]

@echo aurora thinking models when

[dreamy sparrow]

haha deep think is better

🤑🤑🤑

[hushed gyro]

Which one is better guys

https://019ab778-ff51-767f-9d35-681ea0f92e99.arena.site

https://019ab778-ff51-7427-8244-fb72338fdc2b.arena.site

[hushed gyro]

OMG we need deep think!!!!

[queen veldt]

Oh niceeeeeee

Thanks gemini

[spare mango]

Gemini now engages positively with the hard-R, when previously it only interacted with the soft-R, not the hard-R.

[primal orbit]

are we going to get opus 4.5 thinking in lmarena?

[vapid elbow]

are you one of those indians that create AI political ragebait?

[spare mango]

As someone that solely uses Gemini, I see this as an absolute win. Less censorship.

I had a feeling from a couple days worth of usage that the model was less censored now, but now I've confirmed it.

[hushed gyro]

No... 😔

[fiery gull]

opus 4.5 never in ranking ? lmarena is paid

have 34 respost is good to put it in rank

[spare mango]

I mean, as long as I don't use the hard-R in a hateful way that is, in which case the model is totally justified in refusing to co-operate with me.

[obsidian shell]

Do you find the latest one a great improvement in coding tasks?

[fiery gull]

the 2

[fiery gull]

never lol

[spare mango]

No idea, I don't use it for coding tasks.

[hushed gyro]

I thought so too 😂

[obsidian shell]

Ah...

[spare mango]

It's far less censored, I can guarantee that, I spend most of my time discussing social issues with it.

So I've seen a stark shift.

[obsidian shell]

Its easier to jailbreak too.

Although you don't really have to jailbreak models anymore.

[fiery gull]

ahh..ahhhh a AI user don't work with code too?? I tought I'm the unique here

🤧

[obsidian shell]

You 2 are quite rare 😂

[fiery gull]

nice to meet you hambuguer

[spare mango]

Basically it's a lot more willing to entertain right-wing perspectives.

[obsidian shell]

Apatite...

[spare mango]

Bro thinks I'm food 💀

[obsidian shell]

Amen...

[spare mango]

I would say it's in the political center now, before it would constantly tug you towards left-wing politics, even if it were subtle, not anymore.

[fiery gull]

is maked from billionarie company, is weird an AI is left, wtf? the company don't want profits?

[obsidian shell]

Sonnet 4.5 was a great improvement on that.

[obsidian shell]

Eh, the training data tend to be left-wing...

[fiery gull]

ahhhhh sure sure

[native yarrow]

GPT was very left wing

[fiery gull]

why? lol

[obsidian shell]

A lot of cry idiots fill the internet with left-wing opinions, like 60% of the userbase, thus the result. (I think).

[native yarrow]

it just very obviously gave pretty biased information based on that kind of agenda, i dont care specificallyt that its left wing but that it gave biased info

i want it to be completely neutral

like, a model

[obsidian shell]

Its very good at sounding right.

[spare mango]

Me too.

[native yarrow]

then a quick research dive of your own and its complete bs lolz

[fiery gull]

the left AI will tell them to release everything for free, and companies will go bankrupt because in the future people will only listen to AI blindly (like me) 😭

[spare mango]

Gemini 2.5 was much more biased than GPT 5.0, but now it's a lot more balanced on Gemini's end.

[native yarrow]

i honestly still dont trust ai for news or info

i always look it up after

💔

hallucinations are a big issue

[obsidian shell]

Critical thinking will leave the chat.

[fiery gull]

do you trust in what?

[obsidian shell]

Trust copilot with your emails 🙂

[native yarrow]

i typically tend to trust more in my own research

lolz

[hushed gyro]

Should I do more news snippets or nah

[native yarrow]

do a WW2 one

[spare mango]

I mean this is insane.

The model is calling the doctors that mindlessly told you to vaccinate yourself, despite seeing side effects from it, "Robot Doctors".

Unimaginable for me just a month ago with Gemini 2.5 Pro.

And called the practice "medical gaslighting".

Feels like I'm talking to Grok, not Gemini.

[obsidian shell]

Dat very true.

[queen veldt]

😂

[obsidian shell]

Where did you find the models?

[queen veldt]

I just edited the image made by claude

[native yarrow]

okay so who do we think is better claude or gemini

[golden ocean]

didnt someone say theres more censorship in gemini 3

[golden ocean]

oh nice then

[spare mango]

not at all, you can check the examples I gave above.

it's supporting the anti-vaxxer narrative, and engaging with the user even if they say the hard-r to the AI.

[golden ocean]

w

[native yarrow]

did they lower the rate limit for nb pro lolz

[spare mango]

sure, it's the fact that it's willing to say that, that I find surprising.

[vapid elbow]

yes

[spare mango]

It shows that it's not being censored top-down.

[native yarrow]

aww man

[vapid elbow]

i think u can only gen 5 pics with nano banana now lol

lame

[native yarrow]

that mustve been the fix for the lag

thats weak asf.,.,

[vapid elbow]

hopefully not permanent

[native yarrow]

i would ping the mod

but

meh

[echo aurora]

Ping mods for moderator related stuff please.

Ping me for questions/feedback/etc.

[native yarrow]

do you know how long the new rate limit will last

[queen veldt]

Someone ran custom bench for the model

[vapid elbow]

Idk, even with the new limit gens take like 120 seconds lol.

[native yarrow]

nah mine was fast

[echo aurora]

I do not, specific rate limits for models can change over time.

[native yarrow]

its just i think thats what they did to fix the server issues