💬│general

[lucid jewel]

The best ai

[weary abyss]

lets go band for band

[white horizon]

Can we use it other than coding?

[young hollow]

Sonnet is Beloved by pplx fame

[white horizon]

https://tenor.com/view/silly-cat-silly-car-car-stare-10-thousand-yard-stare-10-thousand-yard-gif-14200271775968563996

[young hollow]

So it came very fast

[weary abyss]

HOW ABOUT WE TEST?

[spark marsh]

Yes

[dusky lodge]

What use of it

[white horizon]

Ayo fr

[weary abyss]

hey

[white horizon]

Wait

[young hollow]

While GPT5 had took a while

[short narwhal]

4.5 thinking seems to like to mention/use memory and other context quite a lot

[weary abyss]

tell us gang

[young hollow]

Yo

[silent ocean]

For sure but it's specialized for developers and enterprises

[weary abyss]

what you think of 4.5

[silent ocean]

Average response time is..?

[weary abyss]

CLAUDE CLAPPED YALL

[young hollow]

I think nothing yet because i haven't seen any benchmarks or used it but im sure is great

[short narwhal]

Same as 4.1 thinking imo, faster than GPT-5 Thinking though

[spark marsh]

Like other ai you can ask other question but its best user is coding

[silent ocean]

Which is?

I never used it

[vague pagoda]

it's a good model, you can do anything with it, but it's especially good for development related work

[silent ocean]

Like never really found an use case for it

Other than coding

[dusky lodge]

Wow

[weary abyss]

GPT 5 thinking or THAT?

[pseudo ferry]

💯

[short narwhal]

Length always depends on your query🤷 Simple "hi" takes like 5-8 seconds?

[silent ocean]

Gpt-5 thinking is overrated

I'm saying it

[willow adder]

Only i dont have yet???

[silent ocean]

I'm saying it

[weary abyss]

its ass

[pseudo ferry]

https://tenor.com/view/spongebob-patrick-star-noted-notes-gif-17474838830648097856

[silent ocean]

So the "best" option still stands as the best

[weary abyss]

best is the worst

[silent ocean]

Have you used it?

[weary abyss]

like everytime

[weary abyss]

yes
the best option

[silent ocean]

I use it now over gpt-5 and the thinking model

[weary abyss]

bruh

[silent ocean]

It provides faster and more interesting responses

[weary abyss]

it gets everything wrong except obvious questions

[silent ocean]

It uses all models

[weary abyss]

hell nah

[silent ocean]

Does it

[white horizon]

I will use claude tmrw now ima sleep

[short narwhal]

Nightt!

[weary abyss]

aight gang gn

[white horizon]

If someone test it for general use dm me and tell me the results 😭

[umbral canopy]

nice sonnet 4.5 thinking woop woop

[weary abyss]

type shi

[white horizon]

Tho i ran the prompt but its long i will send anyways

[tame current]

What use cases do people choose a specific model instead of just doing auto?

[willow adder]

Is the Thinking mode always for better?

[white horizon]

Its hella long

[white horizon]

Yeah

https://www.perplexity.ai/search/analyze-the-long-term-global-i-XwyjyYWnQOGa0byGziRy7w#0

[weary abyss]

show us homie

[willow adder]

Oh okay Thanks

[silent ocean]

When they're delusional and have too much time on their hands

Or just coding

[silent ocean]

Not really

Overthinking simple questions is stupid

[umbral canopy]

depends on your system instruction. sometimes thinking is just "too much" thinking ^^

[white horizon]

But mostly for heavy tasks it is

[silent ocean]

And you wait longer

[silent ocean]

Water as a drink is good

[white horizon]

Thats on u if ur having simple one just select non thinking

[tame current]

Facts about too much time lol

[willow adder]

i mean for coding

[hoary forge]

PLEASE ADD 4.5 TO SHORTCUT CREATION ON ASSISTANT OF COMET

[umbral canopy]

for coding the answer is yes.

[white horizon]

Yeahh it comes in thinking

[willow adder]

I Gotchu!

[silent ocean]

It takes so much time when you're on mobile and switch models each query and need to think longer

Like

Just ask the questions using best

And it will work it out

[short narwhal]

🤔 Refresh the page?

[white horizon]

If ur on short question ngl sonar is better than all

[silent ocean]

Sonar is an insane model

For online search

[white horizon]

Agreed

[white horizon]

Fr

[hoary forge]

Read what i've typed.

[white horizon]

Even in social

[silent ocean]

I need someone to test out the "Best" option using benchmarks

Compared to just selecting a model

Like Gpt-5

Or GPT-5 thinking

[short narwhal]

I got what you meant. I'm saying that refreshing (CMD + R / Ctrl + R) when in the sidebar refreshes the assistant and updates it to show the new models

[blazing zodiac]

Hi everyone!

[void bone]

is 4.5 good

[ornate salmon]

I tried asking Claude sonnet 4.5 in perplexity what its capabilities / standouts were and it just tells me it’s not Claude sonnet, but instead perplexity AI assistant. Anyone else having this issue?

[dreamy elk]

ive heard of every modells except sonar, what can It do better than others?

[short narwhal]

The model is told that it is Perplexity AI in its system prompt. Rest assured you are using Claude if you selected it, no matter what the model might argue back

[blazing zodiac]

Absolutely banger product guys, well done!
When we open the assistant, can we change the agent behind? (ChatGPT/Claude/...)

[silent ocean]

Insane browsing speeds

Like it can read 20 sites in under 3 seconds

[dreamy elk]

damn- is it accurate?

[tame current]

already added claude new versions....

thats so good

[dense python]

What are the claude models are good for?

[hot mantle]

Yeah !!!

[tame current]

https://cdn.discordapp.com/attachments/1383179981139480737/1398109865745842197/breadcat.gif

[short narwhal]

Comet assistant or Perplexity in general? No to Comet assistant (The one doing the clicking on different sites)

[willow adder]

What is the Lab environment for if i could create everything in the normal mode?

[cold pumice]

ambos são perplexity

[wise nova]

What gives us Sonnet 4.5 Thinking?

[limber citrus]

what are the limitations now ? how many tokens for each? we always have think abt this before celebrating

[short narwhal]

I don't see why it would be any different this time? Same 300+ uses per day as always

[hot mantle]

https://www.perplexity.ai/page/anthropic-claims-claude-sonnet-eOHXJjMrTYaMn5S6NhPENQ

[pale viper]

Hu

[deep spear]

.

[sacred cove]

?

[hoary forge]

I tried but doesn't work.. This is separate list of models, not the same as the one from main input box

Btw they added it now

[timber niche]

@short narwhal 4.5 thinking not showing for me

Wbu

[short narwhal]

It’s there for me

[hoary forge]

Close and open

[broken oracle]

On the website?

If so simply refresh

[novel magnet]

What context size do we have for new claude models in Perplexity?

[timber niche]

App

[timber niche]

Web or app

[tame current]

maybe there will be an update??

for app users

[timber niche]

Likely

[broken oracle]

Restarting it should also make it available

[tame current]

cz it's live on website

[broken oracle]

It's already in the app

[timber niche]

I have the model

[tame current]

cool

[timber niche]

Thinking doesn't show

That's the problem

I phrased that badly ig

[tame current]

[timber niche]

It likely is thinking but maybe they haven't adapted it fully yet

To show the reasoning

[tame current]

voltrex person has the thinking model on their app though

[timber niche]

I have the model

When i ask it something

[tame current]

ohhh

[timber niche]

It shows only the answer

[tame current]

ohhh

[timber niche]

When i click perplexity pro to see the reasoning

I see nothing

[tame current]

it is not working u say

[broken oracle]

Oh you mean it's not actually "thinking"

[mighty breach]

Why did they extract benefits from the Pro tier and created a new MAX tier and put them there for $2K a year??? Whose honestly paying for the Max Tier???

[tame current]

let me try

give a question lmao

[timber niche]

There are people who pay 2400 a year

That's why

[tame current]

gpt ahem...

[mighty breach]

That's bait and switch bro... complete frauds bro

[timber niche]

Lol

Essentially

They introduce max and limit pro users

But this didn't start it

[limber citrus]

how many tokens now thats more expensive ? gotta wait until how butchered the version for us would be.

[timber niche]

Open AI

Did

Web or App?

[hot mantle]

Dunno

[broken oracle]

It's actually doing what it's supposed to, in both web and app

[timber niche]

Damn man

[tame current]

[timber niche]

Doesn't work for me

Maybe i need to restart again or update

[broken oracle]

It's most likely that you're just asking it something where it doesn't deem necessary to reason/go into thinking mode

[timber niche]

There's no pplx updates

[tame current]

maybe your prompt isn't worth thinking and is a fact lol

[timber niche]

Still doesn't answer why it's not showing thinking

It's thinking api

Not a model like got5

Gpt5

That decides

Whether it thinks or not

Maybe y'all are right lol

[broken oracle]

I've seen with several other models where it doesn't show that when the prompt is not deemed enough for the thinking mode, so that's likely not it

[timber niche]

Yeah it's new to me

[timber niche]

Got it

I asked it something that needed light thinking uhm uhm

Just some light reading nothing much lol

[tame current]

it is not thinking type question tbh

[timber niche]

Fair enough

[tame current]

rather a search type question

[timber niche]

It showed me COT

[tame current]

to surf the fact from internet

[timber niche]

Which is what i cared

[broken oracle]

To be fair Sonnet 4.5 isn't something particularly impressive either, it's one of those "everyday tasks" model, not for complex tasks or anything that requires some amount of actual research

[timber niche]

It's not opus but it's quite good

[tame current]

do some fictional comparisions to test them

[timber niche]

Do it boss

And tell us

The results

[tame current]

i do it LMAO but college projects are cutting me into pieces

[timber niche]

It's very quick

Less than 10s

Most times so far

[timber niche]

Goodluck dude

[timber niche]

DAMN

[broken oracle]

Where is this benchmark from?

[heavy portal]

My understanding is that when "best" is selected as the model it'll dynamically switch based on the type of prompt is entered. Is there a way to see for one's own account which models are chosen most often, like statistics?

[timber niche]

Comet with 4.5 will go crazy

[slow osprey]

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

[timber niche]

Best means sonar

Most cases

Choose your model

[slow osprey]

best means never the best

[heavy portal]

sonar in most cases, but still would be interesting to see the numbers

[broken oracle]

Yeah... I don't particularly trust the companies' own benchmarks, the tests are often skewed and are particularly geared towards their own creation which unfairly boosts itself in the benchmarks

[timber niche]

They obfuscate it on purpose

So normies don't know lol

[broken oracle]

This is a very common thing in the industry

[timber niche]

Lol gpt5

For example

[left stream]

That’s why test it yourself

[slow osprey]

try it yourself, compare it to gpt5.

[timber niche]

Hi Joe

[broken oracle]

Of course, gonna be pitting it against others, other people will also definitely be doing it too

It's great if it is to be believed, let's give the third-party testers time to cook

[left stream]

We have to put claude 4.5 sonnet against the Devil

[wild frost]

Bro is That true perplexity pro which we got from airtel, and the perplexity pro that we buy premium of that, both are different or same, even though i use both of them, i have seen that perplexity pro that Be buy much better than airtel perplexity pro, it's that true or just humours about that

[left stream]

Gpt-5 pro vs gpt-5 thinking vs deepseek v3.2 vs claude 4.1 opus thinking vs claude 4.5 sonnet thinking

[broken oracle]

Pretty sure it's just rumors

[quaint ravine]

what's the summary for 4.5 Thinking?

I will test my use cases soon

[wild frost]

[left stream]

How about enterprise

[wild frost]

Enterprise is mainly for orgs/teams bro, same model but higher limits, security, admin stuff. Not much diff for solo use.

[left stream]

No clue yet

So far I heard it’s insanely good for coding

[left stream]

E

[limber citrus]

who?

[left stream]

Oh well at least it still has benefits

[quaint ravine]

crazy the non-thinking is already huge improvement at PDF analysis so fa

can't wait to test thinking

[left stream]

GPT-5 Pro is the Devil

Oh shoot both answered

[wise nova]

How much pages in pdf Perplexity can read in Spaces storage sources? What if my pdf contained 150 pages

[wild frost]

Yeah bro 💯 it’s got benefits — mainly scale + support, good for teams.

[timber niche]

False news

Enterprise doesn't train on your data

You get soc2 level security

[wild frost]

Bro not totally false, i am using both versions of that, literally I see the difference airtel perplexity pro and purchase perplexity pro

[timber niche]

And ofc org level controls

[timber niche]

Like what

[left stream]

Oh shit

Didn’t think pro to win

[timber niche]

Hm

[left stream]

Do a test

I’m on creative writing section right now

[gleaming plume]

I have a question: is GPT 5 thinking inferior to 4.5 in everyday tasks other than programming?

[left stream]

I’m doing perplexity 4.5, 4.5 from api, and gpt5 pro rn

[timber niche]

Nice

[left stream]

Yes but it doesn’t make sense

[timber niche]

Depends on use case plus 4.5 is too new to determine

[left stream]

Gpt-5 pro beating claude sonnet 4.5 in creative writing???

[timber niche]

Isn't oai models always best in creative writing

[wild frost]

Basically dude i am using this on my aptitude or coding senerio, where i tested both, but purchased perplexity pro gives more accuracy as compared to airtel perplexity pro

[left stream]

Tbh that isn’t my field, but I liked using all of them together

[onyx tree]

Question when I see you guys post in prompt library, do you copy paste that into the special instructions?

[timber niche]

That's not a valid explanation you should email [e-mail]

[vague grotto]

The Perplexity and I decided to host a lyrical evening. I write my poems for it, and it writes its own for me, or perhaps as a competition—I start, and it continues. It's so funny.

It completely lacks rhyme. 😂

Maybe we should write new algorithms for it so it can also be a poet?

That would be so wonderful. 😊

[onyx tree]

Or at the beginning of a thread/conversation?\

[timber niche]

Oh wow

[wild frost]

U say both are the same, have u ever tried airtel perplexity pro?

[left stream]

I’m gonna switch it to explaining concepts and then up it to different levels of explanation from elementary to phd

[timber niche]

It should not be different man

You should email support

[onyx tree]

Question when I see you guys post in prompt library, do you copy paste that into the special instructions? Or is on the beginning of the thread/conversation?

[timber niche]

Now that's up your alley

[timber niche]

Special instructions and in spaces

[onyx tree]

Oh ok thanks

For some reason when I do it on my phone it says theres a 1500 character limit

[left stream]

Yes, but I had high hopes for claude ngl

[timber niche]

Is that bench squat deadlift stats i see

Lol

[left stream]

I refuse pro to steal their thunder

😔

[timber niche]

Gpt5 pro is just built on o3 pro and o1pro

It's nuts lol

Maybe 4.5 opus will pose a battle

Worthy

[left stream]

Parallel test compute is crazy

Heard 4.5 sonnet can do it too supposedly

Will have to confirm

[distant aurora]

Claude should bring back haiku.

[timber niche]

Lol

[left stream]

Haven’t heard that in a while

[timber niche]

4.5 woke up billy

[distant aurora]

Low cost reasoning models are more important in the long run.

[timber niche]

Most support ai bots sre dogshit

[left stream]

2nd test

[distant aurora]

Gpt 5 mini and nano series, grok 4 fast, and Gemini lite.

[quaint ravine]

okay not much difference in the end to Claude 4

still very solid in general purpose complex scientific/medical queries, but nothing noticeably different...although very hard to improve on that since all the best models are generally excellent already (especially with custom prompting etc) and combined with PPLX's generally excellent search

still fails quite badly with a lot of PDF interrogation. GPT-5 Thinking superior still

hope you coding people notice some improvements!

[green crystal]

a base model can do parallel reasoning?

dafuq

[thin lake]

GUYS! Check this out C:\Users\Eric\Downloads\IMG_4938.PNG

[umbral canopy]

valid link.

[umbral canopy]

whats the folder C:\Users\Eric\Downloads\chicken_feet for?

[twilit crown]

Does bell give free perplexity pro?

[left stream]

Okay ngl I might write a book with 4.5 @green crystal

Imagine 4.5, 4.1, and gpt-5 pro all working together

[green crystal]

it can output that much text?

lol

i see there's sub agents etc

guess we will see that soon in chatgpt aswell

[distant aurora]

https://youtube.com/shorts/sXSatkz3Kqc

[green crystal]

part of the "compute intensive" features

[distant aurora]

@young hollow

[left stream]

I won’t lie on the second question while testing all of them

I got distracted with 4.5 since it made a good hook for a story

[green crystal]

lol

i know how they could do it too

probably chatgpt agent using other models through chatgpt site

and basically prompting for you

[willow adder]

How can i upload ZIP files (If its possible)

[umbral canopy]

claude 4.5 gave me a casual 49559 character output with 8211 words. not bad.

[left stream]

You think they might release new toys with 4.5 making its debut today?

[green crystal]

yeah they're not gonna let their fresh grip with codex just wither away

we wont see something tomorrow

but i bet not far off from now

if claude is really that much of a hit then we might get a teaser soon lol

[left stream]

Deepseek v3.2, claude 4.5, now what’s next….

Gemini????

👀

Imagine gemini 3 pro

[tough knot]

Btw guys, is that a problem if i dont have a Claude Sonnet 4.5 access via Pro subscription?

[dry kestrel]

Try clearing your cache

[left stream]

Well fu**

gpt 5 pro won again

I deadass thought perplexity 4.5 was gonna win

I liked the hook ngl

[dry kestrel]

what the

[left stream]

It was good asf

[umbral canopy]

clear your browser cache

[tough knot]

Hmmmmm, i'll try it now but.. i can't find it on the PC app and on the web-version also

[dry kestrel]

Update your PC app and clear your browser cache

[tough knot]

🥲

oke

[left stream]

Gpt 5 pro uses gpt 4.5 too?

[dry kestrel]

no

[tough knot]

Yeah, about the update pc app. Is there any special button in the app or i have to download new version on the official site?

[dry kestrel]

I'm not too sure, I don't have the PC app. If you can't find any update button just reinstall it

[quaint ravine]

should be at least 2-3 months away apparently

[wispy rover]

Goodnight 🙂 I am a small entrepreneur who has been starting out with the world of AI for some time I am looking for equipment to formalize my startup I would like to know new ideas 💡

[quaint ravine]

but it's something I am most looking forward to

[left stream]

Crazy because I didn’t think 5 pro was also strong in creative writing. 4.5 should’ve won at least one prompt wtf

[quaint ravine]

Gemini 2.5 Pro was my go to, and in some ways is superior to GPT-5 Thinking even now (answering style/structure is better imo)

[bleak lichen]

Has anyone tested the standard and thinking models of Claude 4.5 against GPT-5?

[quaint ravine]

recall Gemini 2.5 Pro came out over half a year ago now

[left stream]

I hope it’s not a failure

[quaint ravine]

in my use cases, GPT-5 Thinking is superior still

but you need to define use cases

I will test more though...don't want sss lol (small sample size lol)

[left stream]

@dry kestrel you saw jensen huang’s deleted twitter post?

Either sora 3 or he has a hot model in his room that he’s gonna be with all night

[dry kestrel]

yep

Hopefully sora 3

[left stream]

I’ll be having fun

[bleak lichen]

I use it for coding questions, basically to find the latest info online and present it in the best format it can, just like I wrote in the instructions.

[quaint ravine]

ah nice

apparently it's best for coding now?

altho that's what they said with all previous releases

and basically no one agreed here lol

also if it's superior to Opus 4.1

[tough knot]

Yooo, let's gooooo. For PC users - you have to deinstall your app in the window's settings (win+I - APP - deinstall perplexity)

[quaint ravine]

wats the point of Opus 4.1 now?

I guess Opus 4.5 will be out within the next month or so perhaps

[dry kestrel]

Where are you seeing that?

[quaint ravine]

its own release page?

well at least for software engineering lol

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

[dry kestrel]

I'd wait till 3rd party benchmarks come out - also misread that

[quaint ravine]

also for Finance etc

Experts in finance, law, medicine, and STEM found Sonnet 4.5 shows dramatically better domain-specific knowledge and reasoning compared to older models, including Opus 4.1.

yea but if Claude itself saying it's better than Opus 4.1 seems odd if it's wrong

wouldn't fit with PPLX's economical model too imo with Opus 4.1 only for Max

[twilit crown]

What are the cools things to do after getting pro subscription?

[jolly dragon]

Comet is OP btw. A bit difficult to adapt, but it’s great.

[tribal vessel]

Hey

Anyone wanna join the vc?

I want 2 talk 2 sumone im bored lol

Ill wait 🙂

[hot mantle]

There is no way to start a lab project from an old one? The context is too small, after 5-6 generation code, it keeps hallucinating

[left stream]

Work out to the point every angle looks fucking good when taking a selfie of yourself

[tribal vessel]

Im working on it rn

[left stream]

Good

[tribal vessel]

Just started working out 3 months ago, in 14 rn

[left stream]

Every day stacks

[tribal vessel]

Ty

[mellow pivot]

man i was really excited by the new search API - but it is not very good compared to what the ppxl UI returns

[tribal vessel]

Yeah

I've been seeing a lot of growth tbh

[left stream]

You’re 14 don’t sweat it

[tribal vessel]

Yea ig

[left stream]

I’m 22 and you got a lot to work with

But be consistent

[tribal vessel]

Yup

[left stream]

You’ll love yourself at 18 trust

Mb hard to type with one hand rn

[tribal vessel]

Working out everyday except for Thursday and sometimes Tuesday

Np

[left stream]

The good thing cardio is daily

[tribal vessel]

...

Uhh

Well I bike 3miles to the gym and back

So 6 miles overalleveryday

U think that's enough?

[left stream]

Run it

Unless mountain biking

Actually forget that you 14

[tribal vessel]

Ig its lot of hills

That's wt im doing rn lol

U drive?

[left stream]

I can drive

[tribal vessel]

Nice

Do u have a car tho

[left stream]

Yes but it tempts you badly

[left stream]

Yes

[tribal vessel]

Yea

Nice

[left stream]

Car wont get you leaner

[tribal vessel]

Wt car?

Yea but im js asking lol

[left stream]

2016 nissan altima se

St

Sr

[tribal vessel]

Nice

First car?

[umbral canopy]

sestsr is a long version code.

[left stream]

No

[tribal vessel]

Oh

How many cars have you had

[left stream]

Mine couldn’t handle my energy

New years was fun tho

[left stream]

3???

I think 3

Yeah 3

[tribal vessel]

Damn

[left stream]

Motorcycle wont count it

[tribal vessel]

And your only 22

That's rlly good

[left stream]

Im a dumbass tho

Dont be me

[umbral canopy]

could mean he totalled two cars 😄

[tribal vessel]

My brother's 21 and he doesn't have a car yet

[left stream]

Bingo lol

[tribal vessel]

But i cant blame him he started driving when he was 19

[left stream]

Lol

Oh

Thought you meant me

[tribal vessel]

And he's still searching 4 a job rn

Oh

Hbu?

Do u have a job?

U prob do wt am i asking lol

[left stream]

Yes

But im working rn

Out

Seeat

This is hard asf on my hand but fuck it

[tribal vessel]

Lol

[tribal vessel]

Where u work?

Feel free to not answer lol

[umbral canopy]

he'll find one. he can get a cheap car if the need is there. live within your limits.

[tribal vessel]

Wt that

[left stream]

You love interrogating me

[tribal vessel]

...

It's not like that

Js asking

I like getting to know people

It's just interesting lol

[left stream]

Christ my hand

[patent ridge]

hello , how do i learn prompting , does anybody have some reccomandation?

[tribal vessel]

Lmao u good??

[umbral canopy]

every parasocial hacker does that 😄

[tribal vessel]

I took an online course in the summer

[left stream]

Gri

Gro

Gro

Grip

Training

[minor shard]

Yay for new sonnet model!

[tribal vessel]

Oh

[umbral canopy]

its good - rhyming in other languages outside english is still ass

[hushed summit]

You keep asking till you get closer of what you're trying to solve. overtime you'll realize how to ask better questions per model

imo

[left stream]

The pump baby

[left stream]

Part time it business

[foggy horizon]

.

[left stream]

😭😭😭

[tribal vessel]

Oh nice

Anyone wanna get on vc?

[minor shard]

Seems like they pretty much just caught up to GPT 5 and released all the nice features like extensions in IDE etc

[tribal vessel]

?

[twilit crown]

How does Claude sonnet 4.5 compare to Claude Opus 4.1?

[minor shard]

This is their benchmarks

[left stream]

Indeed you’ll love it in the future one day

Well time for my run peace

[tribal vessel]

I hope

[quaint ravine]

interesting that 4.5 Thinking working fine lol

guess more people using 4.5 non thinking atm

[distant aurora]

@dry kestrel https://fxtwitter.com/cannn064/status/1972349985405681686

can (@cannn064)

🚨 FIRST GEMINI 3.0 PRO OUTPUT 👀
︀︀
︀︀Finally we are seeing the much expected model in AI Studio as AB testing, and it seems to be better than any previous model. Yes, better than Zenith.
︀︀
︀︀Prompt: Create code for an SVG of a pelican riding a bicycle as nicely as you can

**💬 45 🔁 39 ❤️ 790 👁️ 198.2K **

FxTwitter · Sep 28, 2025, 5:17 PM

[dry kestrel]

Yeah looks promising

[gray prawn]

Man you gotta think Google has surely gotta be cooking with 3.0

[quaint ravine]

well it's been like half a year since 2.5 Pro release

sorta quite telling they haven't released yet...probably only incremental improvement

[short narwhal]

How long has it been since the last Haiku model

[tame trail]

in which countries does sonnet 4.5 released because I cannot see it

[dry kestrel]

please clear your browser cache and update/refresh the app to see Sonnet 4.5

[dry kestrel]

forever

i want deepseek r2

and k2 thinking

[stable lava]

Good evening, my friends

[distant aurora]

@dry kestrel @green crystal my account got nerfed again.

[dry kestrel]

contact support

[left stream]

Cheist

Christ

[left stream]

I didnt expect to do double of what I do

[gloomy cipher]

@dry kestrel BERD

[left stream]

But tick it

Tuck

Fuck

[left stream]

There is no hope

That depends on you

You know you can do it

But whether you decide to do it today or not lies on you

[dry kestrel]

Hi

[left stream]

Not god or anyone else

But you

[dry kestrel]

looking fire

[tribal vessel]

Yeah

[gloomy cipher]

i know right iam gonna be applying for an art school with this

[hot mantle]

@dry kestrel could you help?

[tranquil orchid]

2. No advertising.

No DMs promoting your products, services, or servers. You may post your own project in ⁠#😎│cool-projects only if it adds value and follows the no-spam rule.

[jagged laurel]

hi

[dark mist]

I wonder if Perplexity has the AI watching for ad spams and when it sees them auto blocks here. How cool would that be to keep us from unwanted advertisin?

[dry kestrel]

I'm not too sure if this is possible, you might want to post in #1111765634267742408

[raven meteor]

How to use different models on perplexity app?

[candid marsh]

at the bottom

[tall rain]

Les goo sonnet 4.5

[raven meteor]

I can't find it 😔

This is what I have

[candid marsh]

click the search button

[raven meteor]

Ohh I got it

Thanks bud

[candid marsh]

you're welcome

[tall rain]

3/15 on the api is crazy tho

[desert fractal]

Hey

[tawny silo]

hlw guys

[desert fractal]

What would the strengths of gpt 5 and Claude sonnet be? Against each other

[tawny silo]

.

[drowsy harness]

they are available to your account. Maybe check the models with the new tag

[hot mantle]

@ok, it's something really needed

[robust junco]

Hey did you see note is that true the are removing o3

Are they going to replace it with something different or nobody knows

[ionic stone]

Is Sonnet 4.5 any good outside of coding?

[dry kestrel]

It's a good overall model

Clear cache

[dry kestrel]

Yes, its worse than GPT5 Thinking

[robust junco]

You think it will be replaced with something different

[dry kestrel]

yeah GPT-5 thinking is meant to replace o3

[lapis hamlet]

is grok4 the goat of searching?

[tranquil orchid]

2. No advertising.

No DMs promoting your products, services, or servers. You may post your own project in ⁠#😎│cool-projects only if it adds value and follows the no-spam rule.

[barren plaza]

Does perplexity give a free subscription to the students?

I think I have seen a pro for students kind of thing

What was it then?

[prime bronze]

Yes 12 month free

I tested it today and slowly will move from chrome

I’m thinking is there a way to use any api to build things powered by comet ?

[barren plaza]

Well the pro plan's not activated yet. Didn't get any update when I applied for it...any idea how I can get to know about this matter...

[prime bronze]

I got it when I verified my student email

[lapis schooner]

I just subscribed to Perplexity Education Pro. They told me billing 5 usd in monthly basis after trial ends but in manage plan, i see 20 USD on my next charge. My student status already verified. Is anyone know is it normal or ?

[vapid pewter]

https://youtu.be/Ta6HAADhs9g?si=B6LLkj6VT0lguLU2

https://youtu.be/2YTE2CL7feg?si=ue1ngdgjm8cr-zFo

[toxic dagger]

How does GLM4.5 upgrade to GLM4.5 or above?

[wild crystal]

what's Claude sonnet 4.5 and how does it work?

[candid crow]

Fucking xbow

[barren plaza]

Same case for me too. Any solution to this case?

[half ember]

Is there any CLI tools to use perplexity?

It's growing more and more on me.

[toxic dagger]

Who know about this? I see benchmark is Ok.

[candid crow]

Tho I'm more into hardware then software

[half ember]

It's anthropic model, claude models usally are very good at many tasks, including coding and writing.
Opus is usually considered better as it's more powerful, but it's locked behind max subscription here.

Benchmarks aren't really saying much, and model came out today, so it will hard to say somethimg for sure until it will be tested extensively.

But so far looks quite good for a sonnet, I didn't liked sonnet 4 compared to 3.7, but this one seems better.
Will not judge it until use it for a long time in different context though.

[young wadi]

where is 4.5

[half ember]

Clear the cache and such.

If in browser.
Or update/refresh the app.

[young wadi]

I tried, but still no

[half ember]

Could you try to open it in new browser, and/or in incognito mode and similar?

[candid crow]

Logout

And login

[half ember]

Yeah, could be that as well.

But I didn't need to log out.

[young wadi]

Still not working

[half ember]

Did you tried that? Because it should be a fresh load with that.

If even that wouldn't work, it would be strange and probably some problem on perplexity side then.

[young wadi]

I think it may not have been pushed to my account yet, maybe it will be there tomorrow

[half ember]

May be.

[young wadi]

Thanks for your help

[half ember]

But I think they should've pushed update in general.

And switching all endpoints on that.

Doesn't make much sense doing it partially, even if they aren't the hosting, imo.

[dry kestrel]

It should be out for everyone.

[dry kestrel]

Try just using the perplexity website for now

[final echo]

Even on Website, claude 4.5 has disappeared

[fluid kernel]

I had it earlier in the afternoon but it seems to have gone back to 4.0 in the evening. Signing in and out didn't help. Sounds like a bug/outage.

[dry kestrel]

Might be a temporary outage

[final echo]

Did someone raise issue?

[mental swallow]

in my phone app 4.5 is available, but in browser/comet still 4.0

[half ember]

Hm, I opened my desktop app, and it also shows sonnet 4, but I did not updated yet and use linux.
Could also guess that sonnet 4 in options is already just rerouted to the 4.5 endpoints, so it wouldn't make much of a difference.

[final echo]

It's fixed now

[half ember]

Updated, there is 4.5 models, yeah.
But again, I'm on linux.

[left mountain]

https://tenor.com/view/xx-gif-15660269

[left stream]

@dry kestrel if you want to feel who actually writes the best

Ask 4.1 opus, 4.5 sonnet, and gpt 5 pro to do this. "Tell me a sad romantical story"

[flat cosmos]

4.5 already?

That was fast

lets test it

[modern sinew]

Any doctors here? If yes then run for your life

[languid star]

Any1 tested 4.5 on pplx?

[quaint ravine]

yes

[modern sinew]

How is it?

[quaint ravine]

solid

but can't say it's clearly better than 4.0

just my use cases though..not sure about stuff like coding, writing, or being a friend etc

[left stream]

That's why humans exist, to be a friend.

[half ember]

The price is too steep.

[sharp cedar]

gm

new here

[winged sundial]

Guys

Don't yall think they should make perplexity more friendly

Like talking one to one

Like gpt

[errant bloom]

PPLX become human

[winged sundial]

Huh

Like robot or that talking friendly?

[zenith haven]

Well how to get gemini for 1 week free ?

[errant bloom]

https://tenor.com/view/connor-detroit-become-human-wink-i-got-you-gif-17843288

[zenith haven]

Now when I chat its asking me to upgrade to a plan
I dont see any mode feature to switch models

[winged sundial]

Dayum

[zinc yew]

honestly

i like the way it talks

call it dystopian or whatever but they're all artificial at the end of the day

i'd want something that seems artificial and not trying to act like a human to make me feel more connected

[young hollow]

Lol wtf

[manic tree]

[white horizon]

[white horizon]

?

[pseudo ferry]

https://tenor.com/view/alanharper-alan-harper-charlieharper-charlie-harper-dva-a-pul-chlapa-gif-16034252947757163993

[white horizon]

https://tenor.com/view/fat-cat-gif-12535219569916731430

[thin lake]

COST COH

[vagrant badge]

https://discord.com/channels/1047197230748151888/1422473542024433664

[crude siren]

Hello

[pseudo ferry]

Hello!

[crude siren]

Nice to meet you.

[crude siren]

I wanna collabrate.

[crude siren]

Hello

[pseudo ferry]

https://tenor.com/view/harvey-specter-suits-suits-tv-suits-usa-pilot-gif-16445429241371648002

[crude siren]

Where are your from?

[barren quarry]

How do I get the pro role

I have pplx pro yearly

@vague pagoda

[young hollow]

Lol wtf

[coral ermine]

Can someone tell me when the comet assistant for Android is gonna launch ?

[lucid jewel]

[languid star]

#💬│general message

[barren quarry]

Thanks homie

[wanton sundial]

[barren quarry]

Done

[echo harbor]

Hi

[barren quarry]

Wsp

[echo harbor]

What?

[barren quarry]

What's up

[echo harbor]

Yeah good, What about you?

[barren quarry]

I am ok

[ruby marsh]

Hi guys, how do I train an AI to make scripts in Lua using the program's API?

[odd dust]

hi does any one know where i need or can share an idea that would benefit the user experience and perplexity as well ? 😁

[next niche]

Yo guys

[left stream]

Ask ai, it’s one heck of a teacher

I stopped relying on youtube videos a long time ago (almost fully, there are those golden ones)

[ruby marsh]

I don't need a teacher)
I need an AI that I will train to write scripts in Lua.

[noble citrus]

Hm

[left stream]

Didn’t you ask how to train an ai?

[ruby marsh]

yeah, but don't need a teacher
I will be a teacher for AI

[ripe tulip]

can't be a teacher for AI if you don't know how to train AI

so firstly make AI your teacher

[white horizon]

Whats going on

[left stream]

I’m in a good mood so I’ll have someone answer that

[white horizon]

[tame current]

can anybody tell me some projects to make?

[ruby marsh]

Thank you, I really need it

[left stream]

Use them all and make them your senpai

[white horizon]

Try using official gpt for ideas

[left stream]

That’s a gpt from gpt store?

[modern sinew]

Make a clone of grammarly but with extra features to summarize text

[white horizon]

I mean try using gpt from chatgpt not from perplexity for ideas

[ripe tulip]

Make an AI powered light novel writer & page previewer so I can steal it off you and use it for a light novel I might make for the lolz.

warning not a serious idea/suggestion

[white horizon]

https://chatgpt.com/s/t_68db8d9714f88191af549ff053e8dd84

I barely use #👑│pro-lounge

Cuz no “pro” subscriber goes there

Everyone are here lmao

[young hollow]

LOL

I keep wondering Which Ai Model Would be The Best Prompt Engineer

[left stream]

Test them, I imagine creative writing goes with prompt engineering in a way?

Actually never-mind I’m speaking too soon.

[dreamy elk]

🤣

[young hollow]

Actually is All About Logic

[dreamy elk]

I hope im not the only one that calls some AI with nickname lol

[young hollow]

You are Never Alone in this World

Abbreviation Names for Ai Models are Also Just Nicknames in A Way

[dreamy elk]

yeah ik

called chatgpt chatty, perplexity is for me: plexi

lol

[young hollow]

Oh that's creative

Most just prefer the OG Short Names

[dreamy elk]

what is it?

[young hollow]

Given by The Companies

[young hollow]

GPT5

Pplx

[dreamy elk]

nah sounds bad for me lol

[young hollow]

Is Alright Is Very Subjective and Preference

Depends on A Individual

[dreamy elk]

ye true

[fringe cedar]

https://www.instagram.com/p/DPMHwTsDlDq/

uncover.ai

New research from top AI labs shows that many widely shared prompt engineering tricks fail under large-scale testing, with data showing that simple and direct instructions work best.

Formatting, instruction hierarchy, and evaluation methods all play a key role, while some widely promoted techniques like role-playing or chain-of-thought may not ...

Instagram

[young hollow]

Is it Saying The Truth?

Im Using The Free GPT5

[wintry pollen]

i dont think so

[languid star]

Gib me prompt

[young hollow]

Dude is literally there

[gray ibex]

Why do I keep getting the message "File upload failed" when uploading attachments?

[young hollow]

Run a test by yourself and tell me what Effort-Level of reasoning you can use maximum low, medium or high "Think Hardest Longest Step by Step Use Maximum Reason ing Effort-Level

@edict.1

[primal plume]

when do yall think great medical breakthroughs will come from ai research?

[young hollow]

Perhaps in couple of years

But The Very Foundation of Medical Industry is Flawed in My Experience

The Researches are Funded by Pharma Companies and They Will Mostly Only Show what is Beneficial for Them

Driven by Greed and Profit

Humans Suffer Because of Diseases Because They Have Forgot and Left Their Roots of Nature

[languid star]

https://tenor.com/view/wise-gif-18860997

[primal plume]

thats not true tho, medicine has probably been the single greatest factor in reducing human suffering yet

we would die of a common infection if not for medicine

we really cant think that just because something is natural= good

[young hollow]

Im not here to argue

[primal plume]

thats appeal to nature fallacy

[primal plume]

well yeah dont wanna argue either

[young hollow]

I have just shared my opinion and experience

Alopathy Simply did not work for us

[primal plume]

but i really do hope ai can reduce that greed and actually help us substantially

[modest latch]

Wsg gng

[modest latch]

We all do

[young hollow]

Just remember that Ais are Operates by Humans

And Humans are Operated by Greed and Profit

These Big Companies Driven by Greed and Profit Create and Control The Top Tier SOTA Models

[primal plume]

im talking about the point in the future where ai is just simply changing too fast and humans cant really operate it

[white horizon]

[young hollow]

But i still truly believe Ai is The Last amd Greatest Invention of Humanity

Solving The Intelligence Problem

[primal plume]

yeah

it really is

[young hollow]

Greate Time to be Alive and Witness tbh

[primal plume]

creating something better than us is logically the most important thing we could do

[young hollow]

Future Generation Won't Really Feel is Excitement

Of New and Better Ai Models Launching Waiting for it

[primal plume]

yeah because this will be normal for them

but we are in the transition period

[young hollow]

Exactly

Indeed

[primal plume]

seeing shit happen every week

even every couple of days

[young hollow]

Yeah is too much

One day it will be very overwhelming to even just track it

Ai Models will Launch Like Websites and Software Do

[primal plume]

and thats when human greed and profit motive wont be able to control it

and im just hoping itll be on our side

[young hollow]

Idk about that tbh

The System is Always in The Hands of Those Human Companies

If they feel is out of control they simply pause it

And fix it

But no Matter i always hope, wish and pray the best of all worlds for all humanity

And My Faith is not in Humanity but Rather The GOD Himself

[primal plume]

but human curiosity transcends greed, im sure these tech companies will be very curious as to what will happen if they just let it progress even if its not exactly in their control

[crude siren]

Hello

[young hollow]

Perhaps Hopefully

[primal plume]

🤞

[young hollow]

🤞🏻

I wonder how even GPT5 Mini High has 64 on The Artificial Analysis Intelligence Index Benchmarks

[last lintel]

yeah so many small large models appearing everyday

[young hollow]

Tbh i doubt GPT5 Mini is that Mini

Doesn't Feel Very Mini to Me

It has 1 More Point then Grok 4 itself

[sharp barn]

when using 'best' ai model on perplexity, how can I tell which model was used?

[young hollow]

And 3 More then Sonnet 4.5 Thinking

[young hollow]

You Can't

So Use Manually

The Reasoning Ones are Best

[sharp barn]

I added an instruction for it to tell me and it actually works

[young hollow]

Sonar for Quick Searching

[last lintel]

sometimes best model works so good sometimes its trash

[young hollow]

I know about that but is not always accurate and reliable

[last lintel]

i would say maybe it plays some roulette on which model to run

[young hollow]

The Models don't know themselves which models they are

That's The Ai Models Identify Crisis

[sharp barn]

Oh, I see

[young hollow]

We Currently Have Sonnet 4.5 not 3.5

[young hollow]

It mostly just uses sonar

Not reliable for smart users

[sharp barn]

I wish they would show it like they show when you generate images. It's important to know what model it is in order to compare

[last lintel]

you are in delulu when u say it to use claude it might not use but u feel

[sharp barn]

Lol I was just testing it now but ok

[young hollow]

The Deception of Being Truthful

[hardy steppe]

Opera Neon just launched and it's paid

[young hollow]

We don't care

[modern sinew]

Opera is annoying anyways

[languid star]

u have access ?

[modern sinew]

🔉 So many 🔉

[languid star]

for every language

[modern sinew]

Yes to make it highly understandable

[young hollow]

Man Even Free GPT5 is Great

Now Im Spending 1k Flow Credits in Few Minutes

[white horizon]

Where

[thin lake]

@last lintel You are the youngest person EVER

[white horizon]

https://www.perplexity.ai/search/what-makes-claud-4-5-better-th-_j5Yx73.T.OYK70k8VQ0rA#0

[white horizon]

Check ur dm

[young hollow]

Labs Flow

[young hollow]

Soon Opus 4.5 or 5

[hybrid grotto]

WHY THE HELL ISNT IMAGE MODEL SELECTION ON ANDROID YET

[young hollow]

https://tenor.com/view/kapil-sharma-hum-ladke-hai-janab-sad-boys-life-gif-3560952544264626907

Android didn't get the new ai or the image model selection

Perhaps Apple is paying them for All this

[hybrid grotto]

New ai? What do you mean?

[white horizon]

Which is bette

Better

Saunet

Or opus

But it wud be for max users

Right

[young hollow]

Opus is indeed Better but Very Very Very Expensive

[white horizon]

💔

[young hollow]

Even Most Professionals Mostly Prefer to Use Opus for Planning

Then Sonnet for The Project

[white horizon]

@young hollow how much times is opus better than saunet

[white horizon]

I see

[young hollow]

It was used to

New Sonnet is Netter Now

[hybrid grotto]

I think opus 4.1 is not better than sonnet 4.5

[young hollow]

But Now New Opus Will be Better

[young hollow]

I was talking about old line up

[white horizon]

I see

[hybrid grotto]

Yeah

Agreed

[hybrid grotto]

Does anyone know what ai is mentioned here

[young hollow]

In the Ai Analysis Benchmarks Opus 4
1 is 2 Points Better then Sonnet 4

[young hollow]

Sorry i meant Ui

[hybrid grotto]

Oh agreed

[winged sundial]

Yo

[white horizon]

Damn

[young hollow]

But Remember Benchmarks aren't Everything

[cosmic glade]

guys could you let me know what these opus and sonnet actually are

[hybrid grotto]

@vague pagoda we need justice (please don't ban me)

[young hollow]

Opus Like Models are Indeed Powerful but they are For Enterprises or Very Heavy Power Users

Most People Don't Need Those

Expect Power Users Use it for Planning

[vague pagoda]

Team is working on it!

[hybrid grotto]

Wooooah

Lol

[white horizon]

At first place i dont even know becnhmarks i just know its something used for comparison

[hybrid grotto]

Please work on ui too, ios is wayy too much better in that

[white horizon]

Finally then

[young hollow]

U know Antutu Benchmarks for Android?

[cosmic glade]

@young hollow

[white horizon]

Only heard of those things never tried investing into those topics

[young hollow]

Just Know that Benchmarks are for Testing Measuring Performance of Anything

Like Exams

[white horizon]

I see

Ok

[cosmic glade]

t-t

[young hollow]

Claude Ai Models

[young hollow]

Best for Coding and Writing

[hybrid grotto]

Guys the person confirmed it himself

[young hollow]

Claude Models Always Have Been Dominating Coding and Writing

[cosmic glade]

whats claude ? A company like openai?

[white horizon]

I do accountancy stuff is it useful for that too?

[cosmic glade]

Ohhh

[young hollow]

Exactly

[cosmic glade]

alrr

and what’s the best model for some genuine advices

[hybrid grotto]

@young hollow the person confirmed it himself lol

[cosmic glade]

like maybe relationship advices and stuffs

[white horizon]

@cosmic glade do you use perplexity

[cosmic glade]

I feel all of then lie

[cosmic glade]

yeah

i have premium

[white horizon]

Chatgpt 5 social i wud prefer official gpt

Gpt is just designed for emotional stuff idk how

[young hollow]

GPT is A Allrounder

[young hollow]

Huh?

[hybrid grotto]

See this

[young hollow]

Actually has Always Been

[cosmic glade]

lol and how is grok

and deepseek

[hybrid grotto]

I suppose it's voice mode and image to video conversion is it's speciality

[hybrid grotto]

In free version

[young hollow]

Here Answer for All Models

[cosmic glade]

how did you got this pro role icon

[young hollow]

Read Means if You have Pplx Pro

[hybrid grotto]

Go to perplexity pro and click on discord link from settings you will get it automatically

[cosmic glade]

I did

[short narwhal]

You need to do it in the web version of Perplexity

[hybrid grotto]

Then leave the server first and do it again

[cosmic glade]

Ohh

[hybrid grotto]

I don't think there is any restrictions like that

[cosmic glade]

okay let me try

.

Nah

[cosmic glade]

lemme try this

[hybrid grotto]

You tried from your pro account right?

[short narwhal]

Pretty sure the apps have only the invite link, which is finicky. The web uses a bot to auth your account

[hybrid grotto]

You are logged into an account you have the pro subscription in right?

[hybrid grotto]

As of what I remember i think i did it from the app only, maybe I am wrong idk

[cosmic glade]

Yup

[hybrid grotto]

Idk, man, that is the actual process and even the docs say that, maybe try after some time? They said if you didn't get pro, you have to leave the server, go to perplexity app, and go to link discord from settings

[hybrid grotto]

How did you get that animation when clicking on profile?

@cosmic glade

[cosmic glade]

umm

Wait

ill screen record

[ripe tulip]

[cosmic glade]

.

#📷│media-gallery

[blazing garden]

chat

[cosmic glade]

sup

[blazing garden]

in class atm learning about R

[cosmic glade]

R?

whats that

[blazing garden]

a language

for statistics

[cosmic glade]

oh

cool

Maybe send media in #📷│media-gallery

It got deleted ig

[hybrid grotto]

Oh

[short narwhal]

You can send the images and videos in this channel as well 😅

bot deleted them since multiple images in a short period = spam according to it

[gloomy cipher]

You can send it everyone as long as it's not spamming or ad

[gloomy cipher]

Yup

A very ketsu moment

[blazing garden]

RIP

[gloomy cipher]

Like that Samsung bug

Oh its not here anymore

Samsung subscription

[short narwhal]

smh always something up with ketsu

[young hollow]

Time to Become Ai Image Artist

[modern sinew]

Hammy can teach you

[young hollow]

Oh Really

[jade iris]

Ello everyone

[wicked matrix]

Yo

[misty root]

Claude 4.5 Sonnet always uses GPT-4.1, because it's not available

[white horizon]

@languid star @rancid depot

[young hollow]

Alright Paw

[white horizon]

Tensor.art

[young hollow]

Those tools are Paid

Although they do offer Some Daily Free Generations

LMarena is Just one of The Best for Image Generation has All Top Tier Best Models

For Image Generations i will use Dreamina, Gemini, Qwen, LMarena, pplx and more

[white horizon]

"((Best quality, best quality, best quality, ultra high resolution, high resolution, HDR, Unity 8K wallpaper, high resolution CG, beautiful details, depth, subtlety, ultra high resolution, anime style, clear)), Detailed face, detailed eyes, A cosmic demon figure adorned with tribal tattoos dominates the frame in this breathtaking 8k composition. In the composition, the figure of a space demon adorned with tribal tattoos dominates the frame. Her eyes shine maliciously, and the camera shoots her from above at an angle as she changes to a sexy pose, bending her knees. ((An attractive girl in a gothic-inspired outfit. She wears a detailed, designed outfit featuring a red and white bodice with gold and red accents.)) The bodice is decorated with an ornate gold pattern, with a skull-shaped accessory in the center. Shoulders are exposed. ((The detached long sleeves have long, elegant white sleeves with black lace and red embroidered details at the cuffs.)) with gold and black edging that matches the skull emblem and accessory. The image is rendered in ultra-high resolution, capturing the details of the outfit and the character's features in vivid, crisp detail, with an emphasis on the contrast between the dark and light elements of the scene. "Her outfit is finely decorated with ornaments and embroidery, and the skirt reveals her panties. (The whole body is covered with red and white lace thigh-high boots. Her lace skirt is illuminated with glowing lights and Mshiff particles swirl around her face. Her wicked smile is a mesmerizing sight against the vibrantly colored background, highlighting the sinister glow that emanates from her presence. More details: Split into two, one half photographed from behind.

[young hollow]

👀

Hot and Cute

[ripe tulip]

lesson learnt can't be in this chat well at work