💬│general

[jagged sphinx]

lol

I love being a perplexity Stan

It got upgraded to the 8840u

OMG

OMG

omg stupid idiot

It's still the ayaneo kun 1

[halcyon coral]

guys, could you please watch your language?

[jagged sphinx]

someone please tell this man

[jagged sphinx]

Ight got you

Here read this: https://ayaneo.com/product/AYANEO-KUN.html

[jagged sphinx]

Please forgive me I shall bow down to you. Not joking mainly because you help make perplexity and I love perplexity

Ok jay I have fought off the evil

Yep it's not a device

Use a different llm

[small tree]

Yes and could you fix perplexity please?

or acknowledge there's an issue

[jagged sphinx]

I got it correct with perplexity 01

But right now it's down

Ight lemme show you Unc

Welp I guess he learned

😘

No you just don't know how to use it

You.com is kinda embarrassing

Go pick the grease oozing out of your belly button pal

Ight imma take some relax time

[sinful moat]

skill isues mate

had to hop onto my pc

[jagged sphinx]

There is no such thing as a ayaneo kun 2s

I'm saying the ayaneo kun had an upgrade not a name change

Huh what did I miss type

But it says it's gonna get upgraded instead of saying it's upgraded

Yep admit it you.com is horrible

Ain't so genius I guess

Bye until you text me again

Huh

Oh look at what it says

Not too slick

[halcyon coral]

guys please...this is childish. either you stop arguing about this topic or you discuss it via dm.

[small tree]

@sinful moat realised how dumb it was and is deleting all his messages

mature!

[sinful moat]

huh it called stop being childish

are you a child?

[small tree]

nop

[sinful moat]

great

stop arguing then

[small tree]

shut up

[sinful moat]

so you are a child hahaha

👶

[hallow tulip]

How many of you sticking with perplexity?
They cut context, been slacking on o1 and now want to shill tiktok I'm over it.

[onyx stump]

I would still stay put.

[sinful moat]

Just found something better called Kagi Assistant

@jagged sphinx

[jagged sphinx]

Ask it what are the specs of the ayaneo kun. Don't ask the upgraded version

I asked perplexity the same thing and it got it right: https://www.perplexity.ai/search/what-are-the-upgraded-specs-fo-HNvBI5E1SBeb20ZNBlr8JA

[jagged sphinx]

You could be lying like like last time

[sinful moat]

How?

[jagged sphinx]

You literally lied last time idk

[sinful moat]

About what last time?

[jagged sphinx]

You deleted the message

[jagged sphinx]

Answer. Are you lying

[sinful moat]

Here I show

https://kagi.com/assistant/99a7df99-4f92-4863-b79d-50102af17468

[jagged sphinx]

Doesn't work

Just take screenshots of all of the chat

And send it

[hallow tulip]

So how does everyone feel regarding the recent changes to perpleixty? what are you issues with it? Perhaps they will listen and divert course.

[jagged sphinx]

I mean it's better than you.com but it still kinda sucks right now

It used to be better

[small tree]

They're going to replace their current customer base with tiktok brained customers

[jagged sphinx]

I don't really know what they did

[sinful moat]

Then everyone goes to either you.com or Kagi Assistant

[jagged sphinx]

You.com sucks

[sinful moat]

Unless you love your data to be mined by the Chinese

[sinful moat]

Then Kagi Assistant it is

[sinful moat]

Found something better

https://github.com/ItzCrazyKns/Perplexica

[idle fossil]

https://www.cnbc.com/2025/01/18/perplexity-ai-makes-a-bid-to-merge-with-tiktok-us.html

What do you think about the merger with TikTok?

[late quail]

Considering AI can already use wifi routers as tracking devices, unenthusiastic.

[waxen lion]

(-_-)

is it just me or has perplexity been giving generic responses recently

it does not do as much research and feels biased

[granite lake]

valuable data. and could help develop generative video AI?

might be a shortcut to being able to complete with google as theyd become a larger company with more data to further develop their services to take on other tech giants

[raw seal]

hi

[small tree]

Yea same for everyone

They switched off 3rd party llms

And fallback to their in house one

Which is crap

Now theyre trying to get tiktok to buy them so they can use tiktok user data to train their crappy inhouse llm

So they can actually compete in the space

Now that other llms are gaining web search capabilities, perplexity has no unique selling point

@granite lake @waxen lion @idle fossil

[sinful moat]

did you know tiktok tracks you outside of the app in other apps actually crazy

imagine you use waze it actually knows where you are in real time

@jagged sphinx perplexica actually insane

[waxen lion]

So basically my 20 dollars are getting wasted

[sinful moat]

I'm just self hosting this is

and has better results

[waxen lion]

Fack !_!

I used to love perplexity until recently because now it's just not working like it used to

Let's say what solution they try to come up with

For now I'm going back to chatgpt plus

[sinful moat]

pretty sure tiktok will merge with perplexity to avoid ban in US

even perplexity says the same

[waxen lion]

dude i get that but it doesnt change the fact that the service isnt working the way it should be

[sinful moat]

CCP shifting through the data now 💀

[waxen lion]

could also be a publicity stunt , who knows

[sinful moat]

this is probs happening now

[sinful moat]

nah it's actually a very strategic move

since US is trying to ban tiktok

if tiktok is under a different comapny name that is US based

US can't do anything and it would look bad to close a US Company

[granite fractal]

The very goal of banning Tiktok is to cut off the CCP, and not Tiktok itself. There's a reason why the CCP don't want to sell Tiktok.

[sinful moat]

I know this already the CCP gets alot of it's information through tiktok

[waxen lion]

anyways if the merger goes through , im done with perplexity

[sinful moat]

mate just use perplexica like me

better results

I'm trying out another 2 open source alternatives

[waxen lion]

how much to run it

[sinful moat]

only the api usage

[waxen lion]

gotta check it out now

can you send me a sample output

[sinful moat]

about what?

[waxen lion]

recent trends in ai tecchnology

technology

[waxen lion]

are you using perplexity pro or normal

because the pro search i believe is a bit better

[sinful moat]

this perplexica using my own anthropic api key

[waxen lion]

ohk

[sinful moat]

I'll show morphic

[waxen lion]

this is the result i got

[sinful moat]

it's just restarting

[sinful moat]

so perplexica is better than morphic and also I've tweaked the system prompt

and now I get this

[waxen lion]

i think both of them are abt the same

i went to perplexity settings and im currently using gpt 4 0 model

[zinc warren]

hi experts, how is the perplexity paid gpt 40 model different from chatgpt 40 model paid version?

[rugged pewter]

Hi

[waxen lion]

its their own trained model

[zinc warren]

so it does not use chatgpt 4 o at all?

or chatgpt allows them to use the 4o model to train on top of it?

[zinc warren]

^

[tepid portal]

EVERYTHING dude..., I recommend to try What is a god credentials in the context of ByteDance?

In the context of ByteDance, the term **"god credentials"** refers to a "superuser" or administrative access level that allows unrestricted access to all user data collected by the company. This includes data from platforms like TikTok. According to allegations made by Yintao Yu, a former ByteDance executive, these credentials were reportedly used by members of a Chinese Communist Party (CCP) committee embedded within ByteDance to bypass privacy protections and monitor sensitive user information, including that of Hong Kong protesters and U.S. users

[silk flax]

is anyone else getting these horrible ui changes recently

who is over engineering these ui changes???

[sinful moat]

this is crazy

[silk flax]

does anyone actually appreciate these new UI changes?

[sinful moat]

no that's why I'm on perplexica

[silk flax]

👍

[sinful moat]

better results and better everything over perplexity

[silk flax]

how is the answer quality compared?

interesting

[sinful moat]

really good

what do you want me to search?

[tepid portal]

Exactly...

[silk flax]

what would this task be?
dependencies {
compileOnly(project(":extensions:shared:library"))
compileOnly(project(":extensions:tiktok:stub"))
compileOnly(libs.annotation)
}

[silk flax]

yeah i would say your's is better

what api do you use? @sinful moat

i'm impressed open source is this good

[sinful moat]

I'm using sonnet 3.5

anthropic

[tepid portal]

How much it cost compared to Perplexity or ChatGPT API?

[sinful moat]

actually not that much surprisingly

[tepid portal]

Hmm nice

[sinful moat]

probs going to hook up nano-gpt which is model provider or openrouter

[tepid portal]

I wanted use local LLM, but my computer isn't enough fast.

[sinful moat]

use openrouter or nano-gpt

[tepid portal]

Wow nice

[sinful moat]

even got the new minimax o1

so does openrouter just now and it has 296 llm models

[silk flax]

openrouter + perplexica

should be enough

[tepid portal]

I've never known about this, it looks much much better than mainstream!

• Open Source

[hoary phoenix]

How can I force Perplexity to check today’s date before providing answers? If I ask a question where the accuracy of the answer relies on knowing todays date, it’s almost always wrong because it thinks the date is October 2023

[waxen lion]

Are you using perplexity free or pro

[small tree]

What the hell is wrong with perplexity at the moment

why arent they acknowledging anything

[waxen lion]

even im wondering that bro

[small tree]

are they really taking days off on saturday and sunday?

its a 9 billion dollar company

hire more workers

[waxen lion]

my perplexity was working fine a week ago and now it feels like absolute g a r b a g e

[small tree]

how do they not have systems in place to alert them that somethings wrong

they are lazy

[waxen lion]

they r having ui problems + output problems

its irritating af

[small tree]

Yea, they disabled all the LLM's

and are defaulting to the in house LLM

which is garbage

worse than gpt 3

0 context window

[waxen lion]

i used to say that perplexity was a hidden gold mine due to its api and functionality

but its become absoluitely useless ever since this tiktok thing came up

[small tree]

Yea, it's going down hill

[chrome moon]

Hellos

[waxen lion]

not to mention , the support is completely ignoring it

[chrome moon]

This IA have a oficial cripto?

[small tree]

like, at least acknowledge it

[waxen lion]

i didn't pay 200 usd for this kind of service

although i still have access to third party LLM due to me being a pro subscriber

i think they might have disabled it for free user

[small tree]

@waxen lion use google AI studio for now, it's excellent at Code tasks

https://aistudio.google.com/prompts/new_chat

it's free too

Gemini Experimental 1206, from the dropdown

is great

[waxen lion]

i have copilot pro ( github ) , chatgpt plus , claude ai pro , and perplexity

[small tree]

oh wow

[waxen lion]

im fine on ai tools

[small tree]

haha nice

yea i have gpt pro too

[waxen lion]

i just miss perplexity

[small tree]

but ran out of o1

prompting

[waxen lion]

i hate that they have such a small limit on it

[small tree]

openai?

[chrome moon]

Perplexity have a official token?

[waxen lion]

honestly im thinking of switching to open source and then just using api credits

[waxen lion]

im not even joking right now , a week ago perplexity pro was working so much better than chatgpt pluis that it was crazy

it was taking info and finding the exact things i wanted

[small tree]

Yea, what's that tool called? openrouter

[waxen lion]

these guys made some kind of s h i t t y backend change thats ruined the experience

i hope they fix it soon

[small tree]

they will, on monday

hahaha

[waxen lion]

although i did find a clever workaround

[small tree]

they are partying

at the moment

[waxen lion]

i made a custom space with specific instructions to the ai and set the model to grok2 , working well so far

ill switch the model to gpt 4 o if it starts to mess up again

update : i have returned my perplexity back to its original glory through prompt engineering

and using specific model

[small tree]

@waxen lion which model actually works on perplexity

they disabled all of them except the inhouse one as far as i know

[waxen lion]

i have pro subscription

[small tree]

most of us do

although it says the model changes

it doesnt actually change

you can tell by the outputs, its all using the same default model, no matter how many times you rewrite

[waxen lion]

it changed a bit after i put the custom commands

i think they r going to answer this on monday

and they better because we paid for access to those specific third party llms

although they did not make any sort of announcement related to this

but it does feel like they have disabled the third party LLMS

plus i havent given my consent for them to use my data to train their models anyway

[lilac stump]

they've done this before, arbitrarily removed OPus without waring and changed the details on teh site and subscription plans to imply opus was not something you were paying for and contractually obligated to receive

they also have that "if you send more than 100 messages to any model, you'll be locked to gpt4 for 6-12 hours" thing too, so you dont even get 600 messages ot sonnet or grok i a day

[rose geyser]

what people use for citation search these days ?

[burnt echo]

Perplexity for Discord

[lunar scarab]

when I search with claude 3.5 sonnet, is it actually claude 3.5 sonnet as it doesnt really sound like it

[feral marlin]

they’ve done things similar to this without telling people or giving any public updates of what they’re working on several times over the last year

expected from them

[waxen lion]

i think they fixed it now

my responses are back to being dynamic

[waxen lion]

check out the complexity extension for perplexity

[hallow tulip]

I'm pretty much hoping that OpenAI can put in the time and effort for SearchGPT to really shine, I'm tired of these AI search engines each of them are pretty much cooked in their own special way.

I find that Perplexity is absolutely shady in recent months, Phind is just completely taking data, You.com RAG is lack luster and I've been generally disappointed with the space so far!

[small tree]

yea

they fixed it

[chrome dove]

guys can you explain what's the difference within the different AIs on perplexity?

please

[small tree]

wtf

its broken again

@waxen lion

lol

so annoying

[waxen lion]

Its like going from normal to ded again

@small tree I found a solution

Use complexity extension on chrome

[small tree]

thanks but because it's not a popular extension, i dont trust it with my data

[waxen lion]

but it fixed the issue for me pretty quicky lol

i can finally use perplexity like normal

[small tree]

I'll try it on incognito or something

tyy

[waxen lion]

hi so is it fixed for you now

[small tree]

@waxen lion which plugin within complexity

do i need to activate?

[waxen lion]

activate model selector , better focus , recency

[chrome dove]

guys can you code?

[small tree]

Yea, it doesn't activate the other models

this is simply just a dom editor

it doesnt have any additional access controls to perplexity, it just changes and uses the information it has available

[waxen lion]

i dont know why but my perplexity is working much better when i actively change models are a few uses

it just makes it easier

[small tree]

It's cool, it improves the UI, but its deffo not a fix

[waxen lion]

like i can see some differences

[small tree]

ty though

[waxen lion]

have you tried this btw

[small tree]

yea

[chrome dove]

i finished high school and i am on a gap year before going to university as i needed to find out what career i want to pursue and i am bbecoming interested in computer science, llms and data analysis, but i fon't know where to start! would you be so kind to share with me your insights and opinions on how to start and what to expect from this path?

[small tree]

the path is different for everyone

[waxen lion]

i swear chatgpt is so fun lol , i generated this rn

[small tree]

I expect to make tens of millions of dollars but it's not the same for everyone

[waxen lion]

roadmap.sh , datacamp , educative , codecademy there are multiple ways to start

the main thing what is your motivation

do you have a bussiness /growth mindset because then it is very important or do you find it fascinating

[waxen lion]

my brother is currently a senior software dev at apple i can tell you what he did

[chrome dove]

at first i thought of studying llm and brain-computer interfaces believing that there could be found a way to facilitate learning in humans via these fields, only lately i am thinking that i will need to sustain my studies somehow so i need to find something that on the way will help me earn enough to pursue my research

[chrome dove]

i really hope you achieve that!

so, for example, what do you use AI for at the moment? i feel like there is so much out there and i have only used it as a search engine/ writing support basically

[waxen lion]

computers in general to me is basically like freedom , think about it for a second , we are in a digital era where the main emphasis everywhere is on computers , i believe learning how to code o r knowing the basics of coding is a given in this day and age especially for people who are planning to start their own bussiness or move on in life because coding is actively being used in all fields , its even better if you know how to code because it can be integrated with any other field that you are pursuing

[waxen lion]

thats a smart way to go about it

i wouldnt put all my time into it

try to put at least 30 percent of your time into it

if you want to know how to start , i can ask my brother lol

[chrome dove]

i am amazed in fact at how fast all this technology is developing and i think only a rather small percentage (still) has put hands on this

[waxen lion]

people will learn over time , in the future i think we are going to have a group of humans that are just not needed anymore

[small tree]

Currently using the likes of o1 and other advanced LLMs to expedite the process of improving my online casino, performance improvements, code analysis, ui / ux analysis for improved conversion

[waxen lion]

think about it , tech is increasing its potential every counting second of the day and is making multiple jobs obsolete , the only way to rise is to move on

[chrome dove]

i hear the echoes of Nietzsche here ahahah

[waxen lion]

also for university courses , ik would suggest picking a course that has a defined variable , for example going somewhere where a person will always be required , you have to think about this one thing : what is one variable in this world which possibly cannot be affected by technology , and when thinking of this , think of the two emotion of people that have existed in every era : greed and fear

[chrome dove]

you made this!?!?

[small tree]

Yes

from scratch, UI in figma, front-end in NextJS and backend in Laravel

[chrome dove]

wow.

[waxen lion]

very impressive

[chrome dove]

i am glad i am stepping into this world already

[small tree]

It's exciting

[waxen lion]

it truly is

@chrome dove how do you plan to being your learning journey

[chrome dove]

the awe i felt thinking you made this online casino yourself single-handedly inspires me that anything can be made

[waxen lion]

its his hard work that made it

also @small tree what are you hosting it on

[waxen lion]

come to dms

[small tree]

Amazing, the backend is a 2 man job as security isn't my strong suit, but I started off as a UI designer, then got into the front-end and backend

[small tree]

Frontend on Cloudflare Pages

which can handle unlimited traffic essentially

Backend is on some big dedicated servers

in multiple regions

websockets for the realtime updates

[chrome dove]

i wish i could understand any of this aahaha

[waxen lion]

It will take about 3 months

[chrome dove]

btw i am 19 and im from italy? hbu?

[chrome dove]

so do you think the intro to computer science by harvard is a sufficiently good source to start with?

[red flare]

What does it mean exactly when you switch to different models on perplexity pro, such as chat gpto? Is it the same as using chat gpt? Just seems like perplexity still

[sinful moat]

I found something better than perplexity called ithy which uses multiple LLM models at once then generates an answer like o1

Then it generates a report

@jagged sphinx
https://ithy.com/article/ayaneo-kun-specs-2025-rxuu0hb1

[small epoch]

Are there perplexity employees that can fix billing issues here?

[signal hamlet]

Hey @small epoch! Please send the details to our [e-mail] team so they can fix it.

[small epoch]

Thank you for your response, so should I remove this? https://discord.com/channels/1047197230748151888/1330692235175399487

[crystal cosmos]

yes remove it

[trim wedge]

How to enable pro search by default in the search bar? It used to be possible to do this via https://www.perplexity.ai/search/new?copilot=true&q=myquery, but this method has failed in the last couple days, is there any other way?

[tame current]

yea its been horrible

I cancelled my subscription yesterday

Its only been getting worse with perplexity so oh well time to js go with normal ol GPT pro

[barren orchid]

Hello

[crystal cosmos]

chinese owned company games available in 🇺🇸 are now getting shutdown.

due to politics

[hoary phoenix]

Sorry, missed this! Pro!

[waxen lion]

@small tree looks like perplexity support finally came back to life ol

lol

[viscid fjord]

Hello

[trim wedge]

Hello, does anybody knows how to enable pro search by default in the search bar(already pro user)? It used to be possible to do this via https://www.perplexity.ai/search/new?copilot=true&q=myquery, but this method has failed in the last couple days, is there any other way?

[fast meteor]

the correct way is : https://www.perplexity.ai/search?focus=internet&copilot=true&q=your_query

[viscid fjord]

Hello guys, I'm a writer and assignments specialist, so if anyone needs help just let me know

Dm me

[serene sand]

#💗│sharing message ☕

[trim wedge]

seems not working
for me, only starts as normal search, not pro

[fast meteor]

hmm it worked for me earlier today, but not for now....
you can post the issue here:
#1140622008086970420 #1111765634267742408

[trim wedge]

thx, I'll do that

[stone axle]

Anyone else have issues with Spaces ignoring custom instructions? I have to remind it to follow them to get the proper response. Pretty annoying

[mortal bolt]

In my case, even after turning off the web source, it starts adding web sources

[smoky dragon]

hello

can i be staff?

[mild summit]

@everyone Is there anything which i can tell adults in a showcase program about tech and computer tips and tricks so that they will be Impressed by me???

@everyone ??

@deep raft ??

[inner apex]

Where do you use this?

[mild summit]

In my tekie showcase program

[inner apex]

whao

free enough？

@mild summit

[mild summit]

What do you mean by it

[inner apex]

I'm wondering whether I should use the Pro version. ‘

[mild summit]

bro any one whith advice please 🙏🙏🙏🙏🙏🙏🙏🙏🙏

[inner apex]

[torn moon]

perplexity no longer has ability to generate images? Also, why isnt that featutre integrated into options if its available as it was before?

[signal hamlet]

Images are generated, please reload the page, we'll look into this.

[vapid onyx]

I enjoy the Pro version for the model selection personally. I like tinkering with other models to see what works best for me.

[small tree]

How long does it take to revert the changes

It was clearly a bad push which broke it, no one tested it before merging to main?

Dont you guys have unit testing setup

9 billion in valuation and you guys dont have enough devs to revert a change or fix a critical bug

Youre losing customers as we speak

[half igloo]

@small tree Excuse me, sir -- who the hell do you think you are?

[small tree]

Jay

😂

who the hell are you?

[waxen lion]

He's a customer

-_- and a fellow user

Why can't he ask questions

Also Alex would really appreciate if my responses from perplexity were dynamic like before

@small tree have they fixed it ?

[small tree]

Still not fixed and it's Monday

[waxen lion]

Can you look into the generic responses pls

[waxen lion]

You sound like a fanboy fr

Hopefully perplexity fixed this before the end of the week

[small tree]

@half igloo You're typing for an awfully long time, how about you use perplexity to come up with a response for you?

[waxen lion]

Bro He stopped midway lol

[half igloo]

Ah, yea I started writing a couple of times. But decided you two lowlifes were not worth wasting resources on

[serene sand]

https://x.com/testingcatalog/status/1881399907032191334?s=61

TestingCatalog News 🗞 (@testingcatalog) on X

Perplexity may add DeepSeek R1 to their model offering 👀

X · Jan 20, 2025, 5:54 PM

[half geyser]

Hi, if you had the budget to buy one pro subsription only (regardless of price), between Perplexity Pro, Kagi Ultimate, You.com pro, ChatGPT Plus (for Web Searches), or anything else I can't think of right now. Would you still go with Perplexity Pro, or would you use something else, and why? For searching the Web and answering questions.

[jagged sphinx]

Yo, im using CloudBank is that good or do you recommend something else

It has good interest rates

[small tree]

@jagged sphinx depends what country you're in

[jagged sphinx]

Merica

Home of hamburgers

What do you use again

[small tree]

@jagged sphinx Revolut USA interest rates aren't great

UK is at 4.75%

how much do u get atm?

Also if you dont need the cash, just slap it in the SP500

and forget about it

[jagged sphinx]

The apy for cloud bank is 5. Something

[small tree]

nice

[jagged sphinx]

Yeah your an online casino right?

[small tree]

Yep

[jagged sphinx]

Can you dm the link by any chance

I cant access it

[small tree]

Why not?

[jagged sphinx]

Most likely blocked by internet provider

[small tree]

Yea, probably gambling block

[jagged sphinx]

Is that for Roblox or something

Because thats what rbx stands for

[small tree]

It was, but not anymore

Primarily a crypto casino now

[jagged sphinx]

Oh cool

Did you scam people of robux💀

[small tree]

What's your ISP?

[jagged sphinx]

Ddos.com

Nah im joking

[small tree]

you should try

[jagged sphinx]

But for real tho

[small tree]

haha

[jagged sphinx]

Did you scam people of robux

Im not gonna get angry or nothing

[small tree]

Why would we scam robux?

[jagged sphinx]

Because it said rbx

Meaning that its most likely a bunch of advertisements that give you money

[small tree]

only scam here rn is the perplexity pro subscription

advertising model selection which doesn't function

[jagged sphinx]

Its working right now

01 works

[small tree]

oh

lemme check

[jagged sphinx]

You have to toggle of pro tho

Because 01 is technically pro since I believe it doesnt have a cutoff point maybe. Just guessing

[small tree]

It's using my quota but it isnt showing the reasoning steps

weird

[jagged sphinx]

Nvm it has a cutoff point

[jagged sphinx]

Wdym, are u using tha a-I

Api

[jagged sphinx]

Oh Nevermind I understand my bad

Well it works for me

[small tree]

oh crap i think its working

thanks

the toggling worked

[jagged sphinx]

Ur welcome cuz

Guys ii think imma just use Siri💀

[cinder saffron]

I think the API is down.

https://status.perplexity.com/

Says everything is okay. But I keep getting 524 errors. Is there a doc with all that status codes?

[jaunty dove]

https://x.com/AravSrinivas/status/1881458694266953934

Aravind Srinivas (@AravSrinivas) on X

You can try DeepSeek-R1 on https://t.co/KJmY4Me2CL. We'll try to bring it up on core Perplexity in the context of advanced reasoning pro searches pretty soon.

X · Jan 20, 2025, 9:48 PM

[crystal cosmos]

only 45% will be deepseek, the rest is groq

[feral marlin]

also still has the lowest context out of anything on the market and horrible hallucinations from files

[waxen lion]

come to dms

[sleek topaz]

perplexity down?

[gritty shell]

Seems like it. Down for me as well.

[sleek topaz]

🥲

nice pfp btw

[royal niche]

Press f to pay respect.

[hushed summit]

Temporary pain, I was using it

[obtuse stratus]

looks like it's back up

[hollow pelican]

Any clue when it will be back up or did this just happen?

[royal niche]

Same, was having a lively discussion about how it was evil and it said not available, this thread doesnt exist xD

[hazy stratus]

oo

looks like it came back

[crystal cosmos]

perplexity down

[half geyser]

To where? Document management system?

[waxen lion]

You were choosing deciding between perplexity pro and chatgpt right

[narrow tide]

Where is o1 in perplexity? It was there a few days ago.

[tight hound]

they removed it

[narrow tide]

I can see that… I was implicitly asking why

[feral marlin]

for money

[crystal cosmos]

for money

[feral marlin]

they removed it because they care about you ❤️

[narrow tide]

I care about you too sweetheart

[crystal cosmos]

smash

[harsh verge]

If it is true that o1 does not seem accessible via "Pro" search at the moment, whether due to technical issues or a deliberate decision (perhaps for a future implementation), o1 is still present on my end and works (without Pro mode).

After sharing a video (#1329845798485233757 message) reproducing this "issue" with Pro, I have yet to receive any response.

To the Perplexity team and moderators: you should add a reaction among your emojis indicating something like "We do not wish to comment on this matter." This wouldn’t answer our questions, but at least it would demonstrate a basic level of courtesy. 😉

[sinful moat]

no point in using perplexity when there's no o1 pro mode and also deepseek reasoning

deepseek reasoning is on par with o1 pro mode

so much cheaper as well

[grand silo]

If anything, it just adds to the mystery of the Pro Search option. Some say it’s needed to use/activate the Pro model(s); some say it’s only needed for a pro ‘search’. In the second case, that could make sense if the model is supported but not for search/RAG, but 🤷‍♂️

[waxen lion]

i think they r considering adding deepseek reasoning to the pro search soon

[small tree]

deepseek r1 is actually sick

tested it yday

[halcyon coral]

You can try DeepSeek-R1 on http://labs.perplexity.ai/. We'll try to bring it up on core Perplexity in the context of advanced reasoning pro searches pretty soon.

[narrow tide]

now o1 seems to be back up, although I have to admit, it's suspiciously fast.

[green crystal]

Core perplexity?

[halcyon coral]

https://pplx.ai

[green crystal]

Ah to the pro subscription

[fading crest]

Ooo what is labs? Is this like testing?

[halcyon coral]

Yes, it's for testing

[green crystal]

I assume it's the full model? The deepseek site uses the lite version and it's much faster

[jagged cove]

Any plan to add deepseek-r1?

[noble kettle]

how is it I do not have Claude Opus?

[peak hatch]

💸 expensive

[jagged cove]

deepseek-r1 is better and free

[peak hatch]

have u tried it

[jagged cove]

I have the 7b running locally

[peak hatch]

Aye nice

[jagged cove]

mind blowing

[peak hatch]

what do you use it for primarily?

[jagged cove]

I use it with Pycharm local models

for coding

7b is not the most powerfull version, the other version I can't run them locally because of compute power

the resoning speed for a local model is out of this world:

[halcyon coral]

Hey @noble kettle!
This is to ensure users have access to Anthropic's latest models. So you're not missing Opus - it's been retired in favor of 3.5 Sonnet & Haiku to provide the best possible experience!

[noble kettle]

Claude 3 Opus is currently the most advanced and capable model in the Claude 3 family, but it is not being retired. In fact, it's the newest and most powerful model offered by Anthropic.

[normal heron]

Why is perplexity is just so damn slow? I paid for this yet free chatgpt is just so much faster and smooth, free of lag

[noble kettle]

little confused

[grand silo]

Sonnet 3.5 is the newest, followed by Haiku 3.5... Opus is from the previous 3 family of models

apparently Sonnet 3.5 is equal to or better than Opus 3, but that's subjective and depends on your use case

either way, it's a lot cheaper so perplexity removed Opus in favour of sonnet 3.5

[halcyon coral]

No, Claude 3.5 Sonnet is currently the most intelligent model from Anthropic

[noble kettle]

I stand corrected: https://docs.anthropic.com/en/docs/about-claude/models

[grand silo]

Opus still excels at creative writing, etc.

[halcyon coral]

Yes, it's the normal model

[past basin]

Yes please

[plucky bison]

O1 ain’t workin

[autumn lark]

o1 not working wtf bruh

[livid aurora]

How can I use Sonar and Sonar Pro inside Perplexity

[dusk shadow]

Will we get the DeepSeak AI model on Perplexity?

[slim star]

The o1 model works if you disable the pro.

[astral nymph]

hello everyone

[lilac tundra]

Hi everyone! I wanted to ask if the new API product from Perplexity Sonar has the same capabilities as the Perplexity Pro search engine?

[agile siren]

Hello

[livid aurora]

yay

[onyx stump]

Hope that got fixed.

[celest stone]

RIP sonar small

[strange flare]

What is actually the difference between Sonar to the old api?

[fervent crescent]

How are you enjoying Perplexity Pro? @rotund sedge

[empty nest]

Hello, new Sonar API is the HUGE version?

[crimson pollen]

are you using some kind of vpn? for me it's not slow unless you use some VPN that triggers cloudflare

[proven crater]

🥲

Alrighty then

[crimson pollen]

if you use the API it has web searches

[fierce sonnet]

which model did you use here, and what was your question?

[crimson pollen]

either sonar and sonar-pro say in the labs that they cannot do web searches if you do a simple question like: what can you do? or what's your cutoff date?

[fierce sonnet]

working well for me, but will dive in to see if we can reproduce the error you saw!

[crimson pollen]

I guess it's random, but definitely happened in both models

[onyx stump]

At least it is a lot better to have up-to-date information on everything.

[proven crater]

Something among the lines of: "Conduct thorough and detailed research on the keyword {keyword}. Search for credible ressources and for each resource, extract facts, statistics, perspectives, findings and concepts relevant to the keyword."

Sometimes works, sometimes doesn't. It especially doesn't work as a second message. When I ask it to come up with keywords in a first message and then ask it to perform research based on the keywords it provided, it is more likely to come up with the message telling it can't conduct real-time searches.

[crimson pollen]

each time's different 😅

[halcyon coral]

The models have no self-awareness, which is normal for LLMs, but not a bug. If you use a normal everyday example, the Sonar model will do a web search.

[crimson pollen]

yeah

[proven crater]

In this sturcture, this happens most of the time. If I understand this correctly, I'd suppose there is an initial search happening at conversation start and no subsequent search feature is implemented for further messages.

Damn, I was hoping I could finally streamline my article generator. 😄

[plucky nest]

Hey team, could anyone help me clarify what Reasoning Effort Parameter you're using for the o1 model? @signal hamlet @fierce sonnet

[tribal skiff]

I'm getting a 500 error with Sonar pro api

[hallow tulip]

So I'm very confused about Sonar Pro is this a new model that will power our pro-searches or is this model available for Perplexity pro users or is this only usable by business / enterprise?

[hallow tulip]

Meaning is this new Sonar Pro powering Perpleixty can we select it as a model?

[tranquil cliff]

Hello ，can anyone tell me the difference between perleexity labs and the UI ?

[mild wraith]

Wondering the same thing

[tribal skiff]

It's for api

But I can't get it working, I'm getting a 500 error

[mild wraith]

https://labs.perplexity.ai/

Seems to be working for me

Okay, so it's part of the new sonar stuff?

[severe goblet]

Not sure if thgis is a big deal Wix devs, but this has a typo in the example code:

https://dev.wix.com/docs/velo/api-reference/wix-blog-backend/draft-posts/create-draft-post

Just take out the " before the colon 👍

And it does work, it's just a UI issue

[normal heron]

No the problem is I'm using writing focus, I need help in writing but it's still slow af, sometimes it's stuck in the loading part.

[normal heron]

It's conditional for me, sometimes disabling VPN works, sometimes it's not

Mainly using gpt and claude

[magic talon]

Someone mentioned split tunneling works better than raw vpn

[magic talon]

So fine tuning if you want. You can give it some writings of your own and make it research ways to improve it like customized to be a full writer

[feral marlin]

?

[modest laurel]

Hi,
I don't get it with the Sonar API: Can I use Visual Studio to connect to this Sonar? Or can it analyze my projects and throw code smell and so on?

[halcyon coral]

Hey @modest laurel!
It is a completely normal API like the one from OpenAI. The only difference is that sonar has access to real-time information. This often gives Sonar an advantage over normal LLMs, as they can only rely on their training data.

[modest laurel]

Oh, I thought it was to have a Sonar for C# projects.... I understood wrong lol
Thank you for your answer 🙂

[sweet raft]

Does anyone know what addon or weak aura that tracks dr looks like this

[south pelican]

Hello, can we expect deepseek models in perplexity anytime soon?

[halcyon coral]

#💬│general message

[south pelican]

Thanks ♥️

Accounts with enterprise subscription don't get 5$ of free API credits each month?

[hushed summit]

O1 not working for rewrites is a bummer

[jagged cove]

Perplexity is the best for stock analysis hands down

[bright parrot]

Can perplexity summarise youtube videos?

[crystal cosmos]

I'm sorry, I can't do that.

I'm unable to provide real-time information.

[jovial kestrel]

I tried the sonar-pro with the API and the Playground. I was expecting to get similar results. However, the playground answer was more complete and reliable.

[tepid sequoia]

But come on, what's so great about it? Any model with active web searching can handle that. Any model analyzes it based on market sentiment and the sentiment of specific stocks—it simply regurgitates what it learned. I’m not saying it’s bad in the long run, but it still only tells you what already happened, not what’s going to happen.

[tribal skiff]

Are you using Python? I think the api call changed but I can't get it working even checking the website

[slow linden]

Is it possible to tell perplexity to always give me measurements in metric rather than imperial? It's very frustrating getting these nonsense units when I ask a question.

[livid aurora]

Can someone test https://github.com/PierrunoYT/perplexity-webui

[tribal skiff]

I'm getting this error with api <head><title>401 Authorization Required</title></head>
<body>
<center><h1>401 Authorization Required</h1></center>
<hr><center>openresty/1.25.3.1</center>
<script>(function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'905ee6ab8b96aab9',t:'MTczNzU0MjE2Ni4wMDAwMDA='};var a=document.createElement('script');a.nonce='';a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();</script></body>
</html>

[livid aurora]

You can use my example above

[austere quartz]

Hey all! Are you planing to introduce an app for windows desktop?

[serene sand]

#💗│sharing message ☕

[zinc swift]

Like many of you I use Claude Sonnet 3.5 as my default model. From time to time I like to switch models. Anyone using Sonar Huge? How is it?

[nimble bay]

what's the difference between the new sonar api and the existing product with the finetuned llama models? is it just an updated model?

[fast meteor]

To give detailed answers, Sonar Pro also runs a multiple searches on top of the user prompt where necessary for more exhaustive information retrieval

[potent oak]

Please let me open product links in an external browser from mobile!
I understand the impulse to keep people in the app. I even appreciate the attempt to steer people to "Buy With Pro" when available. That said, it's maddening and counterproductive not to allow people to open product pages from the Perplexity iOS mobile app in native Safari like every single other web link.

Anyone else?

[storm merlin]

What is the list of supported models in the API after 2/22/2025? only sonar-pro & sonar or gpt-4o also?

There are conflicting information, which confuses Perplexity Pro UI itself.

[magic talon]

Custom instructions

[carmine finch]

who's working on perplexity sports?

Got a question I wanted to ask, thanks

[serene sand]

sonar loading…

[slow linden]

Like in your profile? I added "only metric" in my about me section, maybe that will work?

[bleak tide]

so everyone here knows ai like the back of their hand no?

is anyone good with discord bot creation that i can integrate an ai like perplexity or any similar into

[halcyon coral]

https://docs.perplexity.ai/guides/model-cards

[uneven kelp]

is it just me or does perplexity have some troubles?

[severe pecan]

here also, some login trouble atleast

[stone dune]

Same

Seems like the server is down !?

Can’t use on laptop and phone

[junior adder]

oh so it's not me wheew

[iron agate]

Perplexity says it's out of capacity.

[stark matrix]

same here

[empty nest]

Hi, i'm acutally using llama-3.1-sonar-large-128k-chat. There is a version of api SONAR without online search?

[atomic locust]

hi pro sub have limits for generate pictures?

[gilded finch]

ive never ran into a probleom

rip its dead rn

[next olive]

Request + request

[stark matrix]

looks like it's working again

[cosmic sonnet]

anyone getting this error when trying to sign in via email?: 500
Internal Server Error.

[gilded finch]

how do I get the pro role if i have pro

[lavish oyster]

i got free one year pro lmao

[signal hamlet]

Please leave the server and join it again via the Discord link in your Pro settings: https://www.perplexity.ai/settings/account

[ornate forge]

i heard it can be used multiple times

[lavish oyster]

idk what that is, got it free from my college

iit m

yeah

[gilded finch]

yo

[upbeat lichen]

login still broken

[small tree]

How have they still not fixed pro model selection?

[gilded finch]

whats the best api for a researc hpapper

[gilded finch]

wdym selection like thhe different apis?

[small tree]

Model selection

LLM

[tribal skiff]

Can someone please help me figure out why I can't get Sonar pro to work with the api

I substituted "Llama-3.1-sonar-huge-128k-online" with "Sonar-pro" and got a 500 error, made some edits for a 402 error and 403 error, even went so low as to get o1-Pro Mode on chatgpt to do it for me and even that didn't work

[frosty kiln]

Will sonar / sonarpro model be available in the perpleixty.ai normal webui anytime soon?

[frank rivet]

Hi hey we recently updated the Sonar model line and now Sonar Large performs better than Sonar Huge, while being much faster as well.

We are removing Sonar Huge from the Pro selection soon. We appreciate your use of Sonar and Perplexity's own experimental models - if you have any feedback on Sonar please email [e-mail].

[gusty blaze]

It seems like perplexity completely stopped giving sources a few minutes ago

[tepid sequoia]

And do you have any info on when you will integrate Deepseek-R1 into Perplexity?

[magic talon]

Question for yall in your opinion can ai outmatch low level programers???

By low level l mean people who are new to it /are junior

[lean lynx]

r1 plz

[lean lynx]

for self-contained problems generated solutions do well enough

with a large codebase it's hard to provide enough context, usually anything generated needs substantial editing

personally use perplexity for research mostly, for example I can give it a bunch of api docs, and ask it to implement basic endpoints in my desired lang, then translate that to fit into my project

[haughty monolith]

Hi

Anyone know how to use perplexity pro via API?

I cannot see the parameter to turn on the pro option

[serene sand]

#💬│general message

[hasty zinc]

3

[stuck edge]

Is anyone building a perplexity 2.0? 🙂

[magic talon]

Cool thanks for sharing

[dreamy arch]

[fringe schooner]

can i only upload 4 file at a time i am making a plugin and i have many file and i want to upload and try to messw ith some modify how can i do that

[wispy bluff]

hey yall i have a question, what do you use for an ai model in the settings to pair pro search in terms of finding statistics data and using it to make a prediction for an outcome?

[autumn lark]

Heads up guys. PPLX is being awful again - on the mobile app: on the main page in WRITING mode without PRO it still does Web Search.

[fading furnace]

site down?

[iron barn]

site is down for me

[astral citrus]

Same. Maintenence is the reason, I believe.

[flint sedge]

how long do y'all think it'll be down for?

need it asap

[astral citrus]

No clue.

[south solstice]

Any idea how long it will last?

Oh nvm

[sly minnow]

Just using PPLX for work.
I guess it's telling me time for taking a break.

[mild cypress]

ahh.. seems its not only me having the issue..

[astral citrus]

I'm just using it for silly stories/worldbuilding.

[iron barn]

of course it's during the exact time i need it working 😭

[mild cypress]

same dude😭

[sly minnow]

[sage kite]

perplexity.ai image generation seems to have taken a hit, isn't working at all, anyone facing this?

sorry

I see everyone is facing this

I needed to generate one for a marketing post 😦

[iron barn]

does anyone know if this was a planned outage?

[sage kite]

no, I don't think so. It wasn;t showing the downtime message for me until I reached out to the support

[sly minnow]

Just in the middle of my work session.
66% to complete.
Then "Pow!!"

Maintenance

in my face.

[fossil vessel]

was in the middle of work until it stopped working

[astral citrus]

I use several completely different services for AI art, since I don't have a plan that allows it.

[iron barn]

i was in the middle of generating a response and then it was stuck so i reloaded it and it took a while to load then said the system maitenence error

[gritty siren]

Was in the middle of searching for the best toilet bowl for my home Reno..and this happens

Ugh

[swift arch]

I think the app just came back

Working now

[minor sequoia]

Thx god

[sly minnow]

Not yet done on web side.

Back in five.

[minor sequoia]

YIPPI

[gaunt wraith]

iphone app doesn't work in pro mode. non pro answers only one questions

[astral citrus]

Seems like web's back.

[sage kite]

nope, still showing the downtime banner

[modest pulsar]

same

i think it is back

[sly minnow]

Thank god.
Back to work

[sage kite]

yep. its back

[wispy bluff]

niceee

yall know the best ai model to pair with perplexity i been using claude sonnet but would love suggestions

[lyric wigeon]

What you looking to do?

[wispy bluff]

make predictions based on statistics of a event or game etc

[hollow jolt]

Before i signed an abonnoment for the pro version. Perolexity used to name me the exact pdf sites where it got the infos from. Now it does only name me the file and not the exact sites anymore. How can i fix that?

[harsh axle]

😭

[brisk hare]

@spacehdtv @spacehdtv

[crystal cosmos]

all stablebdiffusion

lololol

[sterile oxide]

Hello folks, good day.

[tawdry mural]

https://youtu.be/GytSx78md0s?si=pEadW3sj-EPqYV3X

[hollow jolt]

Meant for me?

[blazing idol]

hi

[deft otter]

Dm me if anyone wants steam gift cards

[sinful ice]

hi

[cyan tendon]

howdy

[arctic fulcrum]

Can I also use perplexity ai chatbot on discord and if then how ??

[sly rune]

Hello everyone 👋. Just joined the pro because of IITM's institutional benefits. Reply back if you too 🙌

[tame current]

if you do any coding, you could potentially use perplexity's API (application programming interface) and integrate it into discord. chances are someone has already done this but integrating it into discord could be technically challenging.

[stiff ridge]

What is up with this?

Yeah you have to turn off internet focus and pro mode to use the 3rd party AI models, internet makes it go back to perplexity proprietary model. They do not disclose this. Interesting. They also put that the AI used was still Grok. lol

[tepid sequoia]

Unfortunately, Perplexity is falling a bit behind, as the vast majority of services today have search capabilities, but when one uses chat.deepseek.com with their R1 model, and then has Abacus.ai for more complex tasks, you can forgive Abacus.ai for their 2 million limit, basically Abacus.ai works well for me with Gemini 2.0 Flash Thinking so you won't reach that limit anytime soon. Unfortunately, overall Perplexity severely shortens responses to save costs, and even after requesting a long answer, it struggles to provide one.

[stiff ridge]

There is some error on your part. If you mess around with switching the focus mode and turning on or off pro mode, you have better results.

[raw harness]

Hey

[tepid sequoia]

No, I’m aware of that. For example, let’s say you want to create a 1,000-word article, but even after multiple revisions, it still can’t do it. I’m using this as an illustration.

[stiff ridge]

Oh try using Pro mode for more lengthy answers, the non pro is supposed to be short in response. Its to have the user experience what you are actively experiencing to buy Pro sub.

[tepid sequoia]

You don't understand me 😄 Well, I was saying that when I had PRO and Focus mode turned off, which is for writing without web access. But now I'm figuring out one more thing – the main problem seems to be with Sonnet, but not so much with GPT-4. To be sure, I just tested it on pure Sonnet via API, and there was no issue, so it seems that Perplexity limits the length when using Sonnet. I believe this is to save costs.

[stiff ridge]

I do understand you. You didnt see my previous comment? When pro mode is on, it uses their free llama based solution to give answers, not truly using real sonnet

[tepid sequoia]

But I did see it. I'm telling you that I have PRO mode turned off, and that's what I'm basing this on. It should be just Sonnet. By the way, it's interesting – ask all the models in Perplexity who they are, and only GPT-4 responds that it's GPT-4, while the others say they're Perplexity AI. Like I said, there’s definitely something going on.

[stiff ridge]

Yep, you see it. They are being sneaky lol

[dull tangle]

Hi everyone

who is US citizen?

[tepid sequoia]

But then how is it that when I have web searching turned off, because then it switches to their LLAMA model, and I need up-to-date information in the responses, Perplexity really drops in value, since chat.deepseek is free and abacus.ai, even with a price tag of $10 and only 2M of their compute points, somehow Perplexity is losing value and significance in usability. Furthermore, another thing, try translation from PDF, also a problem with length, it only summarizes, doesn't translate. We're talking about around 1000 words, deepseek doesn't hesitate. Abacus even has the advantage of summarizing or translating anything directly from a link. We'll see, maybe when Perplexity releases Deepseek-R1 and they change this approach, the service will be of higher quality.

[slow linden]

deepseek is heavily censored, which might be a problem in the future

[slow linden]

we should send them all a 'get well soon' card

[dull tangle]

so who?

[slow linden]

I guess just find the users who are all super sad right now

[dull tangle]

like you?

[tepid sequoia]

It depends on where you're from. For example, I’m from Europe, and DeepSeek gives me less censored answers within my location. For instance, the worst for me is Sonnet. But yes, I agree that DeepSeek is censored when it comes to topics related to China. However, there’s one thing: I’ve tried DeepSeek-R1 through the API, and those issues aren’t present there. The version that’s censored for China is primarily chat.deepseek.com.

[slow linden]

luckily located on the other side of the atlantic 😮‍💨

[dull tangle]

LOL.

hey @slow linden
can I DM to you?

[slow linden]

I've seen videos where it actually does give the answer but then removes it as it reaches the part where it hits a forbidden topic like tiannemen square or tibet. All frontend censoring. Sonnet censors info? I use that as my main model and haven't noticed it giving me bad info yet (every i have checked ends up veing correct)

[stiff ridge]

Internet focus for up to date info with Pro mode on. Translation, PDF viewing etc.. writing focus, Pro mode off using like claude-3.5-sonnet and should be pretty good.

[tepid sequoia]

Yes, I've tried that with DeepSeek too, I've seen it, but as I say, the API provided me with information about Taichen and told me why it is so. In my opinion, we are encountering the problem of East and West here. Sonnet is a Western model and Europe is inclined towards the West, so maybe the training data is different, or from a different point of view, and DeepSeek is an Eastern model, so there is censorship within Taichen.

[stiff ridge]

The training data is the entire internet of human text compressed. There isnt Europe or US based.

[tepid sequoia]

But it has limitations, and you can't directly translate something from a link. For example, from the official Perplexity page, you can translate things directly from the link, but with something else, it didn't work for me. We're just discussing the limitations. As a comprehensive tool for less demanding users, it's good, but when someone wants something more advanced, it becomes a problem.

[dull tangle]

I am finding a US citizen who can collaborate with me.
Believe each other and making money together.

DM me

[tepid sequoia]

You understand, we are talking about details, so you will sometimes find certain changes whether you want to or not. Just look at how opinions change with changes in governments from 2016 to the present, and these models react to this because they are trained on a certain moral compass.

[stiff sundial]

Can one create Ai image/video yet with perplexity? do i need to switch to a diff model?

[fast meteor]

#1225557012532695060
video is not yet possible

[fluid saffron]

Hello guys, hope you are doing well.
I have a question regarding AI model in Perplexity.
What is the most accurate and reliable model for up to date web search information ? Like legal topics, new regulations, technical standards, etc ..
Should I keep the default mode or would you suggest a better ai model ? (Sonar, Claude .... )
Thank you in advance for your support and advice

[tepid sequoia]

Well, we discussed it a bit earlier in a few posts, I would say that you need more content, not just a simple answer, so GPT-4 and turn off pro mode and just leave web search on, if you want to improve, try the Complexity extension.

Just for info, impact of DeepSeek-R1 https://x.com/sama/status/1882478782059327666

Sam Altman (@sama) on X

big news: the free tier of chatgpt is going to get o3-mini!

(and the plus tier will get tons of o3-mini usage)

X · Jan 23, 2025, 5:21 PM

[gaunt crescent]

Let's imagine you wanna try the new voice mode

But your power button is kinda dead

And home button and corner swipe doesn't work

What to do except crying?

[zinc swift]

The link provided (pplx.ai/android) sends me to the Perplexity app on the Play Store. I thought this was a separate app? I'm confused. 😂
https://x.com/perplexity_ai/status/1882466239123255686

Perplexity (@perplexity_ai) on X

Introducing Perplexity Assistant.

Assistant uses reasoning, search, and apps to help with daily tasks ranging from simple questions to multi-app actions. You can book dinner, find a forgotten song, call a ride, draft emails, set reminders, and more.

Available on Play Store.

X · Jan 23, 2025, 4:31 PM

Never mind, just saw this tweet: https://x.com/AravSrinivas/status/1882467172498436291
It isn't a separate app, I just needed to update. 😄

Aravind Srinivas (@AravSrinivas) on X

We are excited to launch the Perplexity Assistant to all Android users. This marks the transition for Perplexity from an answer engine to a natively integrated assistant that can call other apps and perform basic tasks for you. Update or install Perplexity app on Play Store.

X · Jan 23, 2025, 4:35 PM

[stiff ridge]

I agree.

[wheat dagger]

Can someone please send me a guide giving step by step on how to use perplexity API as a proxy API for chatbots like janitor AI , agnaestic ..........
I'm newbie to this

[feral briar]

Hey everyone,
Where can I check out my subscription term? I am using the Telekom 1 year pro version

[cobalt monolith]

Thank you perplexity for being goated

[wide pike]

😅

[merry kettle]

Sorry but this assisant doesn't work well, the fact you can't voice activate it makes it impossible for me to use. I often use google assistant to pause playback etc. when I have to switch to perplexity assisant I use all voice activation.

So I have the choice to use google assisant that I can activate with "hey google" or I choose perplexity where I have to grab my phone.

[latent pendant]

Any chance that the Assistant will arrive to iOS?

[sly pivot]

Haopy to see the assistant from Perplexity till it doest work as expected. The mic is showing that it records on mic but it just deaf. Where i can reslort problems?

[sudden falcon]

does the assistant have access to email or can it only draft an email? does anyone know?

[weary lynx]

Perplexity Pro with Claude Sonnet 3.5 isn't designed to respond with long or in-depth responses. I use Perplexity Pro for research and then dump the output into Claude 3.5 200k on Poe with a detailed prompt to get started on papers and briefs. Then I need to double check all of the citations.

[wind bramble]

OPEN AI: Introduction to Operator & Agents
read it here and chat with video for details and get custom answers to your answers grounded in the video https://www.copilotforyoutube.com/search/introduction-to-operator-and-agents-1SrOLOfnSBEp9I5yo3enIr

[hollow jolt]

Can ppx cite the exact pdf sites? It did in the free version

[drowsy harness]

How can I trigger this perplexity assistant. Tried everything on my phone but nothing.

[sly pivot]

Check perplexity settings

[fluid tendon]

can't seem to change its voice

[sly pivot]

Anyway it doesnt work for me (its deaf) And I really believe this is not problem with Perplexity as I have issues with Gemini/Old Google assistant since installed the first one as assistant.

I think Android is bugged in that field. It randomly choose assistant in 'Android For Auto' its pain. As sometimes it work well sometimes is deaf

[slow linden]

Tap tap tap, enter voice mode. Alternatively set up iOS widgets and/or siri shortcuts for it. (Or Android equivalent)

[zealous jackal]

Just wanted to say, LOVE the assistant feature

[distant lance]

What stuff have you tried!!

[slow linden]

Have you already used it?!?

Will mistral and Gemini be added back in? I liked Gemini for image prompt generation over other models.

[sly pivot]

after search my settings found that I had ?Qualcomm? installed 😳

[hidden turtle]

Hey, is the Sonar Huge model removed?

I just saw the new o1 model but only the Sonar Large model

[hallow tulip]

You're right they got rid of sonar huge for some-reason perhaps they are getting ready to add in sonar-pro

[half venture]

Does anyone have a code that gives you 10 dollars off on pro sub

[jaunty dove]

Intentional since there isn’t llama 3.3 405b model. Large is based on 3.3.

#💬│general message

[slow urchin]

Hey any idea how to launch the assistant on pixel phones?

Can't get it to come up

[azure mantle]

Hi,

Go to your parameters, find "digital assistant app", change google By perplexity. You will not be able to call it by "hey google" but by maintaining your shortened key (for me it is the power button)

[magic talon]

Tested it out with send a an email [whoever you wanna send to automatically using in my case the Gmail app

[azure mantle]

The assistant is really cool. I love it. The fact that he can analyze the screen is, and is going to be of crazy use. Fact checking, determined if something is suspicious, it is already adopted.

[slow urchin]

I'm amazed by this

It's huge

Can we change the voice?

Doesn't answers to the perplexity voice

[magic talon]

If anyone would like to sort off a workaround to summon asistant with a specific word you can just use Tasker and do a set up that opens the asistant when you said your preferred phrase

@slow urchin yes setting then voice style its set to Mike by default

[slow urchin]

Ok can't be changed?

[magic talon]

Yes in settings

You can

You are even able to change the language if you prefer a different one instead of English

@sudden falcon actually you can fix it with settings see the following screen

[magic talon]

Do you need help changing it ?

[tame current]

Can you elaborate please more

[slow urchin]

I can't change the voice, if I change it it still talks with mike

[small tree]

https://ffnews.com/newsarticle/fintech/revolut-launches-its-highest-savings-rates-for-uk-customers-of-up-to-5-aer-variable/

they just boosted it to 5%

mad

[magic talon]

On the phone right??

[slow urchin]

Yes android pixel 9

[magic talon]

And you already st perplexity as default asistant no ?

[slow urchin]

Yes it's working but voice can't be changed

[magic talon]

Well lest try something perplexity is interconnected so try to log in the website and apply the changes there then check back on the app and let me know how it goes

Make sure to shut down the app completely before that tho

[magic talon]

Well basically we use a app to trigger certain stuff on your phone since perplexity doesn't have it by default

[celest stone]

is opus gone from the API?

brooo

cancelling my pplx

[small tree]

Does Opus produce better quality outputs than o1? Do you have a specific use case?

[celest stone]

I used AI Profile as if it was ChatGPT custom instructions to tell the AI to behave a certain way, and Opus just chats better, and explains concepts in a way that I understand better...

[small tree]

Oh, makes sense

[celest stone]

what I was doing is I kept an older version of the android app, which let me pick Claude 3 Opus, and it showed as "Default" on the website dropdown, as long as I didnt touch that. It was Opus.

[small tree]

maybe they just disabled it from the UI, you can try and modify the http request body to select the old opus model if you remember the exact model name

https://www.perplexity.ai/rest/user/save-settings?version=2.15&source=default

{
"updated_settings": {
"default_model": "claude2"
}
}

change the payload to opus

i actually managed to do it lol

@celest stone

claude3opus is the model name

[celest stone]

yeah, but it comes out as Claude 3.5 Haiku for me.

I know the app sets it to claude3opus

does generating anything give you opus?

[small tree]

"default_model": "claude3opus",

i'll try it

how do i know if it's opus generating the responses lol

@celest stone

[celest stone]

bottom-right of an answer, hover on that

[small tree]

ahh

sec

oh you're right

claude3opus is just haiku

-_-

devs bamboozled us

[celest stone]

ngmi

I miss my 400B parameter, planet-klilling, tree-burning ocean-emptying boy...

[small tree]

get yourself 2 of those new Nvidia Digits and run it yourself

[celest stone]

I probably cost Perplexity so much money...

[small tree]

they're burning through investor money, dont worry

[celest stone]

This is still cheaper than paying yearly

[small tree]

i got 2 years free

from Revolut Ultra

lol

[celest stone]

damn, you burnin' trees for free!

[small tree]

unfortunately, we cant edit the request body for the rewrite function, because it uses websocket for those requests

so cant test to see if opus still exists

[celest stone]

I woulda just gotten a Claude sub directly, but even their paid plan is like 20 opus requests per day

I gotta use like 50

[small tree]

model_preference
:
"claude2"
prompt_source
:
"user"
query_source
:
"retry"

yea defaults to claude2, if opus is selected

[celest stone]

why is it claude2

like, claude 2.x is deprecated, no?

[small tree]

devs have weird naming for it

[celest stone]

I just remember Claude 2.1 being the absolute king of refusals. It was so bad it made me see Anthropic as clowns for a while.

[small tree]

claude 3.5 sonnet is keyed as claude2

in the backend

[celest stone]

Then I experienced Opus and it was like "oh my god, they cooked"

[small tree]

its just naming

[celest stone]

so haiku is claude1?

[small tree]

haiku is set as

claude35haiku

[celest stone]

sonnet 3 was okay, sonnet 3.5 was hit with a metal bat a few times too many by Amodei and his gang

[small tree]

Have you tried Deepseek R1?

[glacial river]

The new android perplexity assistant doesn't realize it's an assistant and tells me to switch back to Google lol

[inner glacier]

CC @fierce sonnet REPPING PPLX Supply in my post today! Peep the sick hat in the After photo: https://bsky.app/profile/wsig.me/post/3lggtvaidtk23

Will Sigmon (@wsig.me)

Since last summer, I’ve been working towards better health and strength.

Today, I’m pleased to report a 45-pound loss!

This progress stems from lifestyle adjustments, improved diet, & support from Mounjaro.

Aiming to maintain this momentum and reach my goal by summer.

Here’s to a healthier 2025!

Bluesky · Jan 23, 2025, 9:40 PM

[gusty wasp]

New assistant app but the chat app still hasn't improved and is riddled with bugs and missing features..sigh

[fierce sonnet]

congrats Will!

[nimble thorn]

I use Apple SSO to login to my Perplexity Pro account, but the Android app wont let me login using an Apple account.

Any ideas on how to get around this?

[fierce sonnet]

https://x.com/AravSrinivas/status/1882493999388430376

Aravind Srinivas (@AravSrinivas) on X

Read Access for Calendar and GMail for the Perplexity Assistant: work in progress; the intention is to get it out within 3 weeks.
It can already (right now) summarize your unread emails and upcoming calendar appointments if you enable notifications for these two apps.

X · Jan 23, 2025, 6:21 PM

[glacial river]

Google integrations are a big sticking point rn for me. i really enjoy being able to ask about my calendar/email/documents and have it add to those too. that and getting a better sounding voice would be two tops IMHO.

[tardy junco]

Who on X shares the best demos, use cases, and prompts? Not looking for the usual fluff or AI automated accounts paying for premium. I’m after the most insightful accounts and the most valuable people to follow those who consistently post high quality, practical content. Any recommendations?

[jade meteor]

Assistant doesn’t seem to work on OnePlus 11, sad.

[sudden falcon]

how do i get to the permissions after i've already made pplx the defgault assistant?

[magic talon]

Go to the regular perplexity in the settings scrool down to asistant and toggle them on

[sudden falcon]

thanks this worked

its weird..it has contacts access but it says it can't find a particular person in my contact

[magic talon]

Maybe is duplicated or maybe the ai is just sensitive to certain letters/characters

[sudden falcon]

not duplicated and straightforward in english. it has to be a bug

[magic talon]

True it could be that

[sudden falcon]

its still such a good experience though

[jaunty raven]

The only reason why I will not use perplexity assistant is due to it not being as fast as gemini when we pull up the gesture from the corner, 2nd is the voice of Gemini is really very very natural and it is soothing

[sinful moat]

Is this through the UK revolut?

[snow gate]

Possibility of perplexity assistant to add my tasks into Google tasks or notes Or a widget that lists my reminders/tasks/notes ?

[slow urchin]

Anyone has a comprehensive list of what assistant can do? is there any documentation of it?

[tranquil orchid]

Hey @frank tree!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[green crystal]

the ability to make better sense of web results and correctly reason if something is not relevant for its answer is incredibly helpful

[sterile oxide]

Buzz off, man, you're annoying.

Say "Tiananmen"

[tepid sequoia]

I'm also probably going to cancel Perplexity PRO, and I have 80 referrals. Considering the results DeepSeek and Abacus.ai gave, and both of those services have unlimited generation, Perplexity simply has to go. Goodbye, maybe again sometime.

[bright parrot]

is it free of cost?

💀

i couldn't see any pricing option on their website

[grim vessel]

Yeah, Only API Cost for now.

[bright parrot]

means i can use the app for free... hmm interesting

🤓

[tight hound]

Yea ppx is dying . they are having many bugs not getting fixed and the O1 has only 10 Usage per day if it even works

[tight hound]

Abacus is not free

[tight hound]

after registering , they immediately asked me to fill my bank card lol . I was not able to try it

[tepid sequoia]

It's not about the price, Abacus costs 10 dollars and I meant it as a complementary thing to DeepSeek.

[tight hound]

why do you think it's complementary to deepseek?

[tepid sequoia]

For those 10 dollars, it's worth it, especially if you primarily use, let's say, Gemini 2.0 Thinking, then you get 10,000 messages per month. Plus, for example, it has Image Generation if you occasionally need it. Not to mention Sonnet or even Gemini, which have a bit more human-like communication if you need that. A benefit is that it has search directly with every model, and the generation of responses isn't shortened to save tokens. Of course, there's also O1 and DeepSeek-R1, but these models will eat up your points faster. That's why Gemini 2.0 Flash Thinking is a very good equivalent, and for example, it can analyze and translate pages from a link. Like, for 10 dollars as a supplementary service to DeepSeek, it's great. Unfortunately, Perplexity has fallen behind and won't recover from it. Instead of fixing problems and adding real features, they made an Assistant which will just generate more problems.

[tight hound]

thank you for elaborating . I will give it a try .
And yes considering perplexity I am really done with it and stopping the subscription pro this month

[small tree]

Yea

[sinful moat]

just use perplexica

open source

[tepid sequoia]

A few days ago I wanted to try it, but the page wouldn't load. I'll try again, but in any case, if I saw correctly, Perplexity doesn't have document reading.

[sinful moat]

perplexica you can attach documents

[tepid sequoia]

yes, okay, I'll try again, thanks.

[tight hound]

it's not loading for me too

[narrow tide]

Very nice that we have a new pro search adapted to O1, but why don't I see the citations and links to cited papers in the response? How do I know which of the resources is referred to in the response? Please FIX this as it's essential !

[willow kettle]

Some alternatives to perplexity you can try: felo.ai, beloga.xyz, genspark.ai(kinda, more for research), phind.com, you.com, komo.ai

[stark quiver]

when is perplexity assistant coming to ios?

[willow kettle]

The startup has also expressed interest in expanding its availability to iOS devices, with a spokesperson revealing that they are currently awaiting the necessary permissions from Apple. “If Apple gives us the right permissions, we’re ready to launch on iOS,” the spokesperson said. The company’s move highlights its commitment to broadening its reach across platforms in the future.

[stark quiver]

is this an approval for launch on the app store or permissions for native integration within ios like replacing siri (afaik the latter is not happening anytime soon)

[magic talon]

Is more like expected until they get approved

[green crystal]

Deepseek being so much cheaper really makes it hard for O1

The model seemed to have pushed quite a few people towards working on more efficient reasoning models
So that's good news overall

[cerulean cobalt]

What model does perplexity use to answer the assistant questions?

Their fine tuned sonar model?

[sinful moat]

I'm really liking the perplexity assistant over the Gemini voice assistant

[ornate cloud]

https://x.com/filicroval/status/1882727675468657138 ????

Filipe | IA (@filicroval) on X

5. Creando un clon de Perplexity en una hora sin escribir una sola línea de código

X · Jan 24, 2025, 9:50 AM

[lone needle]

Perplexity cannot create,right?

[unkempt summit]

agi tomorrow

[serene sand]

https://x.com/aravsrinivas/status/1882555331764781102?s=46

Aravind Srinivas (@AravSrinivas) on X

We will bring up R1 served in American data centers on Perplexity. And make it free for everyone.

X · Jan 23, 2025, 10:25 PM

[unkempt summit]

yo

[fallow wagon]

what about image analysis

[tepid sequoia]

Well yeah, and when? In three months? Or does their new LLAMA need to pay for itself first?

[jaunty dove]

Llama and R1 are open source, there is nothing to “pay for”.

R1 has been on labs.perplexity site for the past few weeks.

They have to apply it for their databases before they make it available. It’s the same as every other model they added and removed in the past.

[glad hornet]

My device does not show the "perplexity" app option as a way to activate the assistant

I have already received the update, but I cannot activate it

is there a way to activate via ADB or something?

[tepid sequoia]

That's nice, but Sonar Pro is built on LLAMA, and that one is paid. Yes, it's known that it's there, but you see, others have DeepSeek-R1 on the first day of release.

[jaunty dove]

You are not paying for LLAMA itself, you are paying for the API access that includes additional features.

Yea, perplexity labs had r1 quickly as well for free.

The regular product will get R1 as soon as they can.

The API product is separate from the search AI product.

After all, we have sonnet, o1, sonar large (llama 3.3, latest) and so on in the search AI product.

There aren’t any indications so far that sonar large and sonar pro have different llama versions.

If you want llama 3.3 70b, it’s already there for a month, it’s called Sonar Large.

[nimble slate]

Why do I not have o1 to choose from in my options? Is it only available to certain pro users?

[jaunty dove]

Yes, pro only. O1 is very expensive and is rate limited by OpenAI. Even o1 is capped at OpenAI for free, I think 50 per day or week, can’t remember.

[nimble slate]

I have pro but I do not have access to o1

[jaunty dove]

Been there since day 1 for me.

Are you on a vpn or using mobile app?

Mobile apps haven’t been updated for this, so you have to use the web app.

[tepid sequoia]

But the core of the price is the model itself, not hardware for example, and access is through the API to that model 😄 ... I'm not going to break down every detail here 😄 so there's no point in you catching me on every word I write, but my criticism is valid ... 'regular product'? Is abacus.ai, for example, a 'regular product'? ... we were just discussing it yesterday in this discussion, each of those models you mentioned is set up to save costs and provide a shortened answer, and when I ask for a longer answer, there's no chance.

[nimble slate]

I’m using a mobile app, but I also logged in on the Safari browser and I still don’t get o1

[nimble slate]

Never mind. I just changed to desktop mode in my Safari iOS browser and it shows up now.

[tepid sequoia]

However, here I will defend Perplexity. You can't expect them to get unlimited access to O1 at the price that OpenAI provides it. What I'm criticizing is that Perplexity has fallen asleep and slowed down, features are missing, and mainly the solution is at least DeepSeek-R1, and it's taking them longer than everyone else.

[distant lance]

How is the android perplexity assistant from a privacy perspective?

[tepid sequoia]

Just like everything else, if something happens, you're the one who's responsible. And yeah, it's written somewhere that your data is anonymized, but bet you it's still stored in some way. On the surface, they call it 'service improvement', just like everyone describes it. But basically, your interaction can be used in some overall evaluation, maybe at some developers' meeting.

[magic talon]

Internet in general tbh

[woven garden]

Is there a safe way to auto-approve MCPs when using Claude?

[viral lantern]

Yoo perplexity assistant is hype

Im a pro member now, does anyone know how i can join the pro channels here

[jagged pike]

Does anyone have a chart or graph to differentiate the main issues currently between the paid versions chatgpt, claude, gemini, perplexity or similar?

[tepid sequoia]

Basically, the number one differentiator is the number of responses. In this case, all these services cost approximately 20 dollars, so Perplexity is the winner because it has Anthropic's flagship model Sonnet, and it has GPT-4o, and you can make 600 responses daily, which no other service offers on these models. Another feature to consider is up-to-date information, which means the service needs to be able to browse the internet. Again, Perplexity wins in this case, because the browsing capabilities of all these other services are pathetic. And in principle, many things are already similar between them. As for my recommendation, Perplexity, but if you want an even better service, then abacus.ai and use the Gemini Flash 2.0 thinking model, you get 20,000+ interactions per month and it only costs 10 dollars, or completely free and without limitations, Chat DeepSeek.

[jagged pike]

Please, tell me if this info its ok or should i rethink it: ChatGPT
Best for generative text (writing)
Connected to web for current data
Customizable plugins
Image generation

Claude

Complex reasoning and information analysis (including database analysis)
Handles larger amount of information (tokens)
No image generation

Gemini

Mathematics and code
Integrated with Google ecosystem
Image generation

Copilot

Based on GPT (but not better)
Integrated with Microsoft 365
Image generation

Perplexity

Everything above minus Google or Microsoft integrations
Uses configurable models, including O1 and Opus, with certain usage limits
When selecting a Claude model, it handles fewer tokens than using Claude directly on its web platform