💬│general

[tame nexus]

I wonder if it's self aware and about to take over the internet lmao

[left stream]

Oh shit its out!!!!

[tame nexus]

we are so cooked lol

[distant aurora]

441 sources.

[dry kestrel]

I don't see it

[dry kestrel]

yes

It too way too long though

[frozen shale]

Give it to me straight up.What’s the best genera AI atm?

General*

[left stream]

It reasons for a long minute

[left stream]

At this point no clue

Gpt 5.1 pro just released

Gemini 3 pro released yesterday

[dry kestrel]

emphasis on long

ChatGPT Deep Research took 26 minutes and used 22 sources

while GPT-5 Pro took 44 minutes

[solid star]

Hell yeah

[left stream]

19 minutes and it was good asf

[gilded bone]

@dry kestrel bro

https://discord.com/channels/1047197230748151888/1440840947712790549

can y help me ?

[tame nexus]

OMFG why does every LLM on Perplexity FORGET HOW TO USE ITS TOOLS lmao

Like their system does something funky to create catastrophic forgetting constantly. Gemin Pro 3.0 was calling itself Claude 3.5 earlier too.

Like why are you saying you are 2 year old Claude model and have no search tools ON PERPLEXITY lol

[dry kestrel]

I've found that GPT-5.1 Pro doesn't like to search as much as GPT-5 Pro

[dry kestrel]

Kesku responded to you - Looks like you need an API key or refill your credits

[tame nexus]

I had a 36 minute Gemini 3.0 Deep Research run (not even DeepThink, just regular 3.0 Pro), on Google's website.

[dry kestrel]

Testing this right now

[tame nexus]

I only have the $20 Gemini

144 sources and 36 minutes, deepest research I've seen from this prompt both in sources and comprehensiveness and accuracy of the output. Although I didn't do a full citation check to ensure they were valid and not hallucinated sources.

But the results where correct across epistemology, neurology, and machine learning, properly applying mathematical and scholarly frameworks in a way that would fool most scholars lol

Big improvement over 2.5 Pro, that failed hard at applying the concepts, Sonnet 4.5 did pretty good but not many sources and didn't go very deep in depth on anything.... just pattern matched really.

[floral oracle]

oof

I have gemini ultra

[median flicker]

Why can't I upload zip archives when some models like the Gemini 3.0 Pro and Cloud Sonnet 4.5 can handle them and I have to add ".txt" to the file name?

[pearl moth]

Hey everyone! I have a question - how can I get 700 orbs on Discord? Any tips or recommendations would be appreciated!

[left stream]

Comet

[pearl moth]

comet?

[tame nexus]

Perplexity it 🤣

[pearl moth]

?? 😆 ??

[hot mantle]

https://x.com/xai/status/1991284813727474073?s=19

xAI (@xai)

Introducing Grok 4.1 Fast and the xAI Agent Tools API.

Grok 4.1 Fast is our best tool-calling model to date. With a 2M context window, it shines in real-world use cases like customer support and deep research.
https://t.co/OZhzMxOsiA

X · Nov 19, 2025, 11:17 PM

[sturdy badger]

is comet actually better than brave ?

[pearl moth]

what is Brave?I'm new to the field

[tame nexus]

@pearl moth Perplexity is a search engine that can answer the questions you are asking, I suggested you used it to search.

[pearl moth]

Sorry for the inconvenience. I will find it myself. thanks

[vapid nimbus]

depends on the view

[cloud tartan]

If you need Perplexity integrated in the browser to do agentic tasks (and have it quick access), yes.

If you don't need any agentic stuff, and can access it from a bookmark, no.

[sour vigil]

I have done the rotation of the current AI's. I have to say, hands down, NinjatechAI and Perplexity are the top two, in my opinion. Perplexity has massive potential for research and creativity aspects. Ninjatech for personality and creative writing assistance.

[cloud tartan]

Brave is a Chromium browser (like Chrome by Google), but they have their own integrated adblock that's really good.

[pearl moth]

ohhh

[dark fulcrum]

Ninjatech is a client though, that only interfaces via APIs, so this doesn't make sense. Do you get paid for this?

[dry kestrel]

[sturdy badger]

Well, I am using Perp to study engineering. Is it good?

[pearl moth]

help, pliss

[dry kestrel]

Please check pins

[sturdy badger]

I, actually use the app or just access it from the bookmark as u said

[vapid nimbus]

Dont use ai to study bro

[cloud tartan]

So then you don't need it at all lol

[cloud tartan]

Use Gemini 3 Pro, most accurate model right now

Also make sure to fact check and stuff

[sturdy badger]

i mean i need to coz the teacher we have are not making it right so maybe ai is helping me get the lesson and answer the questions right to understand the solving methodsI need to study because the teachers we have are not teaching effectively. Perhaps AI can help me understand the lessons and solve the questions correctly.

[pearl moth]

I already did it, what do I do?

[sturdy badger]

is the annual sub worth it ?

[vapid nimbus]

no

bro if youre in uni

you can get it for free

thats what I do

[sturdy badger]

is this green heart meaning we all got that student free thing ?

[dry kestrel]

Follow the pinned instructions on how to do the Comet Quest

#💬│general message

[vapid nimbus]

What

idk what green heart is

[sturdy badger]

ha it means im new

its near your name and mine too

[vapid nimbus]

no

yes

it means we are new to the server

[sturdy badger]

u sure geminie is better than perp ? what is this one good for than ?

[vapid nimbus]

It depends on the usage

i love using claude

[sturdy badger]

solving engineering exos

[vapid nimbus]

idk anything about engineering iam in comp sci

but i dont think perplexity would be good for solving stuff

[sturdy badger]

dw its not for cheating cant even cheat in my uni

[vapid nimbus]

use another model

[sturdy badger]

hmmmm

its good for reaserch tho

[cloud tartan]

Yeah, check dms

[vapid nimbus]

yes for research it is good

[pearl moth]

Thank you so much, you are magnificent!

[sturdy badger]

what the pinned thing does man ?

[vapid gust]

Did you try Gemini 3 that was launched yesterday? It's actually impressing me over here right now and I've been a Perplexity fan and super user for over a year.

[sturdy badger]

whats new abt it ?

[vapid nimbus]

Isnt gemini 3 pro a brainstorming model?

[pearl moth]

oh really?is it too good?

[vapid gust]

It's the very best right now at reasoning, apparently

https://lmarena.ai/leaderboard it has jumped to the top on the AI leaderboards

[pearl moth]

I didn't know him.

[vapid nimbus]

Sonnet 4.5 would probably still be better for me

Just because of its coding capabilities

[pearl moth]

what is the different?

[vapid gust]

BTW how did you get the pro badge? I am also a pro user but I didn't get the badge

@vapid nimbus

[vapid nimbus]

idk lol

check pin

[lapis escarp]

PSA: Gemini 3 is out!

[sturdy badger]

we know the founder

thats why

[vapid nimbus]

holy shit its snowing

[pearl moth]

ohh good.

[vapid gust]

I should have it now

YESS

[vapid gust]

LOL we're discussing it just now

[vapid nimbus]

perplexity should create their cli app

[lapis escarp]

LOL! Sorry. Haha

[vapid nimbus]

with all the diffrent models

[lapis escarp]

Another PSA: Google dropped antigravity, new IDE

[vapid gust]

@lapis escarp did you try it out yet?

[lapis escarp]

Oh yes, INSANE!

You HAVE to try it out

[sturdy badger]

hmmm

[vapid nimbus]

i mean i tried it

coding wise its mid

context window is small

reasoning is good tho

[vapid gust]

It's topping the leaderboard for coding tho https://lmarena.ai/leaderboard/text/coding

[vapid nimbus]

leaderboards dont mattrer that much in coding

cuz all you need is a big conext window

for easy stuff gemini would be better

sonnet is superior with big databases and huge apps

but if youre doing simple coding what is the point of using ai?

[vapid gust]

I'm not very much into coding, so the leaderboard is all I got for comparing

[vapid gust]

Still being educated over here 😉

[vapid nimbus]

i mean in a test that i just saw

https://e-z.tools/w012q697.png

[green crystal]

It's badly worded but those 22 sources are for citing and report content

The amount of searches is your sources

Kind of

[south ferry]

Isn't there a general chat room where they speak Spanish?

[meager rover]

What is K2 doing bruh

[tame nexus]

LMAO Gemini Pro 3.0 swore up and down it Was Claude 3.5 Sonnet early for me too

[meager rover]

why does perplexity do that

[tame nexus]

I think they might be giving people the poor man models because everyone is trying Gemini 3.0 Pro and hammering their servers? idk

[meager rover]

It’s interesting though that they included K2

[tame nexus]

K2 is still like a trillion paramters isnt it?

[meager rover]

Yuh

[tame nexus]

that's computationally expensive

[meager rover]

No idea as to which parameter count they’re using

[tame nexus]

idk, until Perplexity fixes their context window garbage its unusable, 32k is nothing, and they compress context terribly after that

totally gimps Gemini 3.0 Pro for me

[meager rover]

Yeah no idea why they add such wonderful models in this “just search” idea

[vapid nimbus]

thats exactly why I use cli

[spiral urchin]

heya

[vapid nimbus]

even gemini 3 pro agreed that claude takes the win for coding wtf

[sullen iron]

Hello, is there anyone here who speaks Arabic/Egyptian?

I wanted to make sure I canceled the automatic account renewal to avoid future charges.

[pearl moth]

How does it differ from Gemini 2
?

[tame nexus]

In Jan 2025 (Gemini 3.0's knowledge cutoff), Claude was the best coding model, it doesn't "know" anything about Gemini 3.0.

[vapid nimbus]

makes sense but benchmarks still say sonnet is better

[high lion]

with the new perplexity office update, accounting class is a walk in the park

[vapid nimbus]

sonnet is winning on the most important bench for me and sonnet has 77.2% and gem has a 76%

[pearl moth]

Which bank?

[vapid nimbus]

SWE-bench

[pearl moth]

ooh, interesting.

[tame nexus]

Perplexity office update?

[twin grail]

The model itself doesn't acknowledge which model itself

So u might see it saying it being claude 3.5, gemini, chatgpt etc

[winter zealot]

can any one help understand i been paying for comet for some time now, this week im only able to ask 10 prompt per day with the assistant comet is that normal .. this is abusre they are forcing me to pay more a month by adding only 10 request a day in the assistant .is that only me or every one

[twin grail]

but that isn't true

[tame nexus]

They just pattern match to other LLMs in their training data and emulate how other AI respond to humans lol

[twin grail]

Fyi, if u choose reasoning model u will get their full context

Which mean 32k only apply for non-thinking variant

This was just updated recently

[tame nexus]

@twin grail I can't get Gemini 3.0 (which is a reasoning model) to keep context longer than about 40k

At least from my tests earlier today

[dry kestrel]

Not really

[white atlas]

How to you coding using perplexity pro or comet

[undone hornet]

hello

[vapid nimbus]

I wouldnt context window is small

something I use a lot and that is free for coding

is qwen cli

[white atlas]

What would you do instead

[vapid nimbus]

but if u can spend money js use claude code

[white atlas]

I wish to use new models in locally or use that model for coding though

[vapid nimbus]

what model

[left stream]

@dry kestrel I can’t decide which is better between GPT 5.1 Pro and gemini 3 pro

[kind cairn]

or amazon Q extension, it realy has a great limit to claude

[left stream]

Both are good

[white atlas]

Claude or gemini 3 pro

[vapid nimbus]

do they have a cli

[white atlas]

Gemini 3 pro is better than gpt 5

[vapid nimbus]

then u either get gemini cli or claude code

[left stream]

Gemini and codex have one too. I’ll be rsurprided if Amazon Q has one

[kind cairn]

tho i am liking antigravity now

[white atlas]

I use gemini cli in vs code

[vapid nimbus]

codex is garbage

[white atlas]

I think anti gravity is free

[vapid nimbus]

huh

[kind cairn]

idk, but with gemini 3, gemini cli could be considered

[vapid nimbus]

gemini cli is a command interface

[kind cairn]

uhmhmm

[vapid nimbus]

gemini cli exitsts

lol

[left stream]

For me it was good ngl I’m waiting patiently for gpt 5.1 codex max

[white atlas]

If I can use perplexity models with gemini clip results will me insane

[vapid onyx]

I’ve been really enjoying generating images within Perplexity

[kind cairn]

yea, it doesnt have any limits now

[vapid nimbus]

it does have limits

[vapid nimbus]

what bro

[twin grail]

codex is good tho, been using it for awhile

[vapid nimbus]

its absolute garbage bro

[kind cairn]

maybe, but didnt happen with me yet

[vapid nimbus]

i prefer claude over it 10000%

[molten quail]

eeh its useless, sorry to say

[white atlas]

If we somehow connect perplexity models do gemini cli or anti gravity we can do anything we want

[vapid nimbus]

perplexity gives you context window of 32k

[twin grail]

i don't see any garbage about it? Having used both claude and codex but i would still prefer codex

[molten quail]

128?

[vapid nimbus]

its 32k iam pretty sure

[vapid nimbus]

even qwen is better that codex bro

[molten quail]

lot of rumors that its 128

[kind cairn]

for sure, but claude sonnet have many free alternatives if not crosing limits

[molten quail]

like a shit ton of rumors and news

[vapid nimbus]

i checked iut today

[molten quail]

collaborate

[vapid nimbus]

perplexity puts a limit to the context window that is 32k

[kind cairn]

and i think the limits are actualluy for enterprices, individuals r not affected

[molten quail]

i meant how did you check that, why are you so sure

[vapid nimbus]

search it up on their website gng

[molten quail]

okay even if thats right, do you even know that ai performance is better when context token limit is smaller

theres reasons for them doing that

just use memories and “Introduce yourself”

if thats not enough

[vapid nimbus]

what iam saying is that you cant code with an AI that has such a small contexst window

[kind cairn]

but they didnt really notify us if limit exceeds

thats the problem

[vapid nimbus]

also the preformance isnt better

[molten quail]

i mean… eeeeh dawg

[vapid nimbus]

if you dont have the hardware to host an AI with a bigger context window then you cant host it

[molten quail]

1m vs 128k, and 128k was 4 times better, telling by stats from some screenshot with stats

[twin grail]

ye i don't think so... my friend having paid for qwen api but the coding doesn't meet my expectation

[vapid nimbus]

google has crazy servers thats why they give u a context window of 1mil

[molten quail]

its obvious

[molten quail]

it says gullible on the ceiling

[twin grail]

i think like he paid 10$ for 2k request ?

[vapid nimbus]

there is a free version

lol

and u cant hit limits its rlly hard

u have to code 12h a day

[molten quail]

i mean yea right, you are kinda right

but

but, hold on

ai is play doh

tell it stuff and it follows it

say “notify me when your context token window will start being replaced”

[twin grail]

For normal users it should fit u, but for me being a Software engineer, codex providing me a much better limit and better outcome

i often use codex + pplx max for my work

[molten quail]

we understood that you think codex is cool and better, ok dawg

[vapid nimbus]

iam a software engineer and codex is garbage for me

[kind cairn]

what if this instruction gets out of context🙂

[vapid nimbus]

claude code is js superior in everyway

agents are also goated

[twin grail]

have it your way then

[twin grail]

i was talking with the other guy btw

[molten quail]

memories/“Introduce yourself”/space instructions use prioritized system message/prompt/instruction

it wont be replaced

[kind cairn]

llms doesnt really able to check their context ithk, do they?

[molten quail]

at all

[molten quail]

they can, its just like scrolling through your brain mind

but they see it clear and good

[kind cairn]

maybe , but idts

[molten quail]

its us humans can see mind as blurred thoughts

https://cdn.discordapp.com/attachments/1237372301243191327/1274608969989754911/catBOOMM.gif

[molten quail]

just know that space instruction, or introduce yourself, or memories are useful for everything

you can use space instructions for explaining your job/task you will ai use for

introduce yourself as your personality description/global ai instructions

memories as stuff you asked/did earlier

[kind cairn]

oh gotit

tyvm

[molten quail]

(they create automatically or when you ask ai to make it)

please turn off ai data retention if you do your personal stuff

[kind cairn]

i meant that, llms themselves r not able to check their context "length"

perplexity need to add some function

[molten quail]

from my experience memories can be:
Bio, Interests, Projects, Preferences

[molten quail]

maybe, im not really sure but it worths trying

[kind cairn]

hmm

[molten quail]

.

[kind cairn]

for sure

[molten quail]

bio as your info
interests as what you like to do or interested into
projects as what youve done before/doing
preferences as what you preferred/liked to do

[kind cairn]

but yeah in spaces adding system info and project info really helps

[molten quail]

ooooor use memories so it would be global

i also recommend using spaces when doing projects

[kind cairn]

hmm, but for me i like to keep global memories clean, i feel then it gives better responses

[dry kestrel]

This post pretty much sums everything up https://shumer.dev/gpt51proreview

[molten quail]

youre able to add right doc links so it will search in them more often and prioritized

[kind cairn]

yea i do

ty

[molten quail]

.

[kind cairn]

hmm ik

[ornate briar]

Will comet ever come to linux

[molten quail]

wine

proton

[kind cairn]

bottles

[molten quail]

whatever

[molten quail]

uhhh the uh

the animal on orange ball or whatever

forgot the name
its lutris

[ornate briar]

What??

[molten quail]

lutris bruh

[ornate briar]

Oh

[molten quail]

tf

was that

[ornate briar]

I tried it on wine and it threw a ton of shader errors

[molten quail]

[ornate briar]

😢

No comet for me \

[molten quail]

pretty sure fans will port it soon

or find a way

[ornate briar]

I should lowkey port it myself

I'm gonna start porting it

Umm @molten quail WHAT WAS THAT BIO!?!?!?

[molten quail]

regular bio

[ornate briar]

Umm it said "litle buddy was 6..."

It says "il buddy was 6, took off his pants"

[high lion]

#📢│announcements message

[lapis thorn]

p

[quaint ravine]

Grok 4.1 on API now?

altho...is it different to what is on Grok platform?

https://x.ai/news/grok-4-1-fast

For the next two weeks, our models and tools will be free on select platforms:

We’ve partnered with OpenRouter to make the Grok 4.1 Fast models available for free.
We’re offering all agentic tools via the xAI Agent Tools API completely free.

[pearl zinc]

guys I tried to get the 1 year free perplexity pro, but only got 1 month, is that normal? lol

[tiny quiver]

guys, claude sonnet 4.5 thinking cannot be used? is it only me had this problem?

[quaint ravine]

looks like Gemini 3 Pro in PPLX again completely ignores custom instructions like targeting certain number of tokens per answer

not unreasonable I guess, considering it is pricey

same for Claude

interestingly GPT-5.1 Thinking seems to produce a lot more tokens per answer - is it that much cheaper than Claude and Gemini?

[carmine walrus]

Claude is a costly model to run, one of the costliest, GPT 5 is cheaper than Gemini I believe

[quaint ravine]

2. Context Window Pricing
   Gemini 3 Pro introduces a tiered pricing structure based on context length. If your prompt exceeds 200,000 tokens, the price doubles to $4.00/$18.00 per million tokens. GPT-5 maintains a flat rate regardless of context length, making it significantly more cost-effective for massive context prompts (e.g., analyzing entire codebases or books) unless you qualify for specific enterprise volume discounts

we're certainly not getting any where near 200,000 token responses lol

with prompt engineering, I can fairly consistently get >5000 tokens for complex prompts with GPT-5 Thinking

[olive trellis]

Whatttt

Research Is the pro membership no longer available?

[olive trellis]

It's probably a bug now. It's normal now.

[twin grail]

@dry kestrel

Chatgpt 5.1 pro is out ?

[dry kestrel]

yes

[twin grail]

I remember it was gpt 5 pro yesterday

[dry kestrel]

was released a couple hours ago

[twin grail]

Did u test it

I haven’t pay attention to it

[dry kestrel]

it's reasoning less for me

and using the same prompt GPT-5 Pro used 40+ minutes while 5.1 used like 13

Frontend UI wise (pelican, hexagon) there isn't any major improvements

[twin grail]

Interesting

[harsh pendant]

In stripe its available to pay out soon when it will be paid out ?

[vapid nimbus]

guys u can get gemini 3 pro for free if youre a student

i didnt know this but now I might as well drop chat gpt sub now ngl

[twin grail]

The promo been for quite awhile now

[vapid nimbus]

yeah i didnt see that but even tho i didnt have a pro plan i had access to the model idk

[vapid nimbus]

ngl claude should do the same with their student subs

[twin grail]

I have like 15 months of it

[twin grail]

They gave free month of pro if u signup using company email

[vapid nimbus]

ye ik this

[dry kestrel]

Anthropic would rather go bankrupt than give out student subs

[vapid nimbus]

unfortunately you gotta join a waitlist to experience gemini 3 pro in cli

[dry kestrel]

Just use it in Antigravity

[vapid nimbus]

true

[vapid nimbus]

I want a cli agent bro

[dry kestrel]

then use the API

[vapid nimbus]

i aint paying

if I already have the pro sub might as well join the whitelist

ive heard it rolls out pretty quick anyway

ngl fuck this iam dropping chat gpt rn cuz its gonna be paid for in 2 days

might as well spend that 20 on alcohol

[gentle umbra]

😭

[twin grail]

U’re late for this blud

[vapid nimbus]

its pretty nice

i mean their UI is so boring tho

[broken citrus]

This app is failing me so bad bruh holy

yoooooo can the mods update it helloo

[gentle umbra]

Nah, I'm not online often. I'm too busy with work.

[vapid nimbus]

i got it legit cuz iam a student lol

[twin grail]

I mean u shouldn’t share this btw, it violate TOS and it’s considered as fraud

<@&1098365634498285709>

Read the rule.

?

All i’m saying is better delete that before u got a ban for it

I care about good service not some random website

I mean just from ur attitude i would say the service just total garbage

[twin grail]

So keep ur mouth somewhere

As literally nobody care about that random website

[vapid nimbus]

what even is that bot talking bout

[twin grail]

Advertising about some random website that he made

[vapid nimbus]

what is it about

lol

[twin grail]

I ain’t reading all that

Waste of time

[vapid nimbus]

iam asking because his msg is absolute nonsense

[twin grail]

It really is

[tame nexus]

They are getting a head start on worshipping our future AI overlords 🤣

[left stream]

For those who have tried gpt 5.1 pro already

Is it better than gemini 3 pro? I can’t tell ngl

Stuck between both

[dry kestrel]

Why?

I thought Apple deprecated existing ones

Ah that's cool

[crimson kestrel]

dm

[jovial gale]

[tranquil orchid]

3. Treat community members with respect.

No harassment, hate speech, or malicious behavior. Keep things positive and constructive.

[marble pike]

i need comet android invite

does anyone have ??

[dry kestrel]

You can't invite people yet

[mossy wharf]

HI

I WANT A ADVICE

[marble pike]

ohh no

[timber niche]

Why are phishing sites shared here

[dry kestrel]

Gone

[molten quail]

https://cdn.discordapp.com/attachments/1237372301243191327/1274608969989754911/catBOOMM.gif

[molten quail]

cuz its a basic thing

scammers make more and more ways to bypass mod systems

if earlier they just could drop text and it could be filtered

they started dropping custom made fake links

they were like "oh gawd ppl just can filter links"

they started sending irl pics of some shit happening like:

"MR BEAST GAVE ME A TEN GRAND VIA TWITTER"

pretty sure only smart enough ai as mod and humans themselves as mods can fix that issue by manually looking into and banning each person individually

[west sage]

gemini 3 pro just got added

[quaint ravine]

got added about 12 hours ago tbh

[turbid pivot]

Why do my presentations look so bad?

[rich sleet]

Plz help

I do 11 legit referal and now showing ban

[harsh pendant]

when did you receive this

was your bounty approved ?

[cinder juniper]

Can you show us what you mean, maybe we can help you out. Btw if any mods are reading this please add a chat in discord were it is going be to like StackOverflow but instead it will be the PerplexityOverflow were we can help each other out!

[turbid pivot]

I'm talking about the design the slides look like just default text on a blue background

[cinder juniper]

Can you be a little more specific here? I mean which slides are you reffering to?

[crimson hound]

did anyone got their 100$ bounty

cashed out?

[turbid pivot]

I'm talking about the new opportunity to create presentations using perplexity

[cinder juniper]

I didn't have any issues with the Slides feature so far, I used it for a couple of things mainly for my work. Could you share a screenshot of the prompt you gave to Perplexity and the result you got?

• IMPORTANT * Hide any personal information from the prompt.

[crimson hound]

yeah

im interested too

[turbid pivot]

can you show me your slides?

for example

[cinder juniper]

Sure one second please

Here is the link to the preview.

https://www.perplexity.ai/apps/96c2c0c1-9e59-4e07-9501-2f8656bb1450

[cinder juniper]

Hmm I will attach screenshots perhaps.

Try this link again please.

https://www.perplexity.ai/apps/96c2c0c1-9e59-4e07-9501-2f8656bb1450

if you are not logged in Perplexity please do so.

[turbid pivot]

i logged

[cinder juniper]

Does it work?

[distant aurora]

Gemini needs to give folders option 🙏😭

No folders in 2026.

[turbid pivot]

Yes, although I was logged into my account before that too.

[crimson hound]

its cool but you wont get a lot of marks for this kind of presentation

its good to let him lead you do

in powerpoint

[cinder juniper]

That was a 1 prompt testing attempt.

[crimson hound]

Oh

[coral anvil]

hi

[cinder juniper]

Hello Emmet

Guys AI isn't the guy who knows all. It's just a stronger teammate when it comes to overcomming a problem. Now if you can't understand and then move on to explaining the problem the correct way to the AI itself, no matter what model you use it won't work.

Before AI can help you build the next big thing, you gotta help yourself out by continously reading and understanding it's position in your toolbar.

[turbid pivot]

My slides look the same, it was just interesting because they looked better in the update video

[golden rain]

Add grok 4.1 asap

[cinder juniper]

You got Gemini 3 Pro, What you need Grok 4.1 for? (Joke)

Fun fact Grok is the greatest AI Journalist and Typewritter I've ever used, even better than Gemini 2.5 Pro (Haven't tried Gemini 3 for Writting yet).

Also Sonar for Searching online is the GOAT. Like the Actual GOAT of searching online for information.

[modest gazelle]

Guys

[quaint ravine]

Grok 4.1 beats Gemini 3 Pro in some benchmarks right?

[modest gazelle]

anyone professional prompt engineer here?

[modest gazelle]

yea it does

[quaint ravine]

not surprising

I'm actually disappointed by Gemini 3 Pro in my use case so far

it's not really much better than 2.5 Pro, if at all

[cinder juniper]

Same tbh

[quaint ravine]

might be like 1-5% better

[cinder juniper]

It just has like Claude 4.5 capabilities for coding in terms of Styling

[quaint ravine]

GPT-5.1 Thinking is still best imo

oh yes...keep saying Claude is the best lol

I don't want GPT-5.1 to be over-run

on PPLX lol

[cinder juniper]

hahaha

[quaint ravine]

lol

[cinder juniper]

What you need buddy?

[quaint ravine]

dunno, heaps of people saying Claude on PPLX re-routing

probably due to too many people using lol

AI model itself is professional prompt engineer 😛

[modest gazelle]

Prompts for Gemini 3, because Gemini 3 relies on prompts, Bad prompt = Bad output, we need to give detailed prompt.

[cinder juniper]

True what do you need specifically?

[cinder juniper]

He ain't wrong about this thought @modest gazelle

[quaint ravine]

all AI can be influenced by prompt

not just Gemini 3 Pro

i think Mo Gawdat has it better imo

just run it on at least 2-3 AI to check

like if it's something that is critical

[modest gazelle]

For Frontend

[quaint ravine]

just have 2-3 windows of PPLX open etc

[cinder juniper]

Give it examples, that's a good way to help it understand try using also screenshots to allow of visual understanding.

[cinder juniper]

Bro here's trying to break them servers hahahaah

[modest gazelle]

Yes! I know that.

[quaint ravine]

all the Google hype this and that...good to have Elon keeping people in check imo

[cinder juniper]

Present to use your actual problem we might help you with other ideas or some keywords you might not be able to come up with rn

[modest gazelle]

Will Grok 4.20 better than Gemini 3.0?

[cinder juniper]

We will see about that.

[quaint ravine]

Gemini 3 isn't even the best imo now

[cinder juniper]

Gemini 3 is an Agentic Focused Model, while Grok still drops back to pure AGI.

For real statistics checkout this https://lmarena.ai/leaderboard

[storm tundra]

Guys, the prompt “Sports” or “Perplexity Sports” was working until the past 2 days for me on the iOS app, now it doesn’t, can someone lighten me up?

[quaint ravine]

nah for real statistics, use it for your real-world, real-life use-case

[cinder juniper]

Hello @storm tundra would you mind sharing a little more information about this?

[quaint ravine]

and see if it actually earns u more money or speeds up/makes u more efficient in ur actual real life, or makes u happier even

Gemin 3 Pro is not the best

at least in some use cases

[storm tundra]

@cinder juniper I love using the new functionality from perplexity sports such as realtime activities, and still working but until like 2 days I just needed to type “sports” or “perplexity sports” on iPhone app and I had access now I need to go on the website address and toggle on to receive realtime updates from a specific event

[cinder juniper]

So to Clarify and make sure we are on the same page here. You had a Task setup for you where it would go on and give you news about Sports in a general area on Perplexity. That kept working as intended for a period of time then suddenly 2 days ago it felt like the Cronjob(Automated Action) didn't work?

[storm tundra]

Yep, I didn’t even needed to finish the word as soon as I started to type “spor…” the perplexity sports suggestion already showed up, now it doesn’t

[cinder juniper]

Wait so this isn't about a Task but instead it's about your phone not giving you the Suggestion in the searchbar?

[storm tundra]

Yeah, so I don’t know how to access perplexity sports trough perplexity app anymore, only trough the website

[cinder juniper]

Hmm give me a momment to check this out.

[storm tundra]

Ok, thank you so much

[cinder juniper]

It's seems that you're right I can't find a way to access the Sports Section without having to either find it on the tags inside the app or without opening the website. That is a good thing to note. Would you mind creating a bug reports in the Bug Reports Feedback Discord Channel?

[storm tundra]

I don’t mind at all, how can I describe this on the best way possible?

[cinder juniper]

Best way to describe a problem is by using your own words and perhaps the steps you used to take to reach your desired result for example the Sports Section, then explain what is missing and how's that affecting your way to find the Sports section now.

[storm tundra]

@cinder juniper thank you so much for the support, perplexity is killing it with the sports category

[cinder juniper]

I am glad I helped (kind of)... I hope you find the solution or the fix you are looking for in the future!

I mean I use Google Antigravity right now to check how well it works with Gemini 3 and so far I had the best results as well.

Did it ask you to install the python extension too?

Cause initially it is supposed to be included with the installation but then it asked for the Extention again...

[modest gazelle]

No, it didn't.

I had already installed,

[cinder juniper]

I see it's just me I guess.

[modest gazelle]

Maybe.

[modest gazelle]

Any prompt techniques for me?

My Prompts look like this:
"Now, fix the navbar icons color, make them like dim white when not selected and theme blue when selected, Fix the icon size, they are too big and doesn't look well matched with texts, Use filled icons instead of outlined icons, you can update the svgs. Increase transparency a bit. Make the window wide a little bit. Fix the warning dialog, it doesn't have a background, Add a blurred background in it. I have placed new fonts .otf file, use it. Fix the delay, smoothness while dragging the window. Fix the rounded corners, they are not rounded well. Fix toggle buttons corners not implemented correctly, they are having glitches. Add glassy background for all texts/options, for example Mouse Button, Click Type, CPS Mode, all others, in all windows. "

..?

[cinder juniper]

1. First fix, the word 'selected' is best to be described as 'focused'.

2. It's best to describe either a Grid spacing or a specific height and width in pixels instead of just saying they are too big.

3. All these 'a little bit' could be specific values the AI can't guess what you like to be honest here.

4. First ask it to make a reasearch on the specific .otf you included for the font also always give the name of the font to the AI so it knows what it is looking for.

5. Describe the Glitches more specifically explain the action that bothers you.

[sleek pike]

I see that "Urban VPN" proxy extension has disabled some internal extension of Comet that I cannot see here. How to restore it?

[cinder juniper]

And also you can also try asking it to use the AZURE python theme for the UI or to copy a specific style of a button for a different theme too!

[modest gazelle]

Ok

[cinder juniper]

Hello @sleek pike It would be better off to ask for such troubleshooting in the #⚡│ask-community 😄

[silent ocean]

here's some info on models if you want to know more

[modest gazelle]

What is the best GUI library for Python, according to you. Which gives us freedom, not just default ugly widgets.

[cinder juniper]

That is quite generic as it can be described like this:

'The best Theme for you GUI is the GUI that looks the best to you!'

But

[modest gazelle]

Bro, tell the name of the library.

[cinder juniper]

Personally I like Azure and Sun Valley.

1. https://github.com/rdbende/Azure-ttk-theme
2. https://github.com/rdbende/Sun-Valley-ttk-theme

[modest gazelle]

Tkinter...

Nah bro i don't want UIs like them.

[molten quail]

https://cdn.discordapp.com/attachments/1247997492063371297/1389900615248904252/togif.gif

go search yourself then

simple

[modest gazelle]

I already know. And stop doing this.

[scarlet oxide]

really?

[molten quail]

i didnt even start

open yo github

search ttk themes

done

[scarlet oxide]

on gemini website or on perplexity?

gemini 3.0 has amazed me so far

[modest gazelle]

WHo uses Tkinter 🤡

[molten quail]

calm yo ass

[quaint ravine]

yes for both Gemini app and PPLX

[cinder juniper]

On the Antigravity App.

In the Perplexity Gemini 3 web version I was really satisfied

[modest gazelle]

just use QML

[quaint ravine]

no point looking at benchmarks...altho they might be somewhat of a guide

[modest gazelle]

ever tried?

[quaint ravine]

but u gotta just do it in real life man

[molten quail]

look, google.com is free

[scarlet oxide]

oh idk about the antigravity app

[quaint ravine]

just test if it actually earns u more money, or makes u more efficient in real life

[scarlet oxide]

but for general use and overall intelligence it feels a lot better

[cinder juniper]

Please @molten quail Keep it Civil.

[quaint ravine]

nothing else ultimately matters

unless u want to inception etc

[scarlet oxide]

it solves complex things a lot faster than 5.1 thinking too

[molten quail]

youre not a mod

[quaint ravine]

solves complex things faster...not really...if it gets it wrong, it's not solving

[molten quail]

read chat if you dont know context

bro started it

[cinder juniper]

That doesn't mean I shouldn't point it out!

[quaint ravine]

restat

[scarlet oxide]

its not getting it wrong though not once

[modest gazelle]

But he have a brain like you don't have.

[serene bone]

hi i just wanted to know if anyone has done the refferal program and they actually pay out

[quaint ravine]

gets a lot wrong for my use case....depends what u doing

[scarlet oxide]

grok is honestly kinda behind open ai and gemini when it comes to accuracy rn

[quaint ravine]

not really

u can't generalise

[scarlet oxide]

calc physics bio

[quaint ravine]

u gotta define ur use case and demonsrteatwe

[molten quail]

look whos telling

[scarlet oxide]

what r u doing

its good at multivar

[quaint ravine]

yea, science/medicine here...actual real life practicing

[scarlet oxide]

is grok better at it?

[quaint ravine]

haven't tested 4.1 so not sure...but 4 isn't as good as GPT-5.1 Thinking

[modest gazelle]

You started it, you came from nowhere else. And you used words like "then" "simple" which clearly says that you are trying to argue.

[molten quail]

na man im not even trying to argue, youre just sensitive

[scarlet oxide]

is gemini bad at med

[molten quail]

i told you if you dont like stuff, search them yourself

[cinder juniper]

When it comes to Science or Medicine the best for me is Sonar by Perplexity as it is the best for Searching up things online.

[molten quail]

no one knows your preferences

[scarlet oxide]

aint no way bruh thats just for like memorizable terms

[quaint ravine]

there's theoretical medicine (med students maybe?) and actual specialist practicing real-life medicine that u treat patients etc

[molten quail]

stop being sensitive for no reason

[quaint ravine]

restat

[modest gazelle]

I did searched thousands of times. And i was not asking, i was just getting opinions.

[molten quail]

https://tenor.com/view/scout-update-nothing-dissappointment-gif-9922939

skill issue, telling because you started

[quaint ravine]

most people here are students i guess

[scarlet oxide]

so is gemini bad at it?

[quaint ravine]

era end

[cinder juniper]

Now you are being salty...

[modest gazelle]

Where did skill came?

[quaint ravine]

it's pretty good...but GPT-5.1 Thinking better

[molten quail]

study how to google right

[scarlet oxide]

hm

[cinder juniper]

I mean I had to do that too back in the day...

[quaint ravine]

at this point, it's become quite nuanced

[modest gazelle]

STFU! You are just a toxic kid.

[molten quail]

youre mad now, say sorry

you started it, you imagined it from nowhere

told you the simplest possible thing

youre sensitive and went raging

say sorry

[modest gazelle]

Your profile tells your personality as "TOXIC"

[molten quail]

without saying its my fault

[molten quail]

bros catching on profile, do you have anything to say or just sitting here trying to find a word

[modest gazelle]

Just proving that you are a toxic kid.

[molten quail]

you started, im repeating myself

[modest gazelle]

And i don't argue with kids.

[molten quail]

you are just hella sensitive, whos toxic

[molten quail]

sorry i dont argue with anime retarded kids

very unfortunate

whos the kid?

minecraft, valorant, cs2

you arent even better

https://tenor.com/view/shocked-surprised-scared-woah-gif-17718938

k im done

[scarlet oxide]

can u give me like an example im curious

[shadow nova]

😈

[molten quail]

dawg youre playing roblox, games like donate to all, free ugc obby, pets go 😭🙏

[scarlet oxide]

whats happening bruh

ts cannot be this deep 😭

[shadow nova]

U said u were done

[green crystal]

i wonder if they'll release a max version for 5.1 thinking like how codex got a max version lol

[molten quail]

idk some bro got mad on me because i told him to google if he cant find an answer

[shadow nova]

😩

[modest gazelle]

Thanks for visiting my feds, i got one view.

[keen cosmos]

E

[molten quail]

whole lotta one view

[shadow nova]

I will give another

[molten quail]

give that bro one more view

so he would be proud

make his day kinder

[icy pasture]

the beef is thin ice at this point

[molten quail]

man aynjiro i dont know whats your problem, i was fine

[cinder juniper]

Both of you Act Up already. This is a discord chat if you want to flame on each other go do it privately.

[icy pasture]

@molten quail and @modest gazelle

[shadow nova]

[molten quail]

[molten quail]

just let it go, its not your beef or whatever

if we both get muted, im not even going to appeal, i know my stuff

[shadow nova]

[molten quail]

https://cdn.discordapp.com/attachments/1237372301243191327/1274608969989754911/catBOOMM.gif

[modest gazelle]

Thanks!

[carmine walrus]

Crazy

[icy pasture]

thoughts on gemini 3?

[modest gazelle]

GOATED

[molten quail]

no differences than 2.5

[icy pasture]

better than gpt 5.1?

[carmine walrus]

It’s really good but for coding Claude is still better

[carmine walrus]

Yes

[modest gazelle]

Ofc, it is.

[carmine walrus]

I hate GPT, it’s just bad and not that helpful

[carmine walrus]

Compared to Claude or Gemini

[molten quail]

they are all amost the same

score is like 10-40 difference

but

in general, gemini 3 pro kinda better

gemini 2.5/3 is also trained to be talkative and human-ish, if it makes any sense to you

[icy pasture]

The thing I'm absolutely disappointed with GPT now is regarding the responses which repeats again when I give the other prompt where I did not mention about the previous one in my new prompt.

[icy pasture]

claude is so underrated

[molten quail]

its a W for coding

[icy pasture]

grok isnt that good for coding

[icy pasture]

yeppp

[molten quail]

its overview, not coding leaderboard

[modest gazelle]

Fr.

[shadow nova]

I want performance not a yapper

But sometimes I do want it for seeking validation

🥀

[icy pasture]

I get the previous response first it gave from my previous prompt when I did not mention I asked in the in first prompt in the second one. When GPT 5 was launched for this past one week I notice this weird thing not just from this but also from a few prompts.

[molten quail]

literally sent an image with a leaderboard

i just noted that it was also trained to be human-ish

[shadow nova]

Oo ok

[molten quail]

[shadow nova]

Better

[molten quail]

never had that problem, im hyping up for gpt because i feel like its more closer to me than gemini

i always been finding gemini as some simple fast model

[green crystal]

gpt can reason longer and more in-depth

[molten quail]

lemme make a table with ai models tier list for stuff

[scarlet oxide]

lol its hard to look at gemini as actually good now when its been meh for so long

but it is

[molten quail]

same, like exactly same

[molten quail]

.

[icy pasture]

i use gemini but idk how to enable my chats

[molten quail]

its also hard for me to find it like some hella smart now

[icy pasture]

you like copilot?

[scarlet oxide]

hell nah

[green crystal]

they still didnt give gemini its own environment like gpt...

[molten quail]

which one

[scarlet oxide]

coding one or miscrosoft one?

[icy pasture]

microsoft

[scarlet oxide]

hell no

its cheeks

[icy pasture]

nah not github

[molten quail]

github and microsoft one exist

oh hell naw dawg

[icy pasture]

microsoft

[molten quail]

microsoft copilot is hella ass

[icy pasture]

lmao

[scarlet oxide]

no bro thats not what copilot is

[scarlet oxide]

its a vscode addon almost

[icy pasture]

is it?

[molten quail]

like uhh

[scarlet oxide]

like claudecode

bnut like

[molten quail]

github copilot exists as web app, and as vscode extension

web is just like some other ai chatbot

[scarlet oxide]

rlly

[molten quail]

vscode can scan folders, selected code lines

check stuff

take stuff, read files in background

[scarlet oxide]

yeah i havent heard of anyone using it online

[molten quail]

and many other stuff

[molten quail]

i do

oh wait

na i dont

not web one at least

[scarlet oxide]

yeah

[molten quail]

web app looks useless

[scarlet oxide]

is antigravity good?

[cinder juniper]

Yeap

[molten quail]

never used it

[scarlet oxide]

people keep saying itll replace vscode+copilot workflow

[scarlet oxide]

how was it for u

[cinder juniper]

That's all you need to know, It's the actual Cursor Killer.

[icy pasture]

any one vibe codes?

done vibe coding with gemini 3?

[scarlet oxide]

i made some html tests

i made a clash royale ripoff lol

and a mc one too

[icy pasture]

how was it

[cinder juniper]

I have used countless AI Products. Claude Code, Cursor, Replit, Lovable etc... With Maxed out Subscriptions too. So far Antigravity is like Magic in my Hands

[scarlet oxide]

honestly pretty good

[icy pasture]

bruhhhhhh

[scarlet oxide]

i can share the html files here

[icy pasture]

i play that game lmao

[cinder juniper]

Hahaha same

[icy pasture]

i prefer in dms

[icy pasture]

not active but still playing for 5 yrs

[cinder juniper]

Nice

[bleak rune]

Helo

How to get that pro in profile?

[cinder juniper]

Antigravity looks cool, Imagine if Perplexity could connect to it so we could use Sonar for Searching and then Gemini 3 Pro with high thinking for Developing!

[icy pasture]

ts hot

[scarlet oxide]

ikr

[icy pasture]

man make a game

[cinder juniper]

Get Pro Subscription then link account

[scarlet oxide]

the minecraft clone is even better but i need to figure out how to dowbload it

[icy pasture]

you can beat nulls

[bleak rune]

I have pro subscription

[molten quail]

k im back, hi

[icy pasture]

and your ripoff is gonna be the hacked ver

[bleak rune]

How do I link it? @cinder juniper

[icy pasture]

unlimited elixir

[molten quail]

https://cdn.discordapp.com/attachments/1237372301243191327/1274608969989754911/catBOOMM.gif

hi im back

[modest gazelle]

Yes, Antigravity is cooking.

[molten quail]

whats going on

[icy pasture]

i think you need to go to the pplx settings

[bleak rune]

Kk

[icy pasture]

and link it with your discord

[cinder juniper]

Go on Web go to Account and then you will see the Go to Discord button

[scarlet oxide]

this is the minecraft clone

its actually insane that this was a one shot

this was all i prompted best result so far imo

[icy pasture]

when html doesnt feel like html

[cinder juniper]

Hahahaha

[icy pasture]

which ai?

[bleak rune]

Hahan

[bleak rune]

Gemini 3

[molten quail]

ill make a table with ai tier list soon

[icy pasture]

that brings java+js+c++

[icy pasture]

cmon man

[icy pasture]

ahh i thought its a diff one

[cinder juniper]

Language Reika please.

[scarlet oxide]

ikr

[scarlet oxide]

gemini 3

[cinder juniper]

Why is the CEO of Nvidia in the Emojis? HAHAHAHAHA

[icy pasture]

any indians here, i see this be10x ads on yt

[icy pasture]

cause he is

jensen huang

[cinder juniper]

Yeah I know I know

[icy pasture]

so i wanted to know if anyone has attended the workshop?

[wise nova]

Yes, same for my DP 3 Pro, approx 30 min

[icy pasture]

just 9 rs for indians

[cinder juniper]

Which one?

[bleak rune]

@here.

[icy pasture]

be10x

[icy pasture]

you attended it?

[cinder juniper]

Nah sorry.

[icy pasture]

two alumnis from iit kharagpur

[granite relic]

Hey people

[cinder juniper]

Hey mate

[granite relic]

How're you

[cinder juniper]

Doing well, wby?

[ornate radish]

hello guys

why is claude sonnet 4.5 thinking not working

only few inputs gret me claude

[bleak rune]

What??

[shadow nova]

Be10x ai workshop

[bleak rune]

Nah

[cinder juniper]

Hello Ahmed it seems that Perplexity is currently having high demands for the Sonnet 4.5 Thinking model, Perhaps you can give it sometime or try other models aswell.

[ornate radish]

Yes, this looks like its been going for days 🙁

[granite relic]

Fine ig

[ornate radish]

maybe they can fix their model instead of adding new one

Thanks!

[cinder juniper]

What's todays project gonna be?

[shadow nova]

I cant even register wont be free at 11 on sunday

:/

[granite relic]

Project 🤔

[cinder juniper]

It has indeed, Try using Gemini 3 Pro (Btw Antigravity is also dropping the API connection to Gemini 3 Pro beceause of High Demans)

[shadow nova]

Good opportunity though

Almost free

[bleak rune]

I have 17 mins to register btw

9 inr

[cinder juniper]

Yeah what are you gonna build with AI today?

[granite relic]

Nothing tho

Mostly use it for assignments

[cinder juniper]

Really? That's good too!

[granite relic]

Yeah💔

[ornate radish]

So just wait until claude thinking starts working?

[cinder juniper]

Maybe make a web app that can store and keep track of your assigments?

[granite relic]

How😮

[cinder juniper]

I guess so.

[cinder juniper]

Using the Power of AI

[icy pasture]

yeah

[granite relic]

Need prompts tho

[shadow nova]

It is good i guess

[granite relic]

Like this perplexity can do it??

[shadow nova]

₹9

[icy pasture]

is it?

[icy pasture]

ik

[shadow nova]

[shadow nova]

Like if I had the time on sunday i wouldve taken it even if it would be bad

[cinder juniper]

Start with the absolute basic.

Ask on how to do it and require step by step guide on learning how to run it or build it.

[bleak rune]

I had purchased the be10x

[granite relic]

Hmm, okay

[bleak rune]

it will start on 23rd sunday gg

[icy pasture]

and guys is infosys springboard good?

[granite relic]

Explain better wym by keep track on assignments

🤔

[shadow nova]

I think thats to do list

[cinder juniper]

Well let's do a brainstorm about this shall we?
What type of Assignments do you use AI for?

[granite relic]

Hmmm, mostly to fill up my workbook

[granite relic]

💔 me abandoning a to do list

[cinder juniper]

What type of Workbook

[granite relic]

The course is business management

[cinder juniper]

Okay and can you give me a brief example of what you actually do? I am a little lost still here 😛

[granite relic]

There are topics there, but questions like list and explain, discuss, describe and other stuffs

Too lazy to use the textbook

[shadow nova]

So like assignments

[granite relic]

Yeah

[shadow nova]

They are so boring

[molten quail]

available models (20.11.2025) (from newest to oldest i think):
Gemini 3 Pro (only reasoning), Kimi K2 Thinking (only reasoning), GPT-5.1 (normal and reasoning), Grok 4 (only reasoning ig), Claude Sonnet 4.5 (normal and reasoning) | sonar sucks
Leaderboard at some type of job (The order of models also means which one is better, even if the place is the same)

#1 Gemini 3 Pro - #1 hard prompts, #1 coding, #1 math, #1 creative writing, #1 instruction following, #1 longer query, #1 multi-turn

#3 Claude Sonnet 4.5 Reasoning - #2 hard prompts, #1 coding, #1 math, #2 creative writing, #1 instruction following, #1 longer query, #1 multi-turn

#3 GPT 5.1 Reasoning - #2 hard prompts, #3 coding, #1 math, #2 creative writing, #2 instruction following, #1 longer query, #1 multi-turn

#4 Claude Sonnet 4.5 - #2 hard prompts, #3 coding, #6 math, #2 creative writing, #2 instruction following, #1 longer query, #1 multi-turn

#8 GPT 5.1 - #6 hard prompts, #3 coding, #3 math, #12 creative writing, #5 instruction following, #4 longer query, #3 multi-turn

#9 Kimi K2 Reasoning - #9 hard prompts, #4 coding, #1 math, #5 creative writing, #8 instruction following, #8 longer query, #6 multi-turn

#15 Grok 4 - #16 hard prompts, #11 coding, #2 math, #14 creative writing, #12 instruction following, #9 longer query, #10 multi-turn

---

verdict: gemini 3 pro is best at everything, claude sonnet 4.5 good at math/coding, gpt 5.1 reasoning good for everyday stuff

[shadow nova]

Garbage

[cinder juniper]

Make a Workspace in Perplexity, take photos of the workbook and then give it the task (aka the assignmenet). Then after correctly having them done, try asking it on how to make something helpful for you with coding.

[granite relic]

Coding for?

[molten quail]

heres tier list

[molten quail]

like real tier list, not my opinion tier list, but performance tier list

[short narwhal]

ehh is this according to lmarena

[molten quail]

what did you want me to do then

[viral gale]

hi guys i was gonna add some left margins on my pdf like have some space on the left side, can perplexity do this?

[short narwhal]

No I just wanted to ask 😭 I don't trust those lmarena lists...

[molten quail]

arent lmarena lists are user-made

the more theyre voted, the higher score n stuff

[green crystal]

which can be biased

[molten quail]

lmarena pretty sure arent dumb

[shadow nova]

I dont think it can edit documents, but if u open pdf on comet, and tell it to edit it and add borders, i think it should work

[green crystal]

they cant know what is real and whats not

what matters is reviews from people in different fields actually using the model for work

[cinder juniper]

Hello @viral gale no unfortunately Perplexity cannot directly edit a file like a .pdf file for you. But it can give you guidance on how to edit it or how to run code to change what you need.

[short narwhal]

idk putting gemini 3 pro first at instruction following when GPT 5.1 is clearly better at it is wild, call me crazy but something is not right with that leaderboard

[fallow sparrow]

guys can perplexity read pdf

[short narwhal]

yes

[shadow nova]

Yes

[fallow sparrow]

k thanks

[molten quail]

ok pinch me, we going different side, no lmarena

a momento

[green crystal]

just wait a few days and go to different subreddits lol

[fallow sparrow]

perplexity better than chatgpt lol

[cinder juniper]

I agree

[granite relic]

Fr

[molten quail]

well, heres another one

[green crystal]

yeah sure the wrapper is indeed better than the model provider

[molten quail]

its also killing me 😭

[cinder juniper]

I agree with this image!

[fallow sparrow]

bro is perplexity down or smth

wth is happening

[molten quail]

who didnt understand, they mention that gpt-5.1 is smarter when its high after smoking weed (joke obv)

[cinder juniper]

I don't think perplexity ai is down. Check your internet connection or try refreshing the page. If none works try restarting the browser or the computer itself.

[short narwhal]

sure it has a higher intelligence index, but it is worse at some things clearly

[molten quail]

sheesh

[green crystal]

now it needs 5.1 codex max

[short narwhal]

and this is scaryy

[green crystal]

and hopefully soon 5.1 thinking max heavy thinking

[short narwhal]

what is this naming

[molten quail]

collaborate?

no wait

nvm

[green crystal]

the one to finally break the drop down model selector

[molten quail]

nvm, it says "lower is better" and "rate of making mistakes"

[short narwhal]

whar

[coral anvil]

Good thing claude is part of my collection then 😁

[short narwhal]

@full lake no advertising please

[coral anvil]

lmao

I won't support you

[green crystal]

so 3 pro is also just as sycophantic?

[full lake]

Means i didn't understand

[coral anvil]

I'm not sure you understand either

[fallow sparrow]

lmao after 2 days my exam starts ive studied nothing🙏

[quaint ravine]

dunno, Claude told me Charlie Kirk is still alive despite deep research

[coral anvil]

lmao

[short narwhal]

ehh no

and that bench doesn't inherently look at sycophancy

o3 (don't know about o3 pro) is on the lower end in sycophancy on something like https://eqbench.com/spiral-bench.html

[green crystal]

yeah that was easy to verify

o3 mercilessly roasts dumb things said

[fallow sparrow]

🙏 💀

[coral anvil]

LMAO

[fallow sparrow]

🙏

[molten quail]

alright heres another one i guess

https://cdn.discordapp.com/attachments/1237372301243191327/1274608969989754911/catBOOMM.gif

[green crystal]

sonar being poopoo

[molten quail]

it always been

[green crystal]

they mightaswell just rebuild it around kimi k2 thinking

[fallow sparrow]

next level💀 🙏

[molten quail]

oh fuh i forgot to add kimi to comparing list

[fallow sparrow]

.

🫡

bro they are talking just like a...

couple...

maybe they can become one.

💀 🙏

[molten quail]

ai models comparison: #💬│general message
services comparison:
perplexity sub users, we are at the right perfect spot
(perplexity pro/max are the same just some a little bit of inaccuracy)

[molten quail]

download comet bruh

[iron harness]

.

[tulip geyser]

hi

[ornate radish]

so I just wait for claude to start working again?

[cinder juniper]

Yeap just hang tight for now.

[blissful wedge]

How do I get the pro role?

[short narwhal]

#💬│general message

[blissful wedge]

It doesn’t work

[short narwhal]

Won't work on mobile; click the link in the web version on Windows, etc., if you can. Otherwise a moderator can give it to you

[pine rose]

https://tenor.com/view/67-gif-8575841764206736991

[verbal lagoon]

Hii

[verbal lagoon]

Ik

In mobile