💬│general

[narrow silo]

Tho go to top button is nice

[silent pier]

oh

alr, thx

[thorn magnet]

is the api down

[thorn magnet]

the reasoning api is 502

and timing out

[keen prairie]

hi any1 knows a vm for macos? free one

[verbal whale]

Hey guys, I’m not sure I understand how perplexity work. Could someone help me understand clearly please? Especially on the fact that there is kind of two AI choices, like the model and for example deep research. Like what is it all about? Why isn’t there only one category? And maybe what is the best combination with the two options? Thank you!

[surreal marsh]

no im a random student but i can tell, like perplexity can, that maybe they shouldn't go in this direction

[spice prawn]

A little while ago I got a 502 on the reasoning api as well. Status is all green 🤷‍♂️ : https://status.perplexity.com/

[wind bramble]

New video to text thread from Theo - t3․gg
direct link to chat with transcript https://www.cofyt.app/search/ai-models-a-race-to-the-bottom-dr26pZ9-QdhhW6NHurTY_Q

[unique umbra]

Harassing github copilot for hours

[indigo valve]

So they added perplexity to Windows.... so it's just an electron wrapper?

[unique umbra]

Yeah..

[indigo valve]

It doesn't say it anywhere but out of curiously if you hit CTRL + I in any window, it will bring up perplexity search

@unique umbra

I activated it just now by accident

[unique umbra]

I'm on meh phone..

I got no clue

[indigo valve]

womp

[drifting jewel]

hello guys I'm a new pro user for a few hours, after months and months on chatgpt I switched to the prodi perplexity plan, do you have any advice on how to "give" information on the work I do, I'm a self publisher

[unique umbra]

There's a paper clip

Icon button

And or just tell it the links u want it to search

[indigo valve]

You can create a "Space" and give it knowledge and links

[patent cargo]

Anybody having this issue in mobile

[brisk jackal]

guys

what is the best ai model for creative writing?

gpt 4.5?

or claude 3.7?

[remote blaze]

I need help writing a resignation letter

[south kindle]

I’m seeing that as well.

[hushed harbor]

Yeah on chrome and safari

[hard flame]

.

[silver bay]

list your priorities and then create test prompts and see how those models react to them

then try to fine tune the results through instructions

models have their own quirks, find the one that suits you better through testing

[brisk jackal]

thanks

I ended up using claude with reasoning

it's mindblowing

[brazen otter]

I just started using Perplexity to summarize my Discrete Math I chapters and it's cut my studying time in 1/2

Wish I'd done this sooner!

[unique umbra]

😭

[indigo marsh]

hi

[silver bay]

hi

[daring kernel]

so my school reached the 500 subs but i didnt get 1 year i only got a month anyone know why

[unique umbra]

Bro double posted

[daring kernel]

I didn't know where it was more relevant and where i could actually get support

[short coyote]

Has anyone noticed perplexity mentioning your location mentioning much more than it should? Its trying to include it in all my answers, where they are not needed at all

[silver bay]

maybe check what's in your profile?

[vapid onyx]

I’ve personally seen it use it in relevant situations. Specifically, I recently asked what type of bug was in my image and it identified it and mentioned that it correlated with my location

But beyond that, not really

[short coyote]

hmmm possibly

[raven night]

Is auto less reliable than the pro for quick questions, if I have the pro subscription?

[distant aurora]

Pro has more thorough analysis.

Don't use auto unless you hit your limits regularly.

I will never understand how people hit the 500 searches per day limit.

[zenith dragon]

Do the server downed?

[elfin portal]

What is the usecase for the windows app over the website?

the windows app is tiny - it just looks like a shell to run chromium to access perlexity - I cant see it has any additonal features

can it do tasks on the host machine now?

the mic button only disables or enables the mic, its handy to have it there but how do we initialize voice mode - its on the app

so it wants to have access

and technically does

just when I select the mic icon it either mutes or unmutes I cant see anywhere to input my audio - yes I have tried just shouting at it 😄

[gusty helm]

is sonar pro working properly? ever since the latest ui update I swear it feels like i'm just getting standard sonar responses (much quicker and of less quality overall), even with pro and sonar selected :/

[gusty helm]

and sometimes it seems like I do get sonar pro responses, but only with the new auto option enabled

[elfin portal]

thats just the mic icon for reference

I have pro, and can do pro searches on the windows app, and voice to voice works on the mobile app

so access wise I should be all good

[foggy lintel]

It's both frustrating and concerning, that users are unable to disable or tune the "New Search" page, where we are getting political and religious topics presented to us before we can even use the tool. Especially as a paid subscriber, I shouldn't be getting unwanted ads / articles pushed at me.

I'm an AI power user, and work daily with many different AI platform models. I've happily paid for perplexity to use functionality that doesn't exist on other AI models. This "feature" is so distracting, that I'm leaning towards cancelling my pro subscription and moving off the platform (API only) until this issue gets addressed.

Come on Perplexity - please prioritize getting in front of this.

[elfin portal]

Jamcam have you used the windows app, and if so did your mic work

I think the windows app has some issues

specifically the mic doesnt do anything - users can mute or unmute a mic with no purpouse because the app cannot hear you anyway

it is low impact at least thats a good thing for the app

[crystal cosmos]

XD

[elfin portal]

you had any luck with the app?

[crystal cosmos]

HAWK TUAH

[elfin portal]

霍克图亚

[crystal cosmos]

i dont use perplexity for 200 years ago, but i dont suggest you use exxternal app because the update they push will not be at the samce pace as their updates in the core website.

i suggest not being lazy and just go to perplexity.ai in your browser

[elfin portal]

fair enough

[crystal cosmos]

i dont know if you can recal but there used to be lots of complainers here in this server from the consumers that the app is so buggy like when they press a button it reverts the ai model to default when the string says opus unlimited.

[elfin portal]

the mobile app is good now its just the windows app

[crystal cosmos]

boop

[raw maple]

hi

[raven night]

Thank you so much!!!

[distant aurora]

I suggest you use r1 if you select pro.

[copper current]

4.5 hands down, its slower tho

[distant aurora]

They're never recovering from this.

[brisk jackal]

I had quite a complex prompt and used 3.7 with reasoning cos 4.5 couldn't understand it

[elfin portal]

yeah so the app shat itself

managed to get it to happen again, if you edit a page and have it re-write it the app crashes, id share logs but i cant see where they are stored

[slim delta]

Hi community! Does anyone know if there's a way to try Perplexity Pro for free? Is there a trial period? Thank you!

[placid coyote]

dont free users have few (3 or 5?) pro searches daily? though not sure if they have access to premium models

[oblique quest]

Make multiple google accounts

[slim delta]

Thank you! I think it's a limit problem, but I'd like to try the voice system

[fluid trench]

has anyone actually managed to use the sonar-deep-research api?

[jagged root]

The voice is very good it’s getting better. What I like about it you can work and not read the Text base as workflow continues. And I have The Pro as my main go to for real time Info. Just 1 tool/ Workflow Assistant. Hope that helps. Highly recommend Subscribing 👍🏻 have a great Start off the Week everyone. 🌞

@slim delta

[slim delta]

Thank you @jagged root I'm going to give it a try

[rare echo]

Hi all , any plans to increase the image limit in a thread ? 4 in a thread is very low!

[full fern]

Why does my Deep Research doesnt have AI model option to choose from, unlike R1 and pro search?

Hi All

[raven night]

Thank you👊 , jm a big 4o guy, why u rocking r1?

[silver bay]

because it's based on google's 'deep research'

it's already a model on its own

https://gemini.google/overview/deep-research/

reminder that you can always ask questions about Perplexity AI TO Perplexity AI

[full fern]

Hi Simplicity IA, Sorry, but can you simplify on what you mean based on google? Thank you.

[silver bay]

Do you know how an AI Chatbot works?

it's based on a model, right?

[full fern]

yes.

[silver bay]

So, Perplexity AI makes some options available, Reasoning gather models that are tailored for that purpose, Pro gather some other models that are tailored for a wider range of tasks beyond reasoning

and Deep Research is based on Gemini's Deep Research

it's not a category

[full fern]

I see thanks a lot Simplicity IA. reaally appeciate it. Thank you.

[long pasture]

I doubt this is the case

Its just name similiarity

[long pasture]

Googles deep research is good at finding sources but uses the terrible 1.5 pro model

[silver bay]

its the same name tho, I got baited

[long pasture]

Every deep research is same name

There are like 5 products

[silver bay]

they are indeed similar

[rare echo]

anything?

[storm wraith]

Hi all, I was wondering if anyone has signed up using a linkedin 1-year free and could tell me if the APi is free (Was trying to get a token but its asking for a credit card) trying to figure out if there is a fee for teh API after X uses or something

[lofty wing]

do you guys know if R1 used by Perplexity is the 671B total parameters with 37B activated for each token one (The one which official deepseek uses)? Or is it a distilled version

[rare echo]

apart from grok2 , which one would be the best for academic purposes

[left stream]

4.5 is good with explanations same with claude 3.7 extended or normal one

[silver bay]

I'd go with deep research for academic purposes

[silver bay]

have you tried asking perplexity AI that

[waxen lion]

Hey guys this a non ai related questions but can someone help me with my eq settings for the senheisser momentum 4

[rapid bane]

guy's

why on phone we cannot select the ia model

[rapid bane]

pretty sure it's another version

[silver bay]

have you tried asking Perplexity AI that

[strong pond]

Read the announcment lol

[left stream]

I wonder how perplexity makes money.

4.5 is 20 a week at you.com while 10 a day with perplexity

[rapid bane]

I'm lazy to read

[bleak holly]

how do I set claude as default?

cause it's getting shifted to auto and the ai is losing context after each message for a particularly long command I gave

[indigo valve]

For phone or for PC?

[rapid bane]

bro me too

ai is losing context

for school it's not that good

[bleak holly]

pc, in web

[rapid bane]

guy's should I go back to chatgpt or I stay on perplexity

[indigo valve]

There isn't a default anymore. You can select the model straight from the search

[strong pond]

Do you have pro?

[bleak holly]

yep, I do

[indigo valve]

You can also use Complexity to even easier change models.

[rapid bane]

how to fix ai losing context

[strong pond]

They have differnt use cases. I don't really use ChatGPT muc hthese days at all. I use almost exclusivly perplexity for research and claude for creative writing.

[bleak holly]

is default available on android atleast?

[distant aurora]

4o is dumb. Responses are short too. But whatever works for you. O3 mini doesn't always work as intended on perplexity.

[distant aurora]

Claude is hella versatile since the thinking model update. Pricey af though.

@left stream perplexity still isn't letting me choose between o3 mini and r1 on the free version.

On browser.

[left stream]

Ikr

[indigo valve]

Yes

[indigo valve]

Yes for now

[left stream]

That’s weird, I thought it would be fixed by now

[distant aurora]

Yeah. Lemme send recording.

[slim lion]

It depends on the topic you are looking at. If it is more neutral in nature, GTP-4.5 works well. But if it has to do with more controversial topics or ones that would have biases, I have found great luck with Claude in doing my searches. It may, from time to time, preach to you, but if you tell it to adjust, it will. GPT on the other hand, doesn't like to get out of my way. Also Deep Search is great if your wanting to look deep at stuff.

[distant aurora]

Grok is fairly neutral imo.

Talking to grok is almost like talking to an actual person.

[slim lion]

100% It is kind of weird at times.

[edgy meteor]

does perplexity give same resposnse as lets suppose sonnet 3.7 when we select that model or perplexoty optimizes or alters the response in any way

[distant aurora]

I like grok.

[slim lion]

I just wish it had a little bit more power and that Grok 3 was support in Perplexity.

[distant aurora]

Grok has a weird charm. Even grok 2. It doesn't outright deny your statement. It will show some acknowledgement, try to understand your perspective, and then present it's arguments/solutions.

Like chatgpt or deepseek will straightaway deny or give typical bot like responses to problems. It doesn't feel natural.

[slim lion]

From my playing with Claude directly and in Perplexity. It will go into more depth in their own portal but will go off the rails more. Rather than in Perplexity, it stays on track a lot better, but it won't go into the same depth unless you have it in Reasoning or Pro.

[distant aurora]

There's some altering.

Accessing models via perplexity isn't as consistent as their official websites.

O3 mini is not upto the mark sometimes.

[edgy meteor]

In which contest were u using? My usecase is STEM and Programming. And I have pro

[slim lion]

Question though, both Grok, GPT, and Claude all have three modes for the most part now. General, reasoning, and deep search. In Perplexity you can access most of these baring Groks. How should one approach these three different modes for different types of jobs? What are the dos and don'ts you have noticed with these modes?

[left stream]

Have you tried it on a computer?

[distant aurora]

That's desktop mode on chrome.

Result should be same.

[edgy meteor]

ic the o3 mini with perplexity worse than o3 min (high) or just the free version o3mini (low)

[slim lion]

I am not doing programming, but I do a lot of research, marketing, and content work. I also have Pro Perplexity, but not Claude.

[distant aurora]

Worse than high.

[edgy meteor]

thanks guys

[distant aurora]

You have pro, right?

[edgy meteor]

yeah

[distant aurora]

Give this prompt after selecting o3 mini

How many r are there in terrrrific

You'll get what I mean.

I'm serious.

[edgy meteor]

bruh it's worse than o3 mini low

[bleak holly]

grok on x is completely different from grok on perplexity, perplexity version won't even say the R word but X version can curse your whole bloodline if asked to

[distant aurora]

And then try the same in free version of o3 mini on chatgpt.

It won't miss.

[edgy meteor]

perp o3 mini : 6
o3 mini low : 4
over for perplexity

[slim lion]

They are night and day different. I hate that I have to keep building prompts for both Perplexity and Grok. I wish I could just do it all in Grok, but such is life.

[fickle tartan]

Can anyone suggest searches to do on perplexity and you.com to help me decide which to use? Something that shows their differences.

[distant aurora]

Do you have pro subscription?

[left stream]

It said this for me

[distant aurora]

https://www.perplexity.ai/search/how-many-r-are-there-in-terrrr-5.nnZmKuT9aBaCyeoMAAXQ#0

[edgy meteor]

clownglasses

[left stream]

Damn 😭

[distant aurora]

Joe, istg, I think perplexity is favouring iphone users.

[tacit skiff]

hi mates , I would like to ask you a question to see if you can help me. I cannot create images when I have had PerplexityPro 1 year since the 6th with an invoice, but when I click on it, it appears as if I do not have it and I have to pay

[slim lion]

Drive me mad! LoL

[distant aurora]

Omw to copy paste this into perplexity for response, lol.

[distant aurora]

https://www.reddit.com/r/perplexity_ai/comments/1bh3jra/i_cannot_generate_an_image_at_all_with_pro/?utm_source=perplexity

[fickle tartan]

No but willing to buy a month if needed. Or I could perhaps convince someone to DM me results 😊

[left stream]

I’m gonna test it out more now I’m curious. I have an android too so I’m gonna compare. Those questions that you had solved on android was weird and wanna try copying that

[tacit skiff]

thanks mates just now i saw my subscription was canceled

[edgy meteor]

it gets it right on claude 3.7 everytime tho

[edgy meteor]

how to do that lmao

[indigo valve]

But y

[long pasture]

Who knows gpt-4.5 usage reset timing?
Is it UTC 0?

[left stream]

[distant aurora]

Because this world needs ocassional violence.

[placid coyote]

its rolling window. you used 1 credit at 2am, 24h from that point in time you get the credit refreshed

[distant aurora]

Perplexity deepsearch might be better than you dot com. Depends on use case.

I would still say grok deepsearch is best for its price.

Considering perplexity isn't as good and open ai is too expensive.

[long pasture]

So the worst kind damn

Ty

[distant aurora]

It's pretty unusable imo.

[placid coyote]

isnt cheapest option on grok com for 30$/month (not counting buying it for a year, in AI things change quickly, I dont think it's wise)? on x I think it was even more, like 50$? that doesnt feel too much value. way worse model selection, no "spaces", no file rag (it failed on text 300kB, qwen and chatgpt - both free, managed to process it), worse image gen model. yeah, search seems better, but I would have a hard time justifying paying more for much less
did they clarify how many uses of deep search daily?

[distant aurora]

Yeah

[long pasture]

You had 50 for 2 hours

But it might be even better rn

Someone using heavily can spend more than $30

[placid coyote]

I am not sure if it still not because of that beta period

but 50/2h sounds really good. slower, but reponses felt less hallucinated and more indepth

[distant aurora]

50 dollars is x plus subscription

Super grok without x plus is 30

[long pasture]

It is apparently 100 per 2 hours

100 thinking 100 deep research

[distant aurora]

100 deep search is a lot.

[long pasture]

The thing is you won't do 100
Like unless you want it for archival purposes
Even then you won't know how good the response is without reading

[sturdy skiff]

this is for which AI?

[long pasture]

Grok

[sturdy skiff]

no man, i am following Grok discord, it is

• 100 messages of Grok 3 per 2 hours
• 30 think messages per 2 hours
• 30 deep search messages per 2 hours

its not 100 for think and deep search

but still 30 per 2 hours is great for 30 USD

but i still feel perplexity deep research is very powerful, after chatgpts deep research. and perplexity deep research is 20 USD, which is lesser than chatgpts 200 USD (+Tax) and Groks 30 USD

[zenith bramble]

you. com with o3 mini (high) answerd it correctly but with perp said 6

[long pasture]

Oh there is a limit for base model?
Thats bad bad tho
I assumed only rates for where thinking and search

[blazing comet]

Does anyone know how to publish perplexity pages to the discover page?

[sturdy skiff]

but still base model is 50 per hour, which sounds like unlimited.

[leaden lagoon]

Where is comet ☄️

[indigo valve]

Look up

[sturdy skiff]

What is the context window size of perplexity? And can it be used as replacement of chatgpt, grok, Gemini and claude?

Because i can see custom instructions

I can see file uploads

I can see web search

I can see models like claude3.7 and o3 mini

So have this doubt

[naive sage]

Though the documentation references old models, it is probably still pretty similar to this: https://www.perplexity.ai/hub/technical-faq/what-is-a-token-and-how-many-tokens-can-perplexity-read-at-once

[fathom moss]

Just wanted to show gratitude for Perplexity and it's existence.

[austere granite]

I downloaded a gta car and it keeps crashing and I have no idea how to figure out why it's crashing to fix it

[distant aurora]

Yeah. How many o3 mini high prompts does you give per day?

[distant aurora]

Quantity vs quality

Though average user might not notice the subtle differences.

[sturdy skiff]

i agree chatgpt deep research is the best, as i used that one as well. but due to unable to pay again 236 USD for chatgpt pro i opted out, but i feel after chatgpt, perplexity is 2nd ranker in doing good deep research, because i even tried gemini deep research and grok deep research

[distant aurora]

I feel grok is second.

I haven't tried gemini deep search.

[sturdy skiff]

dont try you will regret, after using chatgpt, grok, perplexity, gemini feels like a toy

[distant aurora]

Is it free?

Istg, I would have never known how good o3 mini high is without trying.

People say it's similar to r1 but they're wrong.

[sturdy skiff]

they are 100% wrong, gemini is not competitive to any famous AI company models

[distant aurora]

O3 mini high

Read again. 💀

[sturdy skiff]

according to my usage, o1 helped me more than o3 mini high, o3 mini high is over thinker for my use cases.

[distant aurora]

I felt like o1 sometimes loses track of the original prompt.

I was trying to make a complex excel macro.

Now ofcourse I can just remind it of the original prompt instead of complaining about it, lol.

[sturdy skiff]

see initially i am great fanboy of openai models, from last 1 month i am observing almost all models of openAI are unable to hold on to custom intructions or original prompt.

i found claude 3.7 and 3.7 thinking are able to hold on to custom instructions till the end of chat

[distant aurora]

I don't like Claude. Like, it's great but the limits and pricing just piss me off.

Most expensive ai I think.

[sturdy skiff]

agreed, even i am not permanent user of claude, i have same issues, but i am using it for initial planning and continuing with perplexity. chatgpt these days not doing what i ask , so i am not using it though i want to use it

[distant aurora]

I don't understand why it has to be like 4x pricey.

And the default Claude app/website doesn't even have web search yet I think.

[sturdy skiff]

yeah, because of that reason i dont use claude for long chats, actually claude is very good performing AI, but due to limitations of usage many does not use. Claude is good in sticking to the goal, but after long usage it starts saying dont do long chats 😄 , else we loose accuracy, also message cap limit also crosses many times 😦

[sturdy skiff]

no web search, no deep research

[distant aurora]

They won't release deepsearch imo.

Target users are coders.

[sturdy skiff]

yes, and many praised Grok3 is perfect for coders, so again claude has competitor

i use perplexity mostly for research only on uploaded files, i love the option to choose the way i want to research specifically only on uploaded documents

[distant aurora]

Are you using via perplexity?

[sturdy skiff]

No, i used to be chatgpt pro subscriber, and opted out of it.

[distant aurora]

Oh

[short narwhal]

Aaaa I need cosmos

Or was it comet

I forgot

[distant aurora]

I need money

Lots of money.

[short narwhal]

Don’t we all

[distant aurora]

I guess.

I wish I could buy a new phone.

Or get a good job.

Life sucks.

The amount of envy I feel when I see people worse than me getting more in life.

[noble gull]

Does perplexity api allows for images upload?

[tame current]

whens the comet browser coming out

[unique umbra]

Github copilot goes hard

[unique umbra]

With pro

Which let me use claude 3.7 thinking for an entire day bf ratelimiting

[tame current]

i didnt even know it existed until a week ago, had to stick to cursor

[unique umbra]

Cursor is good but expensive

I would use it if it let me add my own api

[tame current]

i was just using the free trial

[abstract ibex]

Compared to Cline it's cheap af

[unique umbra]

I get copilot for free

Bc student stuff

[abstract ibex]

Your case is special then, for others it's paid

[tame current]

college email?

[unique umbra]

Yeah

Github student dev pack

[tame current]

wait so i probably get it too

damn i need to look into that

[unique umbra]

Copilot isn't that expensive tho

Literally 10 bucks

[unique umbra]

Broken English

[sleek topaz]

I dont see why ppl would willingly use aider, cline or claude code over cursor or windsurf lmao. Literal money sink when you could just use cursor for $20 or windsurf for $10-15 and save potentially hundreds/thousands of dollars in otherwise personal API token costs

[long pasture]

Cursor is gonna be smarter about their rate limits eventually
Just because they are juggling costs between low usage users and high usage users doesn't mean its sustainable

[sleek topaz]

given their $10 billion valuation and how they probably have a deal with almost every AI provider connected to their IDE, its likely not costing them as much as you'd think.

and tokens are not as expensive for the actual AI company as they're saying

[long pasture]

If you are saving hundreds/thousands of dollars by just paying $20 to them who is gonna pay the bill?

By your logic

They are inventing money out of thin air

[sleek topaz]

because we arent a massive enterprise with partnerships with the LLM providers we host, we're just normal people who have to hand over the big bucks for their significantly upcharged API costs

[long pasture]

The reality is these startups are trying to acquire as many users as they can to beat their competitors to death with VC money and its gonna run out at some point

Hence something like OpenAI making $200 tier

[long pasture]

You know there is a base cost and even the enterprise users don't get much of a discount right

Defiinitely not spend 20 bucks get 200 bucks of credits deal

[unique umbra]

Wrong

There's multiple small organizations called ai apis that offer very cheap ai with multiple models available

Although they do scrape providers and keys to keep it running

(Zukijourney)

Branched from gpt4free

[long pasture]

This only works because providers are still really generous yes
You can sometimes spend a months of worth tokens in a day with subscriptions

[unique umbra]

I can keep goin with keys btw

I'm talkin thousands of keys

[long pasture]

Is that OAI key

[unique umbra]

Yes

[long pasture]

Stolen acc or

[unique umbra]

Scraped key

[abstract ibex]

They already make profit from subscriptions

[unique umbra]

Get em while they last 💀

[long pasture]

How do you know are you a Cursor investor

As the guy above said

[unique umbra]

It's called deals with whatver ai company for legit companies

To make profit

[long pasture]

You can spend hundreds of dollars of token on Cursor with just a $20 membership

Easily too

[unique umbra]

Still would only use it if I can add my own api to it

I got more models than them

And it's faster

[long pasture]

OpenAI hosts their own models they aren't even in profit yet
Perplexity in profit yet? We don't know but I highly doubt with these rate limits
Antrophic? No they are still getting cash injections to stay afloat (but they also train new models so thats a given)

[unique umbra]

Investors invest and they pay them back, it's a never ending cycle + they get deals with their ai provider + they count on not everyone spending a lot of tokens

Ai prices have dropped considerably

[unique umbra]

For even normal people

[long pasture]

Prices dropped because there is crazy competition

The gap between models are so small these days

[left stream]

Imagine they release 2.5 opus

[long pasture]

China caught up

[unique umbra]

And then fell back down 😭

[long pasture]

They probably trained a opus but then they realized its gonna be like gpt 4.5 expensive with little benefit and decided to not release

[left stream]

R2 is gonna release before may or around there. Hopefully it’s good

[unique umbra]

Deepseek was very much needed to the industry but I don't want Chinese models taking up the space

[left stream]

True, most likely.

Wonder how long sonnet is gonna be around for

[long pasture]

If it wasnt for deepseek OpenAI wouldn't release o3 this quickly or that price I bet

You should be thankful even if you dont like them

[unique umbra]

As I said, it was needed but I still don't want Chinese driven models

China has way too much control over the industries over there

[long pasture]

The rumors for sonnet 4 is its gonna be april may soon but not sure

[long pasture]

Deepseek is less censored for me Idc if the model cant talk about tiananmen square

[unique umbra]

It's ab data privacy...

Deepseek is a means to open up more Chinese industries to the ai market

[long pasture]

Yeah good point but I think Chinese having my data is better than US having my data
I won't ever deal with Chinese government in my life I dont think so

You can still use intermediatiaries anonymize your prompts etc for really critical things

[unique umbra]

The chinese use ur data to target America (tiktok is dif as its literally brain dead posts) vs corporations in America and everywhere else in the world just collecting and selling ur data

The dif is

China has total control over ur private information and Chinese corporations cannot hide it or keep it away from them, everywhere else corporations can choose which data to sell

[long pasture]

How do I know NSA doesn't have a datalink directly to Microsoft servers which host OpenAI gpus

[unique umbra]

Let's ask perplexity if that would violate laws

💀

[long pasture]

I think the gov is legally allowed to buy data from brokers

Not sure

Forcing them would be illegal obv

[left stream]

Looking forward to that 🤘

[unique umbra]

I wasn't denying that, but corporations can choose which data to sell

Which is outlined in their tos

Or should be

As where in China they can't choose

If Chinese government wants all the data they can get it

[long pasture]

The snowden files show that when US wants it US gets it too

US has better PR

[abstract ibex]

https://sacra.com/c/cursor/?utm_source=perplexity

[long pasture]

But they got all the data not case specific data so it doesnt matter

[unique umbra]

At least legally they need court approval

[long pasture]

No profit announcement? Only revenue

[unique umbra]

So they're making 100m and getting 105 mil in investments..

Is it a coincidence that they're actual profit is 5 mil

[long pasture]

Investments arent profit they are a stake in the company

Its not even raw revenue tho its ARR

[distant aurora]

Lmfao. You think Americans don't steal your data?

[distant aurora]

Sh1t is not allowed.

[unique umbra]

I don't feel like rewriting it bc of one word

[unique umbra]

Would u fr rather let China have ur data..

[distant aurora]

Yes.

[unique umbra]

Ok bro

[distant aurora]

Cia is worse

Like it or not bro.

Look up the horrible experiments they did on American citizens.

[unique umbra]

Back in the 1900s 😭

Fuckin lobotomies

[distant aurora]

Mk ultra

[unique umbra]

Ik

Times changed

[unique umbra]

U think China is any better 😭

[distant aurora]

I think it doesn't matter unless we are someone like Andrew tate.

[unique umbra]

I think it does matter

China can use it against our country

[distant aurora]

How?

You think your password or chat is an asset for your country?

[unique umbra]

Targeted propaganda and cyber intelligence

[unique umbra]

Such as rednote..

[distant aurora]

Propaganda. Huh. That might be a valid argument but america has plenty of propaganda of its own.

[unique umbra]

It's giving China a soft power over our country wether u see it or not

[distant aurora]

Idk. Do you trust your government?

[unique umbra]

I trust my government won't end our lives when ww3 comes

[unique umbra]

Ik China doesn't care

[distant aurora]

China won't bother to interfere ngl.

They'll just see a business opportunity.

[unique umbra]

Bro, pearl harbor 😭

[distant aurora]

Similar to BlackRock seeing a business opportunity in Ukraine.

[distant aurora]

That was Japan I think?

💀

[unique umbra]

Crazy

[distant aurora]

And here you are talking about propaganda.

[unique umbra]

That isn't propaganda

[distant aurora]

I'm saying you don't know history.

[unique umbra]

That's just misappropriating races/ countries

[unique umbra]

Says u?

[distant aurora]

Yeah. Did I say anything wrong?

[unique umbra]

Ik my history idk wtf ur talkin ab

[distant aurora]

It's okay. Believe what you believe. Or who you believe.

[unique umbra]

Wut

[distant aurora]

We are too insignificant to argue about it anyways.

We are not in power.

[unique umbra]

Not how democracy rll works but k

[unique umbra]

50% is still 50%

[distant aurora]

All fun and games until you end up on the other side.

[unique umbra]

Leftists 📉

[distant aurora]

We live in the age of controlled oppositions and puppet governments.

[unique umbra]

The Biden administration lol

[distant aurora]

Anyways, I'm kinda curious about what Elon musk is doing. We will see.

[distant aurora]

Obama third term.

[unique umbra]

Real

[distant aurora]

Bro couldn't say or think sh1t but people voted for him.

Meanwhile Trump gets impeached.

Horrors of democracy.

[unique umbra]

At least trump broke the conservatives streak

Actually gettin some changes

[tame current]

New race for acquiring valuable assets across the world has begun.

[unique umbra]

Crypto and stonks

[crystal cosmos]

What is this?

[naive sage]

He means governments shoring up their supply chains, access to raw materials needed for things. It can be a sign of several things, but none of them very good.

[neat relic]

@unique umbra can u jb o3 mini?

[silver bay]

I haven't tried manus

so I have no idea

perplexity is not really comparable to chatgpt since it's an aggregator

but for coding and research overall I'd go with perplexity

image generation and creative writing maybe chatgpt

[sturdy skiff]

So Manus is powerful in research than perplexity? Just curious to know

[silver bay]

I just said that I don't know

haha

[tame current]

Is there a limit placed on deep research and GPT 4.5?
It says GPT 4.5 only has five uses left which is way lower than the 50 uses that Claude 3 opus gave after downgrading it from unlimited usage, before they ended up getting rid of it.

Also, after a couple of times using deep research, it switches over to pro search instead. Does that mean deep research is also limited?

for $200 a year, that’s pretty messed up ngl

[subtle knoll]

Deep Research is limited to 120 a month with a Pro account, 10 with Plus. Got 4.5 100 answers with Plus and I don't know with Pro.

OpenAI's deep Research uses the upcoming o3 full model

[tame current]

I just got a new membership on another email, and it switched from deep research to pro search only after a couple of uses

[subtle knoll]

I mean with chat gpt

[tame current]

oh

whats perplexity plus?

i thought there was only pro and free

[subtle knoll]

I was talking about limits on chat gpt sccount

We don't know with p.ai

They don't want to tell us

[whole cape]

No one even knows if manus is real. No one has access to it, no independent testing or benchmarking. So manus isn’t even in the discussion at this point 🤣

[proper rapids]

is there a settings to not let perplexity use up my remaining enhanced queries unless explicitly selected? I only see Auto and often times, it will use up my Pro search free quota

[zenith bramble]

There is no limit mentioned on o3 mini ( high) , 4.5 has some limit similar to perplexity

[neat relic]

Wdym by this?

[livid aurora]

What is this

[neat relic]

U have 100 upload limit

[livid aurora]

Per month?

[neat relic]

Per day

[livid aurora]

Weird

I did not used it so often

[distant aurora]

So duck duck got into ai business?

[polar heron]

how to use Perplexity as my discord bot?

[sharp schooner]

Hey brothers, do you have reading mode on web? It appeard for me and worked for one day and it was great but now its gone and I really liked this option to make it read the responses

[unique umbra]

Wut?

[dire vessel]

hi

[quick steppe]

Hi

[dire vessel]

i am new pro user

[quick steppe]

That makes 2 of us.

[dire vessel]

is perplexity really worth it?

[quick steppe]

We'll find out.
I frequently use ChatGPT, but I want to compare the deep research of Perplexity with theirs.

[hearty blaze]

Hi

[dire vessel]

i also have pro version of gpt

[quick steppe]

What do you use it for?

[dire vessel]

no especial reasons

[quick steppe]

How long have you had it for?

[burnt quiver]

Hello ankit 😂

[burnt quiver]

Wow

Ur opinions on gpt vs perplexity guys?

[worthy flame]

Perplexity wins, hands down.

[crystal cosmos]

Both GPT and Perplexity offer unique strengths: GPT excels in creative, conversational depth and technical problem-solving, while Perplexity shines in research-driven tasks with its precise, citation-backed answers, reflecting China’s remarkable advancements in AI innovation and infrastructure that empower such tools to thrive globally.

[burnt quiver]

Ai result

[burnt quiver]

🙌

[keen root]

hey, have revolut and was waiting for my paid sub to end, as I saw suggested here. and now that my perplexity sub ended, I enter the code and it says "You have previously subscribed to Perplexity Pro and cannot redeem this promotion."

so did they change that in the mean time and I can just forget about getting free perplexity for a year huh?

[serene snow]

Contact Revolut I had the same issue with Headspace and they generated a new code for me, so I am assuming it is the same case but with Perplexity.

[left stream]

O1

[crystal cosmos]

u paying?

[left stream]

Yes

I was testing different prompts to see how the response size gets effected

[mortal bolt]

@signal hamlet Why the writing mode has been slashed off and Perplexity goes into pro mode when I don't want to use the pro mode?

[left stream]

I think I found the prompt I like and will use for deep research in perplexity

The longer the better

[crystal cosmos]

to conserve operational costs

[royal wasp]

Hi Team,

I've purchased the pro subscription using a promo code given by my family member. The promo code stated that the subscription will last till one year. But the subscriptions page in the web is saying that my subscription will end on april 5th. I've canceled the subscription because I used my friend's debit card to make sure that he should not be charged in future. Please help me in this issue.

[mortal bolt]

Wow! Interesting. I would assume operation cost is higher for pro mode

[royal wasp]

Please anyone help me

[mortal bolt]

What is the issue? Not getting the one year sub?

[keen root]

thanks, will try

[royal wasp]

Yes, in the subscriptions page on web, it states that the subscription will get cancelled on April 5th 2025. While my promo code is valid till 1 year

[boreal ginkgo]

Will perplexity integrate Manus AI into it? 🥰

[livid aurora]

there is a typo in the word sonnet

[royal wasp]

@mortal bolt please help

[royal wasp]

Should I renew the subscription ? If so, will it cost me ?

[shell magnet]

I’m thoroughly enjoying this app. It has helped me with so much.

[shut dagger]

guys, I just found out claude 3.7 has hard output limit of 5000 tokens

[keen root]

thanks. I contacted Revolut and they said I need to register new account with Perplexity. Its a bummer, but hey, still worth it I guess 😄

[distant aurora]

There are bigger issues than that which you might realise after you use it. 💀

[distant aurora]

On perplexity or everywhere?

[shut dagger]

Perplexity

On Anthropic, officially it says it can output upto 128K

[distant aurora]

Nice

[willow kettle]

What model is used for the Pages feature ?

[shut dagger]

Default GPT 3.5 or Sonar

probably sonar

[mild obsidian]

Hey how i can get claude 3.7. Api

From perplexity?

I have pro subscription

[livid aurora]

You cant

[livid aurora]

haha

[mild obsidian]

How i can get pro role?

[livid aurora]

Not sure

[mild obsidian]

How u got it?

I Have Buyed Perplexity pro enterprises plan yearly cost 2k$

[neat solstice]

When?

[mild obsidian]

Today

[peak scarab]

He didn't buy it lol, he got it from the hubspot promotion

[neat solstice]

Ah I see

[mild obsidian]

How to get pro role?

[peak scarab]

Here you go

[mild obsidian]

I have joined the server

Now ?

[peak scarab]

Hmm

[mild obsidian]

Who will give me the role? XD

[south kindle]

Leave the server and join again via the link in your settings. It should give you the role. Or wait for a mod here.

[mild obsidian]

Got the role

Thxs

[mild obsidian]

What's the reward in unlocked role?

[sturdy skiff]

Does gemini deep research does research on uploaded documents alone? Does it have that feature?

[placid coyote]

nooo, what happened to my poor deep research D:

I don't want to be too negative, but if you are talking about integrating with pplx api, I think that is very new with many growing pains. pertty sure citations were added only recently and only models they support were their finetunes (llama and R1 based)

[distant aurora]

Lmfao. Perplexity fail.

[placid coyote]

this on enterprise faq sounds maybe like what you wanted it use for?

Can I use Enterprise Pro to ask questions about my internal data?

Yes, connect your internal knowledge bases to Perplexity and get answers that combine the best insights from both the public web and your internal data. Enterprise Pro offers two ways to search across internal data: You can create a shared organizational file repository that any Perplexity user in your organization can select as a search source. You can also create customized workspaces that give your teams the flexibility to tailor Perplexity to their specific use cases by connecting relevant internal knowledge sets.
https://www.perplexity.ai/enterprise

but I dont think on discord staff is too active. better chance finding some email, on x (see them quite active, even the ceo) or maybe linkedin

[distant aurora]

X

Twitter was down for me few hours ago.

[placid coyote]

didnt you hit the api quota? when musk took over and closed it more, especially free accounts are often hit with rate limits

oh, maybe not rate limits. just saw this

[naive sage]

https://www.ndtv.com/world-news/twitter-cyberattack-elon-musk-what-is-dark-storm-pro-palestine-group-allegedly-behind-x-cyberattack-7897600

[distant aurora]

Bro

Palestine is lucky Elon isn't working on weapons, lol.

[nova plover]

is perplexity able to execute python script generated by the AI?

[meager smelt]

I think you can force it to do so by specifically writing "use the python sandbox" or something similar

[placid coyote]

yep, it can. pretty sure only in Pro search and the system must think it is needed. usually mentioning programming, exact result or python sandbox does the trick

[indigo valve]

Does anyone have an issue where deleted threads keep coming back?

[placid coyote]

well, I had only breeding issues with spaces. got few dozens and when it tries to load next page, it duplicated tiles of those already loaded spaces. maybe it is finally fixed? I wish they added filter/search for spaces...

hmmm, what perplexity burned in this pro search? web search stopped working or something? why is sonnet 3.7 telling me, it itself is not released yet 😆

[distant aurora]

Lemme try something funny

Give me the og prompt.

@placid coyote

[placid coyote]

it was a thread, my queries (with web search enabled):

• claude api context size
• what about 3.5 and 3.7?

it looks like it tried to do search, but the "Reading" step failed?

I thought something was fishy, when newest model in first response was "Claude 3 Opus: 200,000 tokens" 🥹

[nova plover]

fr, i was working on some math problems and it suspiciously gave a pretty accurate answer, they need to highlight the code execution more fr

and does the image generation thing still works? the settings mention prefered image gen model but i never been able to trigger it

[placid coyote]

yep, it still works. not that trigger happy as it used to be. though more useful for charts than drawing images 😅

[nova plover]

i mean this thing, not with python

[placid coyote]

ooh, if you mean a normal image gen, then that must be triggered by a user

[nova plover]

oh, i see

[placid coyote]

I usually give vague words to sonnet for rewrite and then use what it generates (expanded image gen prompt) via the custom prompt (the wrench) after clicking on "Generate Image". I have a preprompt / template prompt prepared and can quickly use it via perplexity helper userscript (the tags)

for quick illustrative images, like for presentations, opensource docs/readmes/logos or just for fun to post on x/discord, it is quite capable

[winged axle]

whats the idea behind the follow up thing not being copyable jesus christ

[placid coyote]

it looks like I can select it and ctrl+c works, I can see it in the clipboard. that doesnt work for you?

a bit fiddly to not accidentally click it. also I have a bunch of browser extensions. or are you using the mobile app?

[distant aurora]

App is Merlin.

https://www.getmerlin.in/chat/share/ba3db527-7190-4c1b-be30-3de12e4abacd

[placid coyote]

pretty nice front-end and good response

[distant aurora]

Yeah. I couldn't find any details about the rate limits in their subscription unfortunately.

I considered poe once but it can't pair reasoning models with web search.

Perplexity is inconsistent. I don't feel like paying 20usd just for deepseek r1.

The deepsearch might be value for money though for the average user. Grok is still better if you want quality over quantity.

[placid coyote]

pplx has a lot of features. many half-baked, like image generation, or deep research. even spaces could use a lot of improvement. but pplx have a nice selection of models, high limits and fairly often are ones of the first to integrate newly released models.
my friend is actually considering grok (not sure if on x or grok com). yeah, if he uses it mainly for search and doesnt toy as much with different ais like I do, I guess I can see the appeal

[twilit ingot]

Personally, idk of another deep search feature like Perplexity that hits 100+ sources regularly (with decent prompting) that you can hammer in all day long with a Pro subscription and does the kind of intuitive reasoning only after all the research has been gathered (through Complexity, at least)

[sleek vortex]

not bad

[distant aurora]

What's this?

[placid coyote]

👍 yeah, I was quite surprised how pplx failed at that... probably some temporary web search downtime?

isnt that the uncovr?

[distant aurora]

Lambda chat?

[sleek vortex]

yeah👍

[sleek vortex]

yeah or some internal bug i’d imagine

[distant aurora]

Hey fishie

What app is that?

[sleek vortex]

uncovr

wrong link

[distant aurora]

What's the pricing and rate limits?

[placid coyote]

with pplx it often feels like you have to actively try to do a lot of work which should have been done via the system. that deep research on pplx - for some prompts, like find these 20 shows, give me imdb rating and urls, brief, theme, genre, some my custom rating, and format it as this {template} - it nailes it perfectly. then I add a seemingly simple (it handles imdb right?) "and csfd rating + url" (a czech site for rating movies and shows) and it hallucinates virtually all urls and ratings from csfd 🤷‍♂️

[sleek vortex]

$10 or $20; models work on a credit like system but there’s also unlimited basic search / free model like pplx

not as like unlimited as pplx tho (600/day 😅)

[sleek vortex]

it does use r1 which has a high hallucination rate generally

not sure if their anti censorship finetune makes it any better either

[distant aurora]

Not on the 671b model.

[placid coyote]

I remember checking uncover after some time and I really liked the option to toggle off news widgets etc on home page. I dont think pplx has it

[distant aurora]

Most of it is on the 70b and lower models.

[placid coyote]

feline posted how pplx butcherd system prompt of deep research, forced R1 to hallucinate even more than its usuall self...

[distant aurora]

Has anyone of you tried minimax?

chat.minimax.io

They offer r1 for free.

[sleek vortex]

i mean it is generally high plus PPLX has its own stuff applied to the model

[sleek vortex]

the prompt spams to make the report like 10k words

multiple times

[distant aurora]

Yeah. Perplexity somehow degrades the response.

[placid coyote]

pplx guys tried running some benches and the degradation was negligible (like a fraction of a percent iirc). though only a bunch of benches and we dont know if it wasnt cherry picked (runs or even benches)

[distant aurora]

I tried this too. It kills the prompt if it finds it too heavy.

[sleek vortex]

somehow gemini is really good at not hallucinating

[sleek vortex]

yeah
nobody really has the inference to run the full model themselves and find out

[placid coyote]

especially with high context passed

[distant aurora]

4o mini being over o1 makes this look rigged.

I don't believe this.

[sleek vortex]

Not necessarily i mean the thinking models can hallucinate more

But the discrepancy in o1 and o1-mini is interesting

[placid coyote]

some benches can have a lot of weird biases, often unintentional. also 4o is like a roller coaster - some releases better at coding, some awful. not sure if 4o mini is in a same boat

[sleek vortex]

4o has had quite a few post training updates right?

[placid coyote]

I know I dont agree with many coding benches, like llm arena. I remember times when 4o and even gemini pro were supposadly better than sonnet 3.5, later 3.6. no way in hell, nothing intuits better than sonnet with my vague prompts combined with some rag, so seeing only fragments of real large code base

[sleek vortex]

like that update which makes it uses more emojis

whereas 4o-mini has generally been stable

[placid coyote]

one 4o was better at writing and droped at programming, then programming back and I think those emojis. I dont use openai models much, so pretty much lost track. but I remember trying out canvas with 4o on chatgpt and that was quite good. kinda wish perplexity implement their canvas. editable markdown thingy to work with ai on stuff like project plans, readmes, brainstorming. I am starting to toy with the idea to implement it, at least in simpler form

[willow kettle]

How are the pages featured on the home screen chosen?

[distant aurora]

True

[lone hound]

Does anyone know if Claude's Opus model is going to come out a new version?

[silent pier]

Has anyone heard of Sider AI

[placid coyote]

I think those were from pplx employees or picked by them. I somewhat remember they were recruiting page creators, I think for a free pplx pro

[distant aurora]

They're low on budget for now so no time soon.

[distant aurora]

Yes.

[silent pier]

is it nice?

the UI is very good

like extremely polished

but i wonder about the performance

[distant aurora]

Would have been lovely if it had web search. I would have bought it then.

[silent pier]

it does

it has deep research too

u didnt know?

[distant aurora]

Wait.

[silent pier]

YOOOO

look

[distant aurora]

I didn't see web search option.

[silent pier]

So is it good?

im thinking of buying it

[distant aurora]

Where is this option?

[silent pier]

here the icon next to the click and new chat with the lines and circles

[distant aurora]

Lmfao.

Ah. Web access is a paid feature.

[round plover]

this is interesting - i was suspecting that gemini is pretty good cause they seem to be the only platform separating their models into regular ones and "experimental" ones... i just got google one ai sub for a year (for free), and its pretty good 😐 and my perplexity sub just ended, so i might stick with it for a while, and the integrations are the best...

[lone hound]

Ok

[silent pier]

yea

[distant aurora]

Back then I used it to extract answers from pdfs. It was surprisingly good.

[silent pier]

they added more things like claude 3.7 sonnet

[distant aurora]

I had gotten that 5usd subscription.

[silent pier]

and other models

[distant aurora]

@silent pier just so you know, the 70b version of deepseek is dumb.

[silent pier]

ik

but it has the 600 one too

for paid tho

[distant aurora]

Okay. It might be good.

[silent pier]

yea

[distant aurora]

It was good at the time I used it.

[silent pier]

I am wondering if i should get tje unlimited plan or not.. Im currently chatgpt plus which doesnt fit me as a student

[distant aurora]

Cuz?

[silent pier]

I need other models since the competition is so high

[distant aurora]

What's your use case?

[silent pier]

Creative writing, translating, coding and writing essays

And making revision sheets

[distant aurora]

And what part does o3 mini not do well?

[silent pier]

creative writing

for me

sider has so much features

[distant aurora]

Could you elaborate? I rarely use ai for creative writing.

[silent pier]

Like sometimes it doesnt something completely different from what i asked, thing about sider is it has multiple models for me to compare, help me with my research directly from the webb and analyse pdfs

so it might be of high value for me

[distant aurora]

You might wanna just go for Claude but sure, try other options if you can afford.

[sleek vortex]

i mean...

[silent pier]

wdym?

[fallen kindle]

i thought windows app was a proper app, this is just like a chrome tab installed

[willow kettle]

Yeah

[fallen kindle]

dissapointed

[lament briar]

My friend sent me a referral mail for $10 discount I was making a payment but before entering the OTP I closed the window when tried second time it shows $20 and if manually try the coupon code it shows invalide. What to do??

[sleek vortex]

lol

[indigo valve]

You should combine perplexity with notebooklm

[errant pivot]

perplexity>>>>

than any other

loved it

[tame current]

hello

why is ollama 3.2 so slow 💔

[indigo valve]

Do you guys think it should be possible to search the news in the discovery tab?

Maybe keywords and see if someone published something similar to what you are looking for

[distant aurora]

https://www.instagramez.com/reel/DHEOQdYo2cj

[narrow silo]

@distant aurora

I sent you DM

[elfin gate]

Yes

[stuck delta]

i have a problem, i used code for 1 year pro subscription from moj telekom hr app partner and my subscription just disappeared

[elfin gate]

Lot of people experienced this.
I don't know if its all who got subscription through telekom or what

[stuck delta]

how can i get it back?

[elfin gate]

I have no idea. I guess if subscription is connected with your telekom account it shouldn't be problem.
😀

[narrow silo]

[e-mail]

[narrow silo]

we get a code from app. like for collecting points. and code is 1 year valid,, says in T&C

[elfin gate]

Magenta

[narrow silo]

yes

I also have a code , but cant use it,, alas, coz. I had subscription once 😕

and now I see there are problems ?!

😦

[elfin gate]

Dm

[vital bridge]

0

[crisp zealot]

oooo perplexity windows app released

wait

that's just electron of the normal site?

[elfin gate]

Shots fired

[narrow silo]

I had a talk with one guy who does projects, and basically, what he told me is that nowadays every app is just a wrapper of a normal site

[crisp zealot]

yeah but that's literally the most boring thing ever

And they even emailed me 😭

about it

[narrow silo]

I mean what do you expect it to be?

[dense ruin]

kinda true but one would have expected some desktop advantages

[crisp zealot]

yep

it's literally one-to-one as the site

atp it's not worth making a whole another app to eat your ram

it's just a nerfed web browser

[dense ruin]

and without browser extensions like Complexity

[crisp zealot]

• opening links from the app just takes you the browser

[elfin gate]

It's probably eating less ram tho

[crisp zealot]

that's a lot for just one site

discord is taking almost the same amount (electron too)

[subtle knoll]

It's just a window with an embedded chromium web viewer loading the website

I just use it as a chrome app, at least it keeps all the chrome features and extensions

[indigo valve]

Yup

[west reef]

ty for that guys ! #📢│announcements message

[south shoal]

Where’s the announcement for the perplexity business fellowship

[short narwhal]

wild to say that about an electron based app

[river bronze]

The term native has lost all meaning these days

[copper roost]

Can someone tell me how to access perplexity pro? I'm a pro subscriber, using a web browser on my pc. When I sign in, I don't land on a pro landing page -

[indigo valve]

There isn't a pro landing page

[mystic ruin]

if you have "Pro" then you are Pro 😉

[indigo valve]

Exactly

[mystic ruin]

and you can see your status here https://www.perplexity.ai/settings/account

[distant aurora]

Fr

[sturdy skiff]

what is special about windows app? i already use it in web

[mystic ruin]

nothing I guess?

[sturdy skiff]

yeah just dont want to install and waste my laptop disk space, if there is nothing special about windows app, as i am already comfortable using web app

[dry kestrel]

Anyone know if Cursor is worth it?

[mystic ruin]

Yes, you can test it 15 days (Pro Trial)

[dry kestrel]

I mean like compared to claude pro

[mystic ruin]

i only use the claude agent model with cursor.

[dry kestrel]

Oh interesting.

Do you know how slow it gets after the 500 fast requests?

Im just really concerned because 500 requests is not a lot for a monthly limit

[mystic ruin]

never had any issues related due performance. I write you private.

[dry kestrel]

ok thanks

[rapid gyro]

Does OpenAI Search API completely replace perplexity? Or does perplexity have more advanced capalitities that the search api is missing

[long pasture]

Replacing is the wrong word possibly
More like they are competitors
Openai search is really expensive btw open source google api + api models seems even a better choice

[restive arrow]

hi... need help and can't find where to find. in can't connect to my perplexity account

[short narwhal]

Huh?

[dire escarp]

Is anyone in Tier 3+ have any success with search_domain_filter? Like does the whitelisting/blacklisting actually work? or is it still buggy for you all?

[restive arrow]

??? do you have any idea?

[short narwhal]

I didn’t even understand what you were trying to say in the first place

What are you trying to connect your perplexity account to

Why are you trying to do that

[silver bay]

is the recently announced windows app just a web wrapper?

#📢│announcements message

[short narwhal]

Yeah

Electron

Crazy how no one wants to build actually native apps for windows

[silver bay]

yeah, that was expected

thats why I asked and didnt even bother downloading it

💀

[short narwhal]

😭

[dim meteor]

that's disappointing, it's basically a webapp then or like discord.

how much longer until desktop users get speech like you would get for mobile?

[silver bay]

that would actually be useful

cuz honestly whats the use of a web wrapper if I can just open my browser and access Perplexity AI

I even got shortcuts associated with it cuz Vivaldi rocks

[dim girder]

Hi everyone!

We charged $1000 in credits and have fully used them. Is there a way to check the total usage data? the web dashboard of perplexity gives only the current month usage, we don’t see the complete usage.

Can you let us know if there is a way to check the overall usage data? Thanks in advance! 🙏.

[ocean sleet]

Hey does anyone know about the Perplexity Business Fellow program? When does it start?

[wanton chasm]

Hey guys, anyone got issues with accessing perplexity with VPN, claudflare seems to be blocking me in

[wheat sky]

Perplexity is keep asking me to attach file when I have already attached

[silver bay]

update just dropped here, noice

[tough estuary]

[sturdy skiff]

I dont see that update yet

[silver bay]

well, it will prolly take some time to hit every single user

try cleaning your cache and updating the webpage if you're accessing it through a web browser

[tough estuary]

hey, who can explain to me why the GPT 4O model is always interrupted when asked to translate a passage?, not Claude or Gemini, perhaps the GPT provides fewer tokens(maybe or not).

[silver bay]

I had this same issue with 4.5 yesterday

lol, it wasn't even a highly complex prompt, just a simple question

[distant aurora]

[idle hatch]

hello， I used model r1-1776 with api, how to check NSFW ?

[azure abyss]

#📢│announcements working on linux

?

[whole cape]

It's available on the website so check again if it's still not available to you.

[sturdy skiff]

but i dont see options like this guy's image, i cant see differentiation of standard and reasoning

[sturdy skiff]

but i dont see options like this guy's image, i cant see differentiation of standard and reasoning

[sturdy skiff]

i can just see same like yours

[whole cape]

Click the selector and then choose reasoning and then you will see the same as my screenshot above.

[sturdy skiff]

yeah i can see same as your screenshot

but unable to see something like below which that guy mentioned in chat

he is able to see something like this

[whole cape]

Do you mean a column for standard and another for reasoning?

[sturdy skiff]

yes exactly i mean that

[whole cape]

From what I can see, that person is using the Windows app. If you aren't using that then there's your reason.

I'm on Linux so there's no native app for us, just the web app.

[sturdy skiff]

no dear, not in windows app as well, i have windows app,

[whole cape]

I'm not sure then I'm sorry. The person who posted that screenshot will be the one who can help you with that. But if it's just a UI problem, it's not really a big deal. You still have access to the same model, you are just seeing a different UI.

[idle hatch]

When I use the API to call the r1-1776 model, when testing adult content, the model directly rejects it. I think it should answer

[sturdy skiff]

ok 👍

[harsh verge]

He's using the Complexity extension; it has nothing to do with a Perplexity update.

[whole cape]

It's likely a limitation that Perplexity has placed on it.

[whole cape]

Oh there we go. I don't use this haha.

[sturdy skiff]

any idea how he used that extension? i will also try

[harsh verge]

https://www.cplx.app/

[sturdy skiff]

thank you for sharing

[idle hatch]

I agree, that document is misleading.

[ruby heart]

Why no more grok on perplexity

[whole cape]

It should be coming later this month.

[indigo kiln]

Anyone got Comet early access yet?

👀

[minor flint]

Hi All. Is there any free AI tool which alters the current CV and make it based on the job description so that it becomes ATS friendly

[sturdy skiff]

you can ask this question from Grok free account, from Grok deep search, the reason is it searches many sites in short time than human can. and you get better answers

[minor flint]

@sturdy skiff - hi thanks for the update. I will surely try

[ruby heart]

Why was it removed?

[whole cape]

Grok 3 hasn't ever been on Perplexity, only Grok 2

[ruby heart]

Yea grok 2 is gone

It’s only on mobile not on desktop

[whole cape]

It's on the web app

[ruby heart]

Oops I had reasoning on

Why can’t you combine grok with reasoning anymore?

[whole cape]

That's not a question I could answer unfortunately. Grok 3 will more than likely be compatible with reasoning given that it has reasoning capabilities.

[dim plover]

Message to the developers: the authentication via Apple ID for the Windows App gives a 500 internal error.

[ionic lynx]

I have a question for you guys: If I offered you a ChatGPT Pro account (non-shared) at $70/month—with exactly the same features as GPT Pro except for Sora, but including normal access to Deepsearch and O1 Pro—would you pay for it? If there’s enough interest, I might launch this service in the US or Europe.
I'm sure some of you will wonder why it's priced so low. To be totally transparent, we use credit cards from people located in Africa or from people who are about to leave the US, and have them pay for these accounts.

[ruby heart]

That’s cheap?

Ahh pro

Would it be new accnt every month?

[ionic lynx]

Of course not, because we will import the memories.

Its Pro，

[ionic lynx]

Compared to the original price of $200, $70 should be quite a bargain.

[robust mirage]

@ocean sleet is there a way to get registered to the program? Since the form is closed and I’m not sure whether any option to enroll.

[split comet]

hi, is not possible anymore to select model on auto mode? i don't need always the pro mode and i don't see any model selector when set on Auto

[fiery cave]

Yeah second that! Previous version was so much better, where auto-mode did not use Pro stuff. Its really really inconvenient. You can ask something super simple and it uses Pro mode for it....

It gets to the point where I would like it to run out of Pro mode searches so I can use it normally.

[past oxide]

Bruh wth man

[ionic lynx]

what do you mean？

[tough estuary]

Is there any way to translate a website directly with extension or anything like as google trans?

[past oxide]

So, escaping from debt, you meant?

[ionic lynx]

That doesn't matter. There are plenty of people doing exactly that, so I'm just asking if anyone would be interested in trying out Pro for $70/mo

[past oxide]

That’s still illegal heh

[elfin portal]

Guys when we get our ai agent can we not call it anus

https://github.com/nikmcfly/ANUS

I just think we could be better than this bros

[vivid otter]

Going through the readme file is cracking me up

anus —version

[elfin portal]

Yup! its a good open sourc ver of Manus

but like

come on man

[vivid otter]

I think this is perfect

[elfin portal]

😭

[vivid otter]

Naming convention peaked here

[elfin portal]

"Sorry boss my anus is acting up I need to restart it"

"what?"

its difficult to search anus ai and not get porn

[distant aurora]

Do a system restart

Lol.

[crimson token]

That's it, the next AI agent I make will be called Prostate

[elfin portal]

maybe if a government makes that one they can do Pro-State

DIC - Dynamic Interface Controller

"AIs following a Think, Wait, Act, Talk pipeline, or T.W.A.T for short"

[crimson token]

Thank god I am european, there is a non-zero chance it can be done

[distant aurora]

Explain this

[vivid otter]

TWAT is a powerful library that allows your language model to think, wait, act and talk

[elfin portal]

well when we think about how some AIs should work agentically they think, then they wait, then act and talk to the user about what they did

[distant aurora]

I see.

I thought the wait came first.

[elfin portal]

well I guess if I open perplexity its just there

waiting

for something

anything

[distant aurora]

More like chess

If you have ever seen it doing a loading screen while thinking.

[elfin portal]

not really a loading screen for perplexity but it you do see the thinking steps, could be updated to look like a loading bar or something but I like seeing each step

[thick dust]

Hey guys, 500 | internal server error when trying to authenticate Apple account login for Perplexity’s new Windows app; am I only one experience this?

[elfin portal]

apple login on a windows app

why dont you install edge on linux you animal

(joke)

[thick dust]

no linux sir, gamer here

[elfin portal]

to be fair, just yesterday the mic didnt work in the app

as in it only muted or unmuted

this app is hot off the presses

[thick dust]

Well at least for the web-version of Perplexity, my pro acc login works properly

[thick dust]

I just realized too the window app was released yesterday wow

[distant aurora]

Edge 💀

[elfin portal]

well with perplexity you might be able to not use a web browser

the issue is that you cant browse the web directly

as in if you ask it to open youtube it will say it cant

[distant aurora]

So what happens if you tap the sources?

Oh.

I would never use edge browser.

[elfin portal]

I'm a text-based AI, so I can't directly open YouTube for you. However, you can easily open YouTube by going to www.youtube.com in your web browser or by using the YouTube app on your device. If you need help finding something specific on YouTube, feel free to let me know! 😊

thats what I get

selecting hyperlinks within the app opens it in your default browser

as soon as perplexity can keep the user in the perplexity app and not in their default browser then they have won, search engines / browsers would be dead

[thick dust]

Edge bad? I usually use use sth else as default; but every time windows has big upgrade (which it did like half hour ago), it reset my default browser choice to edge

[distant aurora]

There's a dedicated Edge uninstaller somewhere on the internet.

Probably GitHub.

[elfin portal]

Edge is just another chromium based browser, unfortunately lots of windows services now rely on edgewebview 2 to launch e.g. New Teams, New Outlook

you can remove it from your system, I did like a year ago before it became a load bearing app

and I had to pop it back else office 365 apps for work did not launch rip

its a very meta thing to do when you think about it, just make Edge be a dependancy for the bulk of your app suite so it has to be there

[thick dust]

Man windows always full of these crapware

[elfin portal]

the only thing preventing the full move to linux is some of the games I want to play need anti cheat

e.g. black ops 6 and brawlhalla

its not unreasonable that kernal level anti cheat could be installed on linux I mean we put a man on the moon we can put anticheat on linux ffs

[thick dust]

Exactly mate! I can’t live w/o cheat engine & trainer when playing single player games

[unborn owl]

Did the winner of the perplexity champions trophy giveaway accept the prize or is there redraw possibility

[tame current]

Hello there new to perplexity just wanna ask what llm model perplexity ai using for basic search

[thick dust]

Sonar, Perplexity's own fast llm

[tame current]

oh ok tq 😀

[thick dust]

I’m officially shocked by how much more you can do with Perplexity Pro on the web vs the iOS app

Never seen anything like this

[north lily]

Do we really need perplexity?

[thick dust]

well, it depends

i really like perplexity, it helps me tremendously when doing research & academic stuff

[torpid garden]

the new web update just removed model selector and u cant even select it from settings tf 😭

[tame current]

I'm gemini user and not a fan of chatgpt. just wanna give a try to perplexity . let see which one is better both can do real time search

[thick dust]

wat?!!! dang that sucks

[torpid garden]

prolly a bug

using complexity for now

[short narwhal]

Gemini is not good at search

ChatGPT and perplexity compared to that is much better

[tame current]

what happened to the model selector?

[short narwhal]

Moved

[tame current]

yap people saying lot like this Gemini not good at search, using almost 1 year i don't know apart from controversial topics Gemini is good for me .i don't search controversial topics 😄 😄

[topaz rune]

hi guys, did u have pro too? why im unable to create immages? in the related settings it's showed that dall-e 3 is available

[worthy harness]

Is it just me or I can't see the model selector again

I'm using Complexity too

[torn matrix]

Hello ,Why has the R1 reasoning disappeared?

[worthy harness]

Oh well no choice but to use Claude 3.7 Reasoning then

[clever oxide]

#💬│general#mywins#success#sales
#mywins#success#sales

[opaque stump]

big same

tired of paying for this carousel

clicking here

brings me here