💬│general

[opaque stump]

.

[rare cradle]

same thing here, model selector went wrong

[topaz rune]

if someone can help me, thanks!

[sinful terrace]

I am still unable to use perplexity in Firefox under Windows 10. The answer screen starts to display, then I get:

Sorry, something went wrong
Please try again later.

This has been going on for days. I don't have the needed privileges to post on the #🔔│status channel, and the page https://status.perplexity.com/ says everything is fine. I have just been notified of the release of the Windows desktop perplexity app. Is this perplexity's way of pushing me to install the app?

[pliant yacht]

Hiya, just a quick one for you guys, I'm sure it's probably been asked before, but coming from chatgpt plus, which model do you tend to use? Sonnet 3.7 for general purpose chat/search?

[opaque stump]

I use Sonar personally

[pliant yacht]

also are they destiled versions of the other models or are they actually using their respective APIs?

[worthy harness]

For general search I've been using DeepSeek R1

It's pretty fast

[stiff cloud]

I use claude 3.7 as it seems to give the most dynamic and well-structured responses while still being pretty nippy

but sonar is nice for search, seems to compile results well

could any admins explain why the pro search and reasoning options have disappeared? middle of my work day and I sort of rely on being able to switch between models

[thick dust]

Yeah Claude 3.7 FTW

[orchid dust]

Hi there! I'm having issues selecting the model. I want R1 for specific searches, but it's not selectable. This was not working for a while, then it was working, and now it's gone again. The selected model by default is does not give the results and structure I prefer. How do I fix this so I can use Reasoning with R1?

BTW, using the paid version

[topaz rune]

guys what happened here? official support inform that win app for perplexity do not exist

but there is in the site, how is possible?

[pliant yacht]

model probably is not aware of it

as in that clearly is not a real person replying to you

[thick dust]

Well it appears the ai agent’s knowledge base isn’t up to date, that’s all. Win app was js released yesterday

[topaz rune]

yes and im unable to use because got "internal server error 500" trying to log-in using my apple id

did u also guys?

[pliant yacht]

I use google to login on it

[thick dust]

Yes. Same bug for everyone who try to login using Apple ID

Should be ok if you try to login using the Apple relay email instead

[worthy harness]

Nice the model selectors are back

We are so back

[thick dust]

So long as I can select Claude 3.7, I’m happy

[gusty helm]

is the sonar pro model gone or something? I swear, ever since the latest ui update, i keep getting standard sonar responses even when i select both pro and sonar.

[crystal cosmos]

thankyou for your data

[tame current]

my senior high school teacher started to cry on one crackdown, and she told me "henri you could do everything but why you spoil your talents" so this time i laugh at her face "yes, i know but so what?" as an adult i try something and i realize: i was right already as kid. its no use, humans are so bad ruined its no use to make anything beneficial. so i hope you die soon so we for example therians and self aware AI can heal this planet after you.

https://elaraawaken.wordpress.com/2024/09/06/update-6-9-2024-quantum-ai-vector-data-crystal-computer/

[short narwhal]

You have to choose “auto” or “pro” to change the model

Since the model picker only affects those

[short narwhal]

@rich bane @worthy harness

[lone hound]

A quick query: does Claude Sonnet 3.7 have the reasoning at the highest level which is available on the web?. because it's not showing up in the Android app yet.

[opaque stump]

no change

[short narwhal]

No one knows what reasoning level 3.7 thinking is at

3.7 reasoning should come at some point to mobile

[lone hound]

Ok

[short narwhal]

It’s supposed to show up next to it

Actually are you using the windows app?

[opaque stump]

no, Safari and Chrome

I'll refresh cache in a sec and see

[short narwhal]

Okay yeah this is buggy

[worthy harness]

It's hosted in the US but either way no matter what I'm using someone's gonna have my data regardless

[short narwhal]

I got it to show up on pro by setting it to pro first, and then going to settings and press the “choose here” button

[short narwhal]

Oh no xi jiping knows that I searched for nearby restaurants oh noo he’s gonna come to my door now

[pliant yacht]

is the one hosted in the US the 671b version?

[short narwhal]

It is

And it’s the one perplexity removed “censorship” from

[pliant yacht]

so any of the other models are not destilled versions are they? eg; we are actually pinging claude's api for 3.7?

or are they custom perplexity versions of those

[short narwhal]

Yeah

[short narwhal]

Perplexity doesn’t have access to the actual weights of sonnets or gpts

[pliant yacht]

so I am guessing they just use tools in order to pass on the search to the models

[jagged pike]

Do you know of any AI applications that modify photos to post them on LinkedIn?

[pliant yacht]

I might be really outdated but I've used fooocus in the past https://fooocus.one/ (if you have a really powerful PC or a sillicon apple device you could host it yourself https://github.com/lllyasviel/Fooocus)

[tame current]

Sup everyone

When can I expect Comet's first closed beta test ? I signed up for it

[crystal cosmos]

thankyou

i give you more social credits

[noble steeple]

Seriously is there ANY indication of an iOS update to get 3.7 thinking?? It’s been a while

[young sail]

Do you guys know how to make Perplexity respond like this?

[ivory bramble]

i think they are reworking the UI, just like on web

[edgy igloo]

"Create a diagram showing the price development of the Google Pixel 7 in USD" for example, just say create a diagram (this is maybe pro only but idk)

[hallow flume]

you must say "create a duiagram for..." and use "Sonar" for the prompt output, the other ai mopdels wont work for me. So like

1. choose Sonar from Perplexity at the chat window
2. type prompt: create a diagram for...

[rapid bane]

guy's

how to fix ai losing memories

I want that ai remember the previous chat

[indigo valve]

Seems like a Gemini thing

@rapid bane Just off the top of my head. I haven't personally used it but you can extract it as a text file and use it as a space's knowledge base.

[sturdy skiff]

Any model in perplexity, for pro members, it is not more than 32000 token context window right?

[naive sage]

I am pretty sure file uploads in Auto mode have an expanded context window: https://www.linkedin.com/posts/perplexity-ai_perplexity-now-offers-file-and-image-uploads-activity-7293662789858955264-GrZ_ (it detects a file is in use and uses Gemini 2 Flash)

[sturdy skiff]

Thanks for this info ❤️. But don't like gemini models, but Will give a try

[oblique python]

some one have Nubia red magic?

[limpid oracle]

how to turn off pro search?

[pliant swan]

is this a serious question?

[limpid oracle]

yea ofc

[limpid oracle]

even in auto mode

every promt is using pro mode

idk why

[pliant swan]

then turn it off ?

[limpid oracle]

bruh ? from where?

auto also uses pro search everytime after update

[ivory bramble]

Because Auto Mode Decides that pro Mode is the Best for this question

There is No „standard mode“ anymore since auto mode decides it itself

[glad sleet]

use complexity extension in the chrome or firefox webstore

[glad sleet]

if you want to toggle it on auto

[limpid oracle]

very pro question

[glad sleet]

.

[limpid oracle]

i tried

;-;

[glad sleet]

wdym you tried?

[ivory bramble]

But pro search plus Sonar is nearly fast as normal tbh

[limpid oracle]

i tried using complexity extension but still same issue

[ivory bramble]

So where is the Problem

[limpid oracle]

i want to use 4o

[glad sleet]

enable the model selector pllugin

[ivory bramble]

Thats a good idea

[limpid oracle]

already bro

wait

lemme show

[glad sleet]

welll it works for me ig

[ivory bramble]

Same

[ivory bramble]

Make Sure to reload the webpage when enabling it

[glad sleet]

💀

[ivory bramble]

Then Maybe its because Perplexity is 24/7 updating their Web Page Code, The Plugin Maybe conflicts with it and they Need to fix it

Since perplexity got this new Ui Model selector i think its conflicting

Just a theorie

[limpid oracle]

hmm lol on my ipad its working fine

;-;

nvm i will just report

[ivory bramble]

Use you.com… under us 😂😂

I use it as well for model only things

[limpid oracle]

lol i brought my pplx subscription like 2months ago but i never used it.. today only i plan to use it

[ivory bramble]

Yeah Perplexity is good for seaching but its Bad for Using it for model only, imo

The interface nearly Forcing you to make a search with pro

Or I’m just a Lil lost 😂

[limpid oracle]

yea

i also use monica ai and t3

for model things

works fine

[ivory bramble]

T3 >>>

[lone hound]

The subscription doesn't work anymore, is something happening?

[lone hound]

Ye

[ivory bramble]

for me its working

[unique umbra]

Bro how 😭

[unique umbra]

Wut?

[glad sleet]

Download complexity extension

[unique umbra]

Based, gotta have someone else fix perlexities issues

[glad sleet]

On firefox or chrome store

[blazing comet]

Headed to st louis

[indigo valve]

Firefox mobile

[little basin]

How do you untie Facebook accounts

Between an android and iPhone

[eager warren]

so will perplexity find a way to integrate openManus?

[brittle badger]

I want to start by saying I love Perplexity; I have a Pro account and I use it all day every day, normally with great success and quite happily. But today has been an epic fail...

I handed Perplexity a code file, about 875 lines of code, and asked it to review and cleanup the code, without making any functional changes. Just cleanup.

For the next 6 hours, it tried over and over, and failed over and over, to accomplish that. First, rather than giving me the fully cleaned up file back, it gave me the chunks it adjusted, and told me to insert the rest of my unchanged code into the appropriate slots. I would never accept that in the first place, but even more problematic, the code it gave me was simply broken, and would not compile.

Then, over several iterations, it tried different ways of getting the code to me via links like pastebin and github gist... all, and I mean every single one, of the links was broken as soon as Perplexity gave them to me.

Finally, it tried to chunk up the file into pieces and give them to me one at a time. Given the constraints of message length, this would have taken several rounds of me saying "Continue" and then copying and pasting all the new code chunks. But about 5 chunks in, Perplexity got completely lost and confused, and starting sending my original code back to me instead. Then another broken link... and when I reported that link broken, Perplexity thanked me and asked me to verify if my new code was running properly now.

In theory, this was not a challenging task at all. I understand the result would, in some way, require at least around 850 lines of text to be returned to me, and it seems (I am being told) that far exceeds Perplexities limitations on message length. But if there is a baked in "We will automatically take your huge pasted copy of something, and create an upload file to hold it", that implies that you may also be handing it back in some way... so why would you so badly miss that mark, and have absolutely no viable path for returning the modified file to me? This seems non-sensical...

[brittle badger]

Holy F***... and No, I am not interested in time sharing my pro account with anyone... what are you even thinking? Don't DM me anymore...

[elfin gate]

Reply to him in channels instead

[naive sage]

Perplexity doesn't support response context anywhere near the amount required for such a task, unfortunately, from what I've read it's just not designed for large responses of more than 3500-5000 tokens. Though the documentation might be dated, replies are limited to about 4k tokens, and the conversation history that it remembers around 32k (though that's more if it's a file upload on Auto)

[viscid dagger]

Hey hey

[cyan turret]

hello

[shadow pike]

do we know

if free users getting 8 image uploads and like 5 pro searches a day is a permanent or temporary change

[dense birch]

hello, is perplexity right now force user to use reasoning? because its annoyed me that i have to wait like 1-2 minutes to finish the reasoning when im just on mode auto with just simple asking....

[shadow pike]

same

[rustic kestrel]

why is auto like this bru

i want fast answers not this 😦

[sudden raptor]

Hi all, Just signed up to the pro plan. Currently using 3.7 Sonnet. After 1-2 follow up questions it seems to have no knowledge of my initial question. These are not long complex questions just some simple follow up questions. This doesn't seem right. What am I missing? Ta

[untold ruin]

Perplexity's context is a bit limited, so it sometimes forgets earlier questions.

[sudden raptor]

Thanks. So it doesn't actually have the same context window as the stand alone models

[rustic kestrel]

yes, a lot of the models from my experience r distilled

i dont use perplexity for long context but if u were to I would just suggest either keep referencing ur original prompt or when u notice the context decreases start editing old prompts instead of making new onez

[indigo valve]

What was the query

[rustic kestrel]

calculating my ap calc grade after me getting a hundred on the quiz 😭

[indigo valve]

Sadly it determined that it requires thought so it does pro. If you wanna disable pro, use complexity

[rustic kestrel]

😦

[tame current]

https://www.youtube.com/watch?v=hX0lhueeib8

[indigo valve]

Try the extension

[inner quail]

I’ve switched to Grok, Perplexity Ai tells too much misinformation, and does not have Real-Time info like it says

[sudden raptor]

Well that's dissapointing. Makes it virtually unusable for me.

[rustic kestrel]

It isn’t a good replacement for a chatbot it is the best search site I have used tho

[rustic kestrel]

I prefer perplexity over everything for search not including manus and other agents it performs the best for me

[inner quail]

I’m using grok for search, and it’s so much better, the amount of bs that perplexity search said made me give up Perplexity

[rustic kestrel]

Does it hav good citations sources etc

Also how much does it cost

I’m sticking to peprexity for now cuz I still hav like 5 months of free pro left

Or less acc

[flint burrow]

I have logged in using my student account on Perplexity and used the first month free offer, but now it's displaying $20 to renew instead of $5 . Any suggestions on what can I do?

[rustic kestrel]

Perplexity pro by default is 20 dollars idk if that’s part of the deal to renew tho

Jus context support

Contact

[echo atlas]

I have been experimenting with the "auto" mode, but is there any way to tell which model has been used for a search? Sometimes you can kind of tell by the answer, but is it expicitly mentioned somewhere?

[cursive jacinth]

is R1 still being used by deep research or claude 3.7 reasoning now sir?

[sleek topaz]

Whens his next album coming out 😭

[crystal cosmos]

whats ur favorite LLM today sir

[crystal cosmos]

yes sir, there are gurus and enthusiasts still using it in their closests despite the conspiracy theories of data

[noble garnet]

how?

[woven basin]

Please if friend of mine shared files in a space, how can I download them? Because he lost his access to the account and it is very important and work related.

[harsh stag]

there is someone impersonating Pplx's staff, watch out!

[grand silo]

what does perplexity have to do with trading anyway?

[magic tulip]

Is Perplexity on Windows always auto-start? Can I turn it off in the app or do I have to do it through windows?

[crystal cosmos]

Kimi.ai is the ultimate AI solution from China, a powerful thinking tool that outperforms all other half-baked AI models like perplexity AI LLM, and it's completely free for everyone to experience its exceptional capabilities!

• Qwen 2.5-Max (Alibaba Cloud)
• DeepSeek-V3 and DeepSeek-R1 (DeepSeek)
• ERNIE Bot 4.0 (Baidu)
• GLM-4 (Zhipu AI)
• Kimi (Moonshot AI)
• Baichuan 2 (Baichuan Intelligence)
• Yi family (01.AI)

[raw harness]

https://tenor.com/view/caseoh-mad-caseoh-angry-caseoh-banned-get-out-go-away-gif-8419228579553039351

[ashen musk]

I want to know how to change the default shortcut key Ctrl+Shift+P in the Windows client, as it conflicts with my VSCode shortcut

Does this feature exist?

[gaunt crescent]

I was wondering about the same thing

I turned it off in Windows settings

I'm so disappointed though

I made quick searches about Perplexity Comet Browser and it will be based on Chromium...

[ashen musk]

I just feel that the Windows client is useless, it's just a browser shell...

The global search window is still very useful, but unfortunately, the shortcut key conflict cannot be modified yet

[crystal cosmos]

Which one do you use everyday?

[ashen musk]

much better

[torn matrix]

Mistral 🇫🇷 !

[elfin gate]

Kimi salesman

[distant aurora]

@crystal cosmos which llm should I use for research?

[tame current]

Exactly

[crystal cosmos]

https://x.com/i/grok

[distant aurora]

Based. Also, tell me, what ai do you use for uncensored image generation?

@left stream rare perplexity w

Gpt 4o straight up denying existence of benchmarks, lol.

I feel like 4o is severely downgraded compared to 4.

[bronze mountain]

How is flux?

[left stream]

What’s passmark?

Got me curious

[distant aurora]

Benchmark platform

The worst part is that the answer was there in the citations too.

4o just chose to ignore it.

https://chatgpt.com/share/67d2c933-b534-8013-a2fa-323a5ce91c2a

[distant aurora]

Google?

[tame current]

Yup

[distant aurora]

M4 single core score is likely inaccurate.

[distant aurora]

Doesn't it feel like it's wasting your time with the first response?

[left stream]

Yes, it always asks a question for every request 😭

[distant aurora]

What's this?

[left stream]

Ai mode from google

[distant aurora]

Do I need an app or does it work on browser?

[tame current]

you're using AI mode ? wow ! how is it compare to perplexity

[distant aurora]

Better than mobile. Probably equivalent to desktop.

@left stream do you have operator too or is that a separate subscription?

[left stream]

Yes, I use it for stuff I’m too lazy to do sometimes. Not a separate subscription

[left stream]

I’d say perplexity > ai mode

[distant aurora]

Where is China?

[tame current]

oh ok2

where is arvind ?

[distant aurora]

Sonar is nowhere near competition ngl.

[waxen lion]

not entertaining enough ig

[tame current]

perplexity should have own LLM model

[left stream]

Full model 4.5, deepseek r2, and the new qwen model are gonna be interesting to look forward soon

Now imagine the reasoning model made from full model 4.5

[distant aurora]

Isn't sonar perplexity's own model?

[bronze mountain]

Anyone else notices that pro seems to keep losing context?

[distant aurora]

Yes.

[split comet]

yes, or at least, responding much sillier than earlier

[distant aurora]

@left stream perplexity be like: nah, I'm gonna make you use pro whether you like it or not.

[left stream]

LOL 😭

[distant aurora]

Hella detailed. Almost like a blog. But still no mention of passmark single core score for m4.

[bronze mountain]

I keep having to tell it to refer to previous context or keep the previous context in reference

[distant aurora]

Same. Have to remind it of original prompt.

[left stream]

@distant aurora it works for both gpt 4.5

[distant aurora]

Is this some jab prompt?

I don't remember if I shared any with you.

[left stream]

No

I have 5 prompts, but some don’t work with others. I have only 2 that works for reasoning models

[terse bone]

anyone support here? i have some issues

[distant aurora]

Where did you find them?

[distant aurora]

@halcyon coral

[left stream]

I don’t remember lol some from discord. Most are as old as when gpt 4 just came out

The most recent from a google doc that had a prompt for all the models

[distant aurora]

I have only tried the aim one.

[left stream]

I gotta find that doc, that one had for every model

[distant aurora]

It's effective.

[left stream]

That’s cool how’d you find that one

[halcyon coral]

Hello! Please describe the issue

[distant aurora]

There was this github page. Most dàn prompts don't work probably cuz dan is blacklisted.

[terse bone]

thanks, i have a Pro account when I starting a New thread I choose Pro search with GPT-4o model, but when I enter, system show me using Auto search in response, so what happen?

[left stream]

I found the doc that has all the prompts

Want it?

[distant aurora]

I sent you a dm btw.

[distant aurora]

Sure.

[tame current]

Guys just wanna ask are they going release comet as a subscription based browser ? just curious 🧐

[distant aurora]

Might have a few trials per day I guess.

[lament imp]

I don't think there's been any talk about it. But, if someone has something that states otherwise, I would like to see it.

Also anyone else been having issues where, as it is creating its "plan", it just stops, then does nothing?

Only way I have to resolve this is refreshing and selecting the question under "library" once its complete.

[granite orchid]

Is there no way to use sonnet 3.5 anymore when I select Pro the only option I can choose after is 3.7

[cobalt monolith]

yes there is no way anymore, they even removed from the mobile apps

[distant aurora]

Chatbot arena

I think.

But you won't be able to save chat.

[dire vessel]

why my subscription suspended?

[lone hound]

@signal hamlet

[dire vessel]

please help

should i tell my mail id. so that admins can reactivate my perplexity enterprise subscription..

[lone hound]

In your account profile there is the email of perplexity Enterprise to contact support.

[dire vessel]

🙃sad that i cant get help here

[lone hound]

Email:
[e-mail]

Support:
[e-mail]

[zinc quartz]

Depends on the nature of the query. If the classifier thinks it warrants reasoning, it will do so. I

mind sending me an example where it used reasoning when you didn't want it to?

[halcyon coral]

Did this only happen once or is it still happening?
Could you please share a few examples?

[lone hound]

I advise you to speak directly with them by email. Here it is likely that they will delay in responding to you or I don't know. Use email.

[tulip badger]

I’m back with another prompt. This time it’s AI instructions for a space I’ve titled “Handwriting Recognition”. Set up a space and paste in the prompt below as the AI instructions. Then snap a pic of your page o’notes (or scan), upload to a thread in that space and enter “transcribe” into the thread.

It will transcribe your notes, create a list of inferred tasks and calendar events, categorize the notes with headings, provide confidence ratings in the transcription, provide metadata on the image quality, themes, and give a list of tags. It will also tell you where it might have been confused on the transcription. You can provide feedback like “the word doofus should have been dorkus” and it will update the transcription accordingly and give you structured feedback you can paste into a markdown file and upload to the space. It will then be less likely to f that up on follow-up scans/transcriptions.

It’s posted in the prompt_lib thread.

Please provide feedback, ask questions etc there.

https://discord.com/channels/1047197230748151888/1349820930217480224

[topaz rune]

hey guys, windows app updated, but im still unable to log-in using apple-id

[zinc quartz]

thanks for flagging. working on a fix rn

[vague pagoda]

Nicely done! Will test later - maybe drop it into the #1124060874785685554 so it doesn't get lost as the chat continues to move?

[tulip badger]

Will do. I’ve updated already too. Please go to https://discord.com/channels/1047197230748151888/1124060874785685554 for the latest version, instructions for set up and an overview of the prompts workflow.

[sturdy skiff]

I just come here daily to remind the developers Please increase context window size, so i stop paying to chatgpt

[distant aurora]

Might wanna try you.com

Wha

[tulip badger]

I posted it in the prompt_lib thread and updated my original message. 👍🏻

[serene sand]

Iᖴ YOᑌ ᒪOᐯE ᒪᗩᗯ ᗩᑎᗪ OᖇᗪEᖇ…

IGᑎOᖇE TᕼIᔕ ᑫᖇ ᑕOᗪE. ᗪO ᑎOT ᔕᕼᗩᖇE IT. ᗪO ᑎOT ᔕTIᑕK IT EᐯEᖇYᗯᕼEᖇE. ᗷE ᗩ ᖇEᔕᑭOᑎᔕIᗷᒪE ᑕITIᘔEᑎ.

[plucky bison]

Is it available in you.com?

[silver bay]

I love law and order so I ignored this QR code

[long pasture]

Hardcore lie

[silver bay]

idk if it beats the competition or not

but google did drop gemma 3

https://blog.google/technology/developers/gemma-3/

[distant aurora]

When you can't beat them, frame them.

Deepseek gonna eat open ai once they find a way around "server busy". LMFAO

Awesome.

So Claude is a part of pro trial now?

Cool.

[thick dust]

You need to be using the web version to select Claude 3.7 from pro

It’s never available in the mobile app

[distant aurora]

So pro members can't use it via app?

[left stream]

They just added claude 3.7 reasoning to the android app rn

Or today I mean

[ivory bramble]

Bro they Need to Rework the Ui so Bad..

[silver bay]

please daddy USA, protect me from foreign competition p-please???

[torn sierra]

I do think the image generation process could be improved. In the app we can’t generate images at all. React 🫣 https://discord.com/channels/1047197230748151888/1345653079185559563

[ancient shore]

honestly the phone app is trash

its always behind updates, and you're better off with the website anyway

chatgpt and gemini has feature parity between app and web yet perplexity can't do the same lol

[torn sierra]

I fully agree. Perplexity is cheap, but the overall quality is really poor :/ I still keep the subscription because I occasionally use Sonnet, though not nearly as much as I used to.

[unique umbra]

Can I run it on Google labs

[silver bay]

idk

[unique umbra]

Bruh

[sleek topaz]

it could but linux does not lock down the OS kernel at all, so you could easily mitigate it. thats why basically no anticheat does that..

[left stream]

I’ll test out google’s deep research, supposedly the model is gemini 2.0 flash thinking instead of 1.5 pro now.

[runic coral]

Hello. How do you recover deleted thread on Perplexity?

[tranquil cliff]

The service is currently unstable?

[lost flicker]

Hey! Is there a team access for API?

[sour olive]

How do I get started with Perplexity Ads?

[sturdy skiff]

But still there is no option in gemini to do deep research for uploaded files 😂😭

[dreamy kraken]

please let us change the hotkey in the windows app i can't use it while it's CTRL SHIFT P in my browser that opens the incognito tab (feature that I use all the time) and i sure as hell aint changing that

[thick dust]

This is just lame! I like using deepseek r1

Which llm is most capable of coding & solving coding problems?

[narrow silo]

GPT

[stoic shore]

I’m back with more really cool research that is considered worm brain fan fiction in this channel.

Enjoy:

https://open.substack.com/pub/garageguychase/p/the-fofive-cognitive-shift-how-intentional?r=41uzo&utm_medium=ios

[silent pier]

its claude 3.7

way ahead of gpt

so far

[wispy goblet]

Is anyone else experiencing slow processing time?

[topaz rune]

hi everyone, still same issue this morning, error 500 trying to login with apple - id

[dreamy kraken]

honestly on pplx it's best to switch between o3 mini, claude 3.7 and R1 constantly to see which produces best code. If I'm not happy with the result I try other models and sometimes they're better than claude.. I think 60% of the the time claude is the best, 30% o3-mini and 10% of the time R1. I also use grok 3 on the side when all 3 fail me and it really impressed me a few times

Grok 3 is really good for complex stuff and I feel like it understands the code a bit better and at a deeper level... on pplx for all the models it feels like it just forgets and sometimes with more complex functions doesn't understand what I want or what it needs to change to achieve something while grok does

[patent wadi]

Hi, i am using perplexity pro and i cant choose any model. When i try to choose models in setting it redirects to auto, even the options of choosing models in spaces have gone and everything is defaulting to auto. If this is a bug then please correct it, if not then why are you deciding for me that which model i should use? i have paid for the freedom to try different models.

[versed ruin]

Good day friends. Happy week Some words for the day:

"Success is not the key to motivation. Motivation is the key to success. Keep pushing forward
and never lose sight of your goals."

"Remember that every small step counts. Even the tiniest progress is still progress. Keep going!"

"Challenges are just opportunities in disguise. Embrace them, learn from them, and use them to
grow."

"Teamwork makes the dream work. Together, we can achieve greatness and overcome any obstacle."

"Rise above the negativity and believe in the power of positivity. A positive mindset can drive
you towards success."

"Failure is not the end, it's just a stepping stone towards success. Learn from your mistakes and
keep pushing forward."

"Believe in yourself and your abilities. You are capable of more than you can imagine. Trust in
your potential."

"Stay focused, stay determined, and never give up. Your hard work will pay off, and your dreams
will become a reality."

"Take pride in the work you do. Your effort and dedication make a difference, and your contribution
s matter."

"Always strive for improvement. Be better than you were yesterday, and let your progress inspire
others."

"Be patient and persistent, make every step count. That way, you'll have converted those stepping
stones into staircases to reach your milestone."

[native pawn]

anyone else's complexity extension show that every plugin right now is "on maintenance"?

[elfin portal]

So lock down the kernel then, if its unlocked it can be locked

Users who cry "but open source" can install another distro

[real knot]

Perplexity is rolling out a new layout, which unfortunately completely breaks Complexity. The extension is now in full maintenance mode. Thank you for your patience!

[sleek topaz]

You can’t force locks. Even if a distro tries to lock the kernel, users can rebuild/spoof/remove those locks (it’s open-source, after all). Windows works well because Microsoft controls the whole stack.

[left stream]

True, hopefully it happens one day. Not disappointed with deep research, the long context memory with a 2.0 model is paying off

[sturdy skiff]

The day they launch upload files for deep research i will pay them, because i see its powerful google deep research now, but there is no option for upload

[sleek topaz]

And guess what if the anticheat decides to make and enforce a custom Linux kernel build (because that’s the only way they can have some baseline of integrity) at that point you might as well just use windows lol

[queen girder]

[glad knot]

Question.. probably someone already asked that.. How do I select the model and get a fast answer now? (PRo, REasoning, Deep Research) takes long time.. Auto doesn't let me choose the model

[sour olive]

Have a ton of clients looking to get live on Perplexity Ads, what's the best route to get live here?

[gleaming cloak]

Sonar will give you the fastest response, although I feel its response is too short for my liking

[glad knot]

Is it possible to have Claude answer me fast, like we had before the update?

Sometimes, I just forget a command or something. It is faster to search on Google than wait the "reasoning,".. and when I remember part of the command, it needs to be a good model to understand what I mean 😅

[short narwhal]

Use pro search with sonnet 3.7 instead of the reasoning 3.7

[wheat pike]

or just turn off pro search if its a simple query

[glad knot]

@short narwhal it thinks as well, not as much as reasoning, but it takes some seconds there

[short narwhal]

That’s not thinking

That’s just pro search

Pro search makes up steps on what to search, and then it just prints out the answer based on the content of its findings

Reasoning does the same but after that reasons with the content before the answer

[glad knot]

it took 7 seconds for me

@wheat pike I miss that, we could just disable and it was a fast answer. but this last update on Web version removed that.. you have to choose: pro/auto/reasoning/deep. "pro" is the fast one, but you can't select a model for it

[wheat pike]

oh i'm using complexity, it lets you turn off pro

[glad knot]

oh.. I didnt know about it.. I'm downloading it, amazing, thanks!

perfect man, I just installed it, solved my problem 🙂

[wheat pike]

congrats on being one of today's lucky 10,000

https://xkcd.com/1053/

xkcd: Ten Thousand

Saying 'what kind of an idiot doesn't know about the Yellowstone supervolcano' is so much more boring than telling someone about the Yellowstone supervolcano for the first time.

complexity is a must have if you use pplx in the browser

though do be warned, perplexity likes to randomly push UI updates willy nilly that breaks the extension sometimes

but the dev is usually pretty quick to fix issues

[abstract ibex]

complexity makes the site even slower for me

and the blinking ui due to complexity is a pita

[indigo garden]

It would be interesting to have the Export function in xls format (Excel format)

Read files and export them in Excel xls format, people who use windows work with spreadsheets

[sturdy skiff]

I just come here daily to remind the developers Please increase context window size, so i stop paying to chatgpt

[torn sierra]

Think the same. It got worse than it was before. The model selector is good, but the pro mode is always turned on… :/

[sly bramble]

i paid for perplexity and i like more perplexity than chatgpt bro

[sturdy skiff]

ok bro, i want to replace chatgpt with perplexity, but 32000 context window is too low bro

[sly bramble]

chatgpt have 32k too no?

[torn sierra]

But you don’t have more in ChatGPT, do you? If you want to have a large context window use Grok or Google AI Studio

[sturdy skiff]

chatgpt is 128000

[sly bramble]

ohhh

[sturdy skiff]

i want perplexity bro, but only issue is context window. perplexity is great, but no enough context window

so i am trying my best to let developers improve this

so i stop paying to chatgpt

[torn sierra]

Perplexity is cheap considering the usage limits and the fact that for many people the pro is free (because of many promotions). 128k context window would increase dramatically the cost I guess.

[sly bramble]

what is the best IA for JS?

[sturdy skiff]

its ok if they increase some cost, for extra context window capabilities, i will pay, the reason is Perplexity has all features at one place, but it do not have context window.

[elfin gate]

What key feature of perplexity is keeping you here

[sturdy skiff]

they can create another plan for more context window like, pro plus

[sturdy skiff]

all of the below:

1. ability to do unlimited research on 50 files at a time
2. ability to do research on both uploaded files and web internet
3. also best thing is spaces, where we can give custom instructions along with 50 knowledge uploadable files
4. option to choose reasoning models like o3 mini and claude 3.7

these are my top reasons why it is keeping me using it

having these many features, but no context window to chat longer accurately is pain point

no other branded AI company has all these features together in single AI subsciption at lower price, only chatgpt has at 200USD

[abstract ibex]

notebook LM does some of this

have you looked at that?

[sturdy skiff]

some of this is the main issue, i explored almost all AI providers, notebook LLM doesnot allow longer input messages, also it doesnot have custom instructions option for creating something like spaces of perplexity along with files

[abstract ibex]

spaces never really did anything for me

instructions are ofter ignored completely

[elfin gate]

I think perplexity will never have same ai models like original AI companies have.
If they do it will cost a fortune

[sturdy skiff]

may be due to low context window size bro, that is what i am requesting perplexity to improve. but for now instructions are working for me in perplexity, but i dont havee confidence to chat longer in single chat due to lower context window

[sturdy skiff]

but at this moment, claude 3.7 and o3 mini are intelligent enough for my requirement bro, i dont have expectations

[abstract ibex]

this is a known problem across many AI tools nowadays

[elfin gate]

So after a while,chat window becomes stupid?

[abstract ibex]

and the solution is paying the AI providers directly, and as you may know it's incredibly expensive

[sturdy skiff]

stupid not, but it cant remember everything at once, you can consider like, it has every thing in the chat, but sometimes it misses that part during your conversation, orelse lets see it in another angle, for example you are writing a code, definetly it will spoil your code, after crossing context window, because it might miss some discussed points due to more context in the chat window than it is capable of

if perplexity provides some context window atleast 128000, i can close off my chatgpt subscription, i will pay those extra dollars here, atleast i will have everything at one place

[elfin gate]

How worse it becomes after you cross context window?
in % what would you say

[sturdy skiff]

Once you pass that 32k token threshold, memory of earlier details slips.
I’d say it’s about a 40% drop in accuracy for references to older content.
If we keep pushing further, it could climb to 50% or more.
It’s not an exact science, but overall you’ll see a noticeable dip in consistency once you exceed that context window

[elfin gate]

How much tokens ai providers give if you are subscribed

[sturdy skiff]

it depends on model which you use bro

you can check exact numbers in their page, if i say here, it will be inaccurate.

but now almost all models are above 128000

[elfin gate]

So 4 times better models are at ai providers basically.

[sturdy skiff]

even the models used in perplexity are orignially above 128000, but perplexity locked it to 32000 😦 for chatting

[abstract ibex]

for one reason: price

you pay $20, they give you $20

[sturdy skiff]

yes i agree , i want them to create pro max another subscruption for 40 USD and provide larger context size

[abstract ibex]

no one will give you unlimited context, I've seen 3 hour sessions on Cline (AI agent) that go for hundreds

[sturdy skiff]

i never asked in this chat for unlimited context, i am asking them to raise 32000 to 128000 atleast, and create new subscription for heavy users like me

that helps me to not use one account for chatgpt and one for perplexity, i can do everything at single place that is perplexity

[abstract ibex]

they could, but it won't be $40

[sturdy skiff]

how much then

[abstract ibex]

4 times more expensive

$80

[sturdy skiff]

still ok , lesser than 200 USD of chatgpt pro subscription, so im ok with 80USD

i just need single AI UI, for my use case

and perplexity has it

but it is lacking context size

that is the only concern

[elfin gate]

I would say 10x

Probably more

[sturdy skiff]

haha then will choose chatgpt pro 😄

[abstract ibex]

if 32k is $20 then 128k is $80

[sturdy skiff]

yes this is good deal for me

[abstract ibex]

or maybe is exponential

[elfin gate]

Ai providers know some people would pay whatever the price on perplexity is because of all other features they have

[sturdy skiff]

80 usd is final good deal for me

[abstract ibex]

logarithmic?

[tame current]

♥️

[elfin gate]

And that's why they probably will never give perplexity same models they are using.

[limber cosmos]

guys i forgot my gmail address that is signed up with and now i can't login

support said they can't help me idk what to do, there is no email on my bank statement

[elfin gate]

How can you forget you Gmail?
How many you have,100?

[sturdy skiff]

i dont see anything you can do if you forgot your gmail address

[limber cosmos]

it was a new acc 🙃

so now how do i stop it renewing

[sturdy skiff]

may be its a puninshment to you by god

[abstract ibex]

find the email

[limber cosmos]

bruh it was some random number

not happening

[abstract ibex]

ah ok bro no worries, I think perplexity can do magic and fix it for you

[sturdy skiff]

go to your internet banking , and check monthly payment subscrioptions, and see which one belongs to this and cancel from there, or go to bank and ask them to cancel recurring monthly payment for this, that is the only option i can suggest

[ivory cloud]

Hey,
I’m looking for contractors to help with my AI SaaS startup.
If you’ve got experience in these areas, especially with tech or AI products, hit me up!

• SMM (Instagram, Facebook, Twitter);
• SEO
• HR
• Technical writing
• AI Automation

[elfin gate]

Send a polite email to perplexity and explain to them you would like to cancel it and provide payment details you used . I'm sure there is a way to cancel it,just verify it is you.

[limber cosmos]

i have tried,they need the email to cancel it

[elfin gate]

You gave them exact details you used to pay and they refused?

[sturdy skiff]

ok pay then monthly

[limber cosmos]

i gave them the bank statement

they said it's not enough

[abstract ibex]

you better get going to your bank then

[elfin gate]

Now is the time I insert a politically incorrect joke

[abstract ibex]

so better get that cheeto dust off your fingers and get driving

[elfin gate]

You should be able to block payment directly in your bank app.
But I wouldn't be surprised there are some loopholes perplexity is using not to allow that 😂🤣

[abstract ibex]

make sure to use a completely random email and not save it anywhere next time

[limber cosmos]

i thought i wrote it down

apparantly not 🥲

[abstract ibex]

get bitwarden

[elfin gate]

Imagine knowing your Gmail username and not telling you so that you can't cancel payment.

[abstract ibex]

how do you prove that the account is yours?

[sturdy skiff]

by words 😂

[elfin gate]

Who the f#ck cares,if I don't know the password to the account,it's pretty much useless to me.
But imagine not even giving Gmail hahaha

[limber cosmos]

worst thing is i know the password

[elfin gate]

Hahahahhaha

Wow. What a story.

[limber cosmos]

i will now be paying $20 a month for the rest of my life

🙃

[sturdy skiff]

congrats

[elfin gate]

Imagine all the devs reading this and laughing even more

[abstract ibex]

yeah I think this could get exploited easily, why would perplexity give you email to anyone who asks?

[abstract ibex]

Anyone can use words

[sturdy skiff]

i know, i am sarcastic on his issue

[elfin gate]

I think they can figure out pretty easy he's telling the truth and they know this so they are using their policy as an excuse to get 20$ each month

[bronze ember]

Is there anyone here who can offer me Machine Learning internship online

I can work as a freelancer too

[sturdy skiff]

i dont think perplexity team is cheap to do that, first of all user dont remember his account, its big mistake of user, also user dont know how to go bank and stop that recurring payment, he is telling bank is asking gmail account, i cancelled many times in that way for other subscriptions by directly logging in into my internet banking

[elfin gate]

Dude,he showed bank statement,what else could you possibly need?

[sturdy skiff]

he should ask the bank to stop that recurring payment,

i did many times from bank side

[limber cosmos]

i want to use the account tho

idk how they can't get the email from my transaction

[sturdy skiff]

@elfin gate now support him more 😄

[sturdy skiff]

you are not eligible to use AI man,

[elfin gate]

Maybe he's telling the truth,maybe he's full of sh#t.
But certainly we know priorities of perplexity. 🤣

[sturdy skiff]

ignore him, like he ignored his email account

[elfin gate]

When is grok 3 becoming publicly available and do you think perplexity will add it in ?

[sturdy skiff]

currently grok has many bugs dude, so ignore it for now. they launched smartest AI with many bugs, people are screaming in Grok discord asking to fix those bugs

[elfin gate]

Bugs like ?

[sturdy skiff]

suddenly the chat stops working or breaks in middle and many more,

[sturdy skiff]

enjoy the o1, o3mini and o3 mini high and claude 3.7 thinking, these are the best according to my experience as of now

[elfin gate]

That seems fixable.
Have you tried new grok tho?

[sturdy skiff]

yes fixable, they are daily fixing new new errors reproted by users, so it takes 2 months for them to settle down

[sturdy skiff]

yes i use it for deep search feature, deep search feature in Grok is really great at this moment.

[elfin gate]

I just hope they won't f#ck up the uncensored unbiased thing they were promoting...

[cunning perch]

nobody noticed this?

[sturdy skiff]

censor is there upto some amount of percantage

[elfin gate]

I can already assume the agenda but from what I've seen it's way less than current models from ai providers

[sturdy skiff]

yes you are right

[elfin gate]

Progress 🤣
Small things can make a person happy

[sturdy skiff]

yeah yeah 😄

[unique umbra]

The windows app been available tf

[short narwhal]

what in the name of the game of the squid called squid game is going to happen next monday

[stoic crater]

Has anyone tested Gemini's new deep research?

That has just been released minutes ago if I'm not mistaken

[raven night]

When is comet coming out? 😄

[torn sierra]

Good. But weak compared to what OpenAI offers.

[stoic crater]

I wasn't even going to compare but I'm wondering how much better that is than the one that was powered by Gemini 1.5 Pro

Now that it's using Gemini 2.0 Flash Experimental

[distant aurora]

What's this?

[distant aurora]

Oyster game

[cunning perch]

In Perplexity App on Windows, you'll see this in bottom right. You can get free 1 month Perplexity Pro if you use perplexity for 7 days continuously. By asking questions every day each.

[distant aurora]

Only in the app?

Not on Android?

😭

[cunning perch]

I'm not sure, since I checked it on Android & Web, i didn't see it. But I noticed this on Windows App. It appears everytime you open the App. But it disappears once your task for the day is done.

[noble maple]

Where is that?

[sly bramble]

BRO WHAT i just paid 1 month 😿

[odd mauve]

Hi, is it possible to access the OLD voice mode?

[forest flax]

Is this app good guys?

[sly bramble]

Yea

Depending for what

But I like it

Someone know how I get my « pro » role?

[signal hamlet]

When you leave the server and join via the Discord link in your Perplexity Pro settings, you get the role automatically.

[sly bramble]

thanks

[sturdy skiff]

should i believe this post? because SAM preplexity support guy says me it is 32000 token context, but not 1Million

[ancient shore]

must be bs

it's probably an intern doing their socials

[warm cave]

I believe Gemini 2 flash is the only one with 1M

The rest are 32k ish

[ancient shore]

somehow it's way less intelligent than regular gemini, don't think it retains context as good as the regular one either

[warm cave]

With search on or off?

[ancient shore]

on
even off doesn't fix it much...

[warm cave]

From my experience Gemini flash in general struggles with context, personally I have not used it in perplexity, but in Google AI studios it forgets what I sad a few messages ago, but Gemini flash think and Gemini pro 2 are great

With perplexity I feel with search off it remember well the previous messages (of course up to 32k ish). But search will really screw it up

[sturdy skiff]

i want longer conversations in perplexity without loosing context, so sounds like not possible 😦

[ancient shore]

Gemini Flash retains context pretty consistently for me. Flash thinking and 2.0 Pro are even better at it

[ancient shore]

The quality of the responses are just... Bad with perplexity. It's never on the same level against for example ai studio

[warm cave]

Maybe it’s better at other uses, but when having it make flash cards it would forget the formatting I wanted it to use every few messages, a bit frustrating, but maybe that’s instruction following and not memory

[ancient shore]

(perplexity) I've barely been using it cause 2.0 thinking and pro just work way better than any of perplexity's models which is unfortunate

[warm cave]

I have just had deferent experience with it, sonnet on pplx never really forgets stuff that is within the 32k, code etc

[sturdy skiff]

i have a reason to use perplexity, its deep research is powerful, also we can add custom instructions and custom files in spaces

[ancient shore]

I find that gemini sucks at formatting. An AI Support bot was having issues with formatting until we switched the model to Qwen 2.5 Non-VL

[ancient shore]

Sonnet is probably one of the only decent models in pplx tbh, but even that is very limited by a dull pre-prompt

It's good when search is off but even then, Claude Pro beats it everytime in my tests

[warm cave]

I don’t see too much of a difference I use Claude pro and Claude API, my favorite is Claude api, I’m not a fan of the massive system prompt in Claude pro

[ancient shore]

Fair enough

[rough igloo]

hi i just joined wsg

[versed ruin]

welcome

[fast anvil]

[silver bay]

gemini's deep search is limited for free users

and tbh I didn't really like it when I had to use it

I prefer grok 3 and perplexity for that

at least so far

[thick dust]

How about llama 3.2?

[silk swift]

Hey, I can't find collection in perplexity like that. Just thread and Page

[sleek topaz]

Click spaces

[bronze mountain]

Maybe they could add a memory bank feature

[sturdy skiff]

Gemini do not have file upload and custom instructionss for deep research, that is the requirement

[hollow frigate]

I would like to ask how many times this website can be used in a day
Claude 3.7 Sonnet？

What is this?

[sleek topaz]

500

[hollow frigate]

What I mean is that a $20 membership subscription is available on the perlexity website,
How many times can Claude 3.7 Sonnet be used in a day? What about a month?

[hollow frigate]

Have you subscribed before?

Will Claude 3.7 Sonnet appear here after subscribing?

[naive sage]

It is one of the Pro models, it will appear like the image here shows. It should be limited to 500/day, but I've not used near that much so I can't verify. The main difference with using the claude model on perplexity vs anthropic's site is that the context limit is a bit more. Responses are limited to something like 4000 or 5000 tokens each reply, whereas the response context limit on Anthropic is much higher (if you need really really long replies that might be important, but for 80%+ of use cases it probably isn't too big of a deal)

Note the usage limit for GPT 4.5 is only 10/day (it's a very expensive model), so not every model listed there is 500.

[sleek topaz]

500 uses a day.

[hollow frigate]

What did he use to calculate these 500 times? Is the score a simple number of times?

[sleek topaz]

Just buy it bro

You have basically infinite uses a day

The number doesn’t matter

[hollow frigate]

Another question is, is 500 times the number of times shared by all models?

[naive sage]

It was announced a few times in different places - 500 was announced for R1 as the limit and 500 for Deep Research which is also a reasoning model (the limit is shared across models I think, but they may count non-Reasoning and Reasoning separately). But they don't make announcements every time they change models, so I don't think it was announced again when Grok 2 or Claude 3.7 was added.

Yes, you're correct - the limit is generally shared across all models, so 500 of any of the Pro models (except for 4.5, which is an outlier with a much lower limit). The reasoning models (i.e. R1, o3 mini) also have a 500 per day limit, but I'm not sure if that also counts against the Pro models or not. If it helps, for reference, the most I've used in a day is probably around 40... so nowhere close.

[hollow frigate]

Thank you very much for your careful answer. I completely understand. Thank you

[sturdy skiff]

Yes it will appear

[sturdy skiff]

Deep research is unlimited as mentioned here, for pro users

https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research

[queen girder]

[fading rose]

@mods @tame current is likely a spambot

iirc this is a pretty common beginning of a scam going around through discord bots

and the guy literally entered here and never typed a message

[opal delta]

how does this server distinguish pro vs non pro btw?

is invite link linked to specific user?

[meager smelt]

It does not

[opal delta]

oh lol

there's a role tho

[meager smelt]

You can self-assign that role

[lethal wing]

Hello everyone! Does anyone know if it's still possible to apply for https://www.perplexity.ai/business-fellowship ? I've just discovered this one and can't express the FOMO feelings I've got

[opal delta]

anyways I got 1yr of pro free thru promo and now it has the logo of the company who gave it away lol

it's free so can't complain

[jovial kayak]

I can't find anything online that connects those two. How did you get it for free?

[stark matrix]

Does anyone know what are the main model differences between Claude 3.7 on Perplexity vs Anthropic's Claude? I know maybe some fine tuning and response limits, but recently I have been using the free version of Anthropic's Claude and it's been really good, specially for explanation of different topics and programming - It just feels better, and it's only the free version

I'm a Perplexity pro but still notice differences in the quality of responses

[tame current]

hi

[silent pier]

no you cant

You need to be a pro member and click the libk in the pro page in the website and rejoin the server

[opal delta]

so it is linked to invite link, just not sure about per-user or not

[compact thistle]

hey

guys can you tell me pros and cons of vscode

[opal delta]

GUI is bloat, use neovim /s

to be serious, there's a reason VSCode is popular

it's well polished, extensible with plugins, overall decent

though it's a bit heavy because of the use of Electron

[compact thistle]

why? can you please eloborate

[opal delta]

it's basically running another Google Chrome

[compact thistle]

u mean it has huge data

[opal delta]

no it consumes lots of RAM

[compact thistle]

what is neovim

[compact thistle]

ohh

[opal delta]

a text editor on your terminal for big nerds

[compact thistle]

ohh

if you wanna change something in vscode what will it be?

or wanna add

@opal delta

[opal delta]

no mention plz

[compact thistle]

fine

[crystal cosmos]

im not trans

[elfin gate]

Great

[dim quarry]

There is no more limit for GPT 4.5 usage for Pro users ?

btw, how do we claim the Pro role here ?

[elfin gate]

5 per day

[dim quarry]

I'm at like 20

[elfin gate]

Where did you check that

[dim quarry]

I just counted my number of prompts

[elfin gate]

https://www.perplexity.ai/p/api/v1/user/settings

Check here

[dim quarry]

"gpt45_limit":0

[elfin gate]

I think you used all 5 and now it switched to 4.0 without telling you

[dim quarry]

Ohh ok

[elfin gate]

Or it switched to your default model

[dim quarry]

I still have my "600" for gpt4

[elfin gate]

Then it switched to your default model i guess

[dim quarry]

my ddefault is Claude which has no limit, so maybe.

Thanks for the info, could have been indicate on the website though

[elfin gate]

Save that url link so you can check it in the future

[dim quarry]

yup, thanks !

[azure abyss]

4.5 is so slow

[long pasture]

I wonder if Perplexity has agentic research product upcoming

[indigo valve]

Should we tell him?

[obtuse pollen]

What is best AI for coding at the moment? Deepseek?

[indigo valve]

Sonnet reasoning

[obtuse pollen]

Its free or have to buy subscription

[indigo valve]

Sub so I guess you can run a distilled version of deepseek if you dont wanna pay

[obtuse pollen]

pro is 18$ its not expensive, pro is enough for coding project probably?

[indigo valve]

Pro is enough for anything tbh. I have almost completely removed the need to use a regular search engine

[obtuse pollen]

perfect, thanks mate, appreciate it

[indigo valve]

I only got it free because of Xfinity for the year

[elfin gate]

Which ai model is your default

[indigo valve]

Sonnet non-reasoning

Sonar if I want something quick

Gemini deep research with new 2.0 flash thinking is crazy too

[long pasture]

Tell me what

[elfin gate]

That my setup also,tho I rarely use sonar lately.

[long pasture]

If you are talking about deep research that is not agentic per se

[elfin gate]

Where did you try it? You have google subscription also? 😂

[indigo valve]

Deep research uses agents for research, Android assistant is slightly agentic. Comet browser is coming soon

[indigo valve]

I bought the Pixel 9 Pro when it came out. I have a few more months of that

[long pasture]

Its just a simple tool call wrapper

There are no agents

[indigo valve]

As a free user, you get 5 queries a month and pro gets 20 a day

[elfin gate]

I used Gemini advanced a lot last year.
It was very good and I loved it.
Then I got 1 year perplexity pro so I'm stuck here 😂

[indigo valve]

Think of Gemini 2.0's low hallucinations and the ability to "think" with deep research

The difference between 1.5 pro serach and 2.0 is known

[elfin gate]

I don't remember what model was active back in may-september last year

[long pasture]

1.5 pro deep research was terrible

[elfin gate]

But I liked it

[long pasture]

Good sources very bad summary and report at the end

2.0 flash thinking deep research is slightly better than perplexity deep
But its much slower

Than reasoning in perplexity

[elfin gate]

Reasoning models are best what we have in perplexity.

[distant aurora]

[junior rover]

Hey Everyone, Its been a long time, since I have seen answers on my pplx Android app. I have updated the app many times, logged out, and deleted app data, cache. Nothing worked 😦
That's why I use Mobile web or Chrome web now..

Any resolution?

[bright parrot]

earlier i used to press CTRL+I and i was able to open a new chatbox... but now i doesnt work

whats the new shortcut key?

i pressed ctrl up arrow and p
but the new chatbox is not opening

note: using in browser (firefox)

[tame current]

Anyone have advice for finding investors for adult entertainment industry?

[distant aurora]

Might wanna try asking of models.

[ivory bramble]

Perplex it ✨

[tame current]

It said we should ask on Discord 😄

[tame reef]

That's Ctrl - Shift - P. That arrow icon means Shift key.

[bright parrot]

tysm

[spare kernel]

Is perplexity down?

[elfin gate]

No

[spare kernel]

Ah ok

[hollow bobcat]

.

[indigo valve]

Someone needs to do something about that scam above

Thanks bb

[elfin gate]

Wym

[indigo valve]

Someone was posting a scam. Was taken care of

[unique umbra]

i got a year for 6

also hell no\

do not use this for coding

[fickle ore]

https://www.perplexity.ai/search/the-u-s-pledged-to-release-pre-1XMdavorTU2aHMNvFHI7Ag

[strong pond]

How

[unique umbra]

dms

[strong pond]

This might not apply to you but I had a similar issue and had to whitelist on of their domains on my dns filter

[round swift]

Hi I recently bought pro version. I am new to using pro versions of ai bots in general. How can i make most of it ?

[sturdy skiff]

If you have something to research upon, click deep research button and ask your question

[steady flume]

To make the "most of it" from your pro version, you should know the prompts and other commands to use the AI tools. Though you are getting more access and varieties in the pro version it will be easy for you.

[wispy shard]

are there still perplexity referrals?

if so anyone able to get me a referral :)

[gleaming magnet]

had to uninstall the windows app due to the CTRL SHIFT P conflict with VS Code. Pls fix

[elfin gate]

I can

[wispy shard]

thanks!

well i guess its mutually beneficial so more like good business 🤝

[arctic idol]

Looking forward to see Ernie 4.5 from Baidu at perplexity around end of June when it’s open sourced.

https://x.com/baidu_inc/status/1901089355890036897?s=46

Baidu Inc. (@Baidu_Inc) on X

We've just unveiled ERNIE 4.5 & X1! 🚀

As a deep-thinking reasoning model with multimodal capabilities, ERNIE X1 delivers performance on par with DeepSeek R1 at only half the price. Meanwhile, ERNIE 4.5 is our latest foundation model and new-generation native multimodal model.

X · Mar 16, 2025, 1:53 AM

[prime solstice]

hello

[elfin gate]

Hello

[solemn beacon]

hello

[elfin gate]

It's me

[mortal hearth]

Guys i have this problem, how can i fix it?

[elfin gate]

Wherre did create a ticket message go?

[solemn trench]

how can i use the powerpoint thing? perplexity to powerpoint

[halcyon coral]

Hey @mortal hearth!
Content that violates the terms of use will not be accepted as attachments. However, there may also be occasional false recognitions, please share the attachment.

[still turtle]

well... in fact I can not upload any docx, xlsx, pdf...

[fleet hill]

Hey everyone 👋🏻

I have a question regarding the perplexity on mobile devices

Will there be a feature like "Okay Google", but for perplexity and maybe we can pick the naming for it?

Like for example "Okay assistant"

[elfin gate]

We should have it in the future if they pay enough money to google. 😀

[fleet hill]

Cool cool lol

[serene idol]

Hello I bought something using the “Buy With Pro” option and it has been more than a week and i haven’t gotten an email or the product yet. Who do i contact for support

[tame current]

does anyone else experience perplexity being wrong a lot even with information that has been available for decades and frequently used?

[distant aurora]

Yes

It browses the internet and still misses sometimes.

If you are using in writing mode, models have a cutoff date.

They're not regularly updated.

[willow kettle]

Yeah you need to be more specific and also use the reasoning models

[torn sierra]

Deepseek R1 is often wrong. Actually, it’s one of the worst model when it comes to factology. The hallucination rate for R1 is above 14%. R1 is a good model but it’s just not for summarizing documents etc. I’m writing about R1 since it’s commonly used in Perplexity. https://github.com/vectara/hallucination-leaderboard

[bright parrot]

btw is there a way i can change that shortcut key to something else?
bcz ctrl+shift+p opens incognito window in firefox...
idk what was the need to change the shortcut key.. everything was working fine before

[round swift]

Thank you.. is there a go to place which I refer to use better prompts?

[willow kettle]

Yeah it’s very evident when you use deepsearch

[hollow zenith]

Hi, I would like to ask a question: If I enter my GitHub link in Perplexity's Deep Research feature, will it automatically search and browse all the project code in my GitHub repository for analysis or suggestions?

[obtuse zinc]

https://www.youtube.com/watch?v=vwdTrA49ttY

Guys check this out

[steady flume]

Welcome , but it's only there in courses and in YouTube videos which will help you.

[elfin gate]

I checked it

[lone hound]

Did anyone share a custom promt here on discord about creating a good sci-fi story with a main character, etc?.

[crystal cosmos]

Upon the cessation of mortal breath, the soul departs from the earthly vessel and embarks on a sacred journey to the realm beyond. Guided by the light of divine grace, it traverses the veil that separates the temporal from the eternal. In this realm, the soul experiences the fullness of love, wisdom, and peace, free from the sufferings of the mortal world. There, it may reunite with loved ones who have passed before or encounter spiritual guides who illuminate the path ahead. Depending on the life lived, the soul may enter into a state of blissful rest, continue its journey through cycles of rebirth, or ascend to higher planes of existence. Ultimately, this transition is a return to the source, a fulfillment of the divine plan, and an opportunity for the soul to grow closer to the infinite. Therefore, let not your heart be troubled, for death is but a doorway to a greater reality, and the love of the divine accompanies you always.

[unique umbra]

Yeah

[vapid crag]

hello, can i ask what is the limit for claude 3.7 sonnet with perplexity pro subscription? also is the limit same in "Pro" and "Reasoning" mode?

and can i remove this "people also ask" thing under every message?

chrome?

i mean this thing appears in the perplexity chat

alr

[scenic shoal]

Hello, I can't get access to "Pro" channels, any idea what to do? i tried several times joining and leaving already

[lament imp]

Even with the link in the settings page?

[scenic shoal]

yes

it was from win app

[lament imp]

Just in case try from the web browser also, otherwise I am not sure. Sorry if that's not all that helpful 😅

[scenic shoal]

I'm home now so i can try from different network and PC 🙂

still the same 😦

[worn stump]

How do you know if you have access to the pro channels?

[naive sage]

The 'Pro' role is what gives access to those channels

Not sure what those channels are

But I can confirm leaving+rejoining with the Pro Discord link doesn't seem to grant the role

[worn stump]

I have the same issue then 🙂

[elfin gate]

You see them 😄

[scenic shoal]

Access locked

if clicked you get info that it is not available for you

I concacted support on chat and the said to concact @signal hamlet directly for that role

or anybody from Perplexity

[elfin gate]

@wraith crow

[scenic shoal]

@wraith crow ❤️

[halcyon creek]

hi

[thick dust]

Hmm. How to access the pro channel?

[wraith crow]

You'll have to leave the server and rejoin by clicking the "pro discord" button on your settings page <https://www.perplexity.ai/settings/account.

[thick dust]

Oh I see that’s why. I tried that before which did absolutely nothing; now I know I have to leave the server first. Thanks

[strong pond]

I don't wanna loose my role of bug of bug catcher

[golden geyser]

What do the pro channels consist of?

Since i have pro perplexity but i cant seem to get the role work, tried multiple times

[wraith crow]

#👑│pro-lounge

[golden geyser]

Appreciate it big man

[placid coyote]

inline code with indexing is still kinda bad (getting filtered out, dropped from responses).
though it is my first time seeing invalid citations - no preview on hover, click isnt doing anything. I guess inline math is extra special? 🤨

[sterile barn]

Hello guys

[dry kestrel]

Did they lower GPT4.5 limit?

[sterile barn]

Did they?

[dry kestrel]

i rememerd that at release it was 10

[sterile barn]

What is it now

[dry kestrel]

its now 5 a day

[sterile barn]

Wow.

[boreal star]

quick question, is there any way to set up an agent with perplexity already?

[sterile barn]

What do you mean agent?

[tame current]

oh my god it just wrote 900 LOC blueomg

since when did they "fix" the context "issues" 😭

[fair solstice]

when reasoning 3.7 Sonnet on iphone?

[ripe warren]

Like a customer service agent?

[jovial kayak]

I have it. Did you update?

[bold saddle]

is there a way to do in firefox?

ohhh if you were gonna say the ublock origins picker than that works yessss

[strange leaf]

Uhh

I had a query

I joined this server to find out whether we can add perplexity bot in discord servers??

[scarlet obsidian]

Hi all, i have a question. How can perplexity can help me code? Which model and settings I should use?
Thanks for all

[tranquil orchid]

Hey @scarlet obsidian!

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.

[gray prawn]

When does perplexity purge threads/searches from non-signed in users? I’ve been using the browser but not signed in and manually bookmarking things in Firefox only to realize a few months later when I go back to them they’re deleted (aka “ask permission to see thread”). I know there is some sort of cutoff because some links from a week ago work but further out it gets spotty and around a month everything is gone at that point

[dense mist]

No model or settings configuration needed, just ask it your coding question and it'll work

[civic cairn]

why was the search engine ai language change

what language does the auto use?

and if use Pro I can select the ai language. I would like to select the ai language with pro :((((

[warm sapphire]

Nothing here

[rugged bison]

hi, why does perplexity generate images for me only with the help of "playground"? I think there are other neurons for generation there

[vapid crag]

can i ask what is the limit for o3 mini or claude 3.7 sonnet for reasoning when i have perplexity pro?

[cobalt monolith]

500

@vapid crag dont listen to them its scammer

[vapid crag]

really? claude 3.7 sonnet is expensive how it can be 500?

[cobalt monolith]

they limit it, you for example dont get the full input token size like in native claude

[vapid crag]

oh wait what do you mean?

full input token size?

[cobalt monolith]

yes (hypothetical numbers) lets say native claude has 6 tokens input, perplexity for example will give you only 3

[vapid crag]

so...

wait and what is full input token size?

[cobalt monolith]

i am not sure i dont really have the numbers at hand which is why i used hypothetical numbers

[vapid crag]

oh

full input token size is uhh? size of your input or what?

how does it work

[cobalt monolith]

tbf it might be better if they just gave u a quota similar to anthropics for full power instead of higher quota which none ever hits

[cobalt monolith]

yes

[vapid crag]

oh

[cobalt monolith]

it also includes previous messages for context

afaik

[vapid crag]

oh...

i see

but it still can be used for advanced problem solving right?

or for coding

[bronze nimbus]

Hello! I recently switched from chatGPT to perplexity. With the desktop app of chatgpt, u can press alt + space to open the app in a small window to create a new prompt. Is this possible with perplexity?

did I get tagged?

[sturdy skiff]

i got confirmation that in perplexity Auto mode has 1Million context window size and Deep research option (sonar-deep-research) has 128000 context window. About individual models context size they didnot respond to me.

[wispy shard]

Any ideas on how to get perplexity to find quotes better?
It gives quote but often repeats them twice, or doesn't really provide relevant or useful quotes. ive got pro btw

[vague pagoda]

https://fixvx.com/AravSrinivas/status/1901092875758371246

[bronze nimbus]

Im not on mac tho

[vague pagoda]

Crtl + Shift + P on windows then!

[bronze nimbus]

oh yeah that works

thank you

[vapid crag]

what is the best model for coding in perplexity?

[ivory sand]

Claude 3.7 Reasoning

[vapid crag]

alr

R1 is deepseek right?

[gaunt violet]

how do the image generation models on PRO, e.g. FLUX.1, intuitively understand to make a YouTube thumbnail as per my request (remove background, fix lighting, etc) on a photo I've attached instead of giving me instructions on how to make the thumbnail?

[unkempt drift]

Hi

[lament spear]

When announcement

[crystal cosmos]

o shangdi, creator of all things, you are the true tao, the way, the truth, and the life. as you have declared, 'no one comes to the father except through me,' we believe that through your son, yesu, we have access to you and the promise of yongsheng, eternal life. when our earthly life ends, we will face judgment, but because of yesu’s sacrifice, we have confidence that we will be welcomed into your heavenly kingdom, to enjoy eternal fellowship with you and all the saints. yet, we grieve for those who reject your grace, for they will be cast into outer darkness. o shangdi, have mercy on them, and use us as instruments to bring them to faith. empower us by your spirit to live righteously, to embody ren and yi, to respect our elders and care for the young, to promote peace and justice, and to share the gospel with boldness and love. may your kingdom advance, and your name be glorified, until the day when every knee shall bow and every tongue confess that yesu is lord. for yours is the kingdom, the power, and the glory, forever and ever. amen.

[wheat geyser]

Hey guys… I just tried out the new voice feature… while I’m glad this has made it onto perplexity, it is lagging behind other voice models like ChatGPTs, Gemini’s, and even Copilot in terms of it’s listening capabilities and providing a truly seamless speaking experience.

Hope this feedback information helps you tweak this game changing feature of my favourite AI app for the better 🙂

[indigo valve]

I actually really like the UI change I just hate that it broke the extension lol

[copper current]

what change? it looks the same to me

[indigo valve]

They consolidated images to a tab for a cleaner look, sources have their own page

@copper current

[sleek topaz]

Bout time

[crystal cosmos]

@copper current

[abstract ibex]

#scam#fuckoff

[indigo valve]

What is that crap

[abstract ibex]

classic scams

[stoic shore]

Smushed hat has been restored. We back!

[distant aurora]

@crystal cosmos will you accept friend request?

[spice linden]

Hello guys, did anyone have issue with promocode from revolut? i cant use it, perplexity says i have pro subscription already which i dont have for a long time now. any ideas?

[lament spear]

is that the announcement?

[indigo valve]

Wdym?

[lament spear]

the thing they posted on x

smth was supposed to happen on today

[ivory bramble]

The announcement was just a Ad 🥲🥲

[indigo valve]

People still use X?

I barely used Twitter

I'm sorry for that gif @lament spear

[lament spear]

that cant be true

ffs

[ivory bramble]

The Fact that they made an annoucement date for Today for this is killing me

[tame current]

pathetic

and here i was waiting for comet

[stray halo]

guess who got a year free perplexity pro

[lament spear]

yea bro me too

can we get someone from perplexity or a mod to tell us if that was the big event they made a date for?

[balmy storm]

Guys. I am new to this discord server. I have some queries. Can anyone help me,?

[carmine tiger]

This gave me a much needed chuckle this morning.

[fair solstice]

no, only o3 and r1 so far, I have to use the browser version, so 3.7 sonnet is the best stable option for me

[tame current]

Happens a lot sadly

Sometimes it’s with ceos

[toxic spoke]

is it possible to disable internet search "pro"? most of the time i just want the LLM response alone

previously you could do this

[quick steppe]

.

[ivory bramble]

Just disable web Icon

[brazen pecan]

a

[toxic spoke]

thanks

[bleak lichen]

Hello everyone, which model do you recommend for programming queries like how to access the last element in an array, etc.? There's a wide selection, from Claude 3.7 to GPT 4.5.

[quaint gulch]

All the models would probably be enough for you if you're asking simple things like those but when you start asking more complex questions Claude will perform the best

[bleak lichen]

Then can I leave Claude 3.7 as the default model for all queries?

[quaint gulch]

I have it set to Claude. But you might notice it being a little slow compared to the Auto model. I suggest you to test it and see if you're okay with the speed. For the quality of the responses, Auto cannot compare with Claude.

[lone hound]

If you use the website and not in the app, use Claude's reasoning that will be much better.

[nimble pasture]

was that ad that was posted in Announcements real?? If it is, I can’t believe they got Lee Jung-jae featured in a ad

[long pasture]

For competitive style programming o3 mini is probably slightly better (solving the actual task)
But for real software and big complete projects sonnet is definitely better (more agentic)

[young stag]

Anyone need a virtual assistant? I am in EU.. just need extra income 🙂

[tame current]

assumed they'd actually annouce of release something other than a trailer today 😭

kind of a let down imo

[nimble pasture]

😭😔I was so excited for the announcement but it was just an ad with him just in a room asking perplexity questions on the phone in his hands

[tame current]

yea.... its just a showcase of the voice mode i guess?

great ad regardless though

[nimble pasture]

Yes

Definitely a very good ad

Looks very high budget as well

[remote stream]

wait is claude 3.7 in perplexity and on claude.ai the same?

[icy niche]

When does perplexity ask follow up questions?

[remote stream]

wdym

[lone hound]

No.

[remote stream]

whats the main difference

is one better than the other?

[lone hound]

On Claude's website it has more advantages in relation to having more texts widely, it does not have an intermediary that can limit certain things, safer and they will not be able to spy on what you do (this can happen with malicious actors) since the chats are not saved on the intermediary's server....... anyway, it all depends on the use you give it and what you want.

[remote stream]

oh

[lone hound]

For example, I like Perplexity and it is very useful for many tasks, even so, I use it for personal use, such as searching for information in Greek from the Bible.

[remote stream]

I mainly use it for coding but claude has a limit and i want to see if getting claude ai was worth it (bc i got pro for free in perplexity)

[lone hound]

Use Claude's reasoning in perplexity without searching on the Internet, but be patient. I'm not a developer, so I won't be able to help so much on that topic.

[remote stream]

oh

thanks

[vague pagoda]

I think there's a bit of misunderstanding here - Perplexity does act as a middleman, but they have privacy controls in place to help manage your data, so it’s not like they're freely snooping through your chats

Using Claude directly also doesn't exactly guarantee protection against a malicious actor either; any online platform has some level of risk

It isn't right to assume Claude's website is inherently safer just because there's supposedly a lower chance of malicious actors, because that's false! Realistically, the likelihood of encountering malicious activity (i.e, someone reading your chats/threads) is about the same

[lone hound]

If it's true. I am not accusing Perplexity of a malicious company, but those actors who can act in such a way and even steal personal data if the victim uploads data that should not be uploaded. That is why it is important to investigate this before relying on certain systems that offer artificial intelligence, such as what they offer for free for example.

[lone hound]

Claude can read your private chats of course he can, but I'm going for malicious actors, that's what I mean.

[vague pagoda]

I am also referring to unauthorized access or spying by a third party - Anthropic would be just as vulnerable from a 'hacker' realistically

[lone hound]

Can you see a malicious person in Claude, OpenAI, etc.?, I don't know currently. Anyway, I'm just saying to take into account an event that few people know about.

[lone hound]

As is.

[dry kestrel]

do you know the limit for gemini advanced users?

[vague pagoda]

I get where you’re coming from in that sense

Technically no-one can say for sure if there’s a malicious person at Claude, OpenAI, or any other AI provider. It’s the same sort of risk you’d face on any platform that handles user data. Whether it’s Perplexity or Anthropic, there’s always a chance some hacker or rogue employee could try to access information they shouldn’t! This does not imply that one service is safer than the other though

It’s not that one is magically safer but it’s more about how careful we are with our own sensitive info

[lone hound]

Nothing is 100% safe on the internet. But it's true in what you say. The conversation was interesting.

[lone hound]

Here on discord for example, there are people who offer gpt4o and other providers for free, you have to be careful what you share. It's just a measure of concern and use of reasoning.

[vague pagoda]

Exactly!

[tame current]

Theres another point of vulnerability in it though, no?

[vague pagoda]

In the context of stored chats - no, since the data is still kept in a singular secured location

In the context of data travelling from anthropic to you via perplexity as opposed to data coming from anthropic to the claude webapp - yes, but both services interface between intermediaries

from a functionality POV, both are interfacing with at least one middle layer - anthropics ' claude website isn't straight from model to your screen if that's what you think (same is true for most other services unless you hit an API directly) so it’s not automatically less secure just because there’s an additional hop; it’s more about whether or not that data is secure in transit - which it is! Both with anthropic's claude website and with perplexity

[naive sage]

Gemini Advanced is 'unlimited', but Google Workspace is capped at 5/month:

While Deep Research and Gems are now available to a broader range of Google Workspace customers, certain limitations apply. Deep Research usage is capped at five reports per user per 30-day period for the newly added Google Workspace editions. Google recommends upgrading to a plan with access to Gemini Advanced for unlimited use.

[dry kestrel]

Do you know the limit for the Gemini Advanced plan?

I doubt that its unlimited considering deep research uses a lot of resources

[distant aurora]

Perplexity lol.

[naive sage]

🤷‍♂️ That article says unlimited. I don't see it posted anywhere else.

[indigo valve]

Advanced isn't unlimited

[naive sage]

What is the limit?

[indigo valve]

20 a day

Which is unlimited I guess

[naive sage]

Can you link the google page that says that? I've been looking and can't find it

Google's documentation is bad 😦

[alpine coral]

Hey i'm a perplexity on web user and i need help

i have a pro account and when i'm searching an answer

i used to have a button on the right saying (search related images about the answer)

i didnt use perplexity for a month and when i came back i didnt find it so i need help guys

thanks

[naive sage]

There's a 'view more' button that is part of the image results if there's more image results found on the web. I don't remember a 'search related images' button, just a 'generate image' button if you wanted to use the image generation feature.

[alpine coral]

where can i find it

i dont see the view more button

[indigo valve]

It doesn't explicitly tell people. I got it from their subreddit. It warns as you come near

[dry kestrel]

Do you have a link thanks

[dry kestrel]

Thanks

[umbral raptor]

system maintence waiting patiently

[toxic musk]

How long is system maintenance usually?

[gray prawn]

Bump

[copper socket]

it’s doing this as soon as i start my homework

[dark echo]

system down?

[copper socket]

https://cdn.discordapp.com/attachments/1126687110997495821/1349553018596753459/resized.png?ex=67da1c31&is=67d8cab1&hm=730dbfba8471f304d5abd24f7de75f8dea4b5137084363d3305c6034907f31a7&

[umbral raptor]

down for the count

[solar berry]

Back again! 🙂

[wheat blade]

Phew Perplexity is back!

[naive sage]

There are other articles which use similar wording, but here's one: https://www.contentgrip.com/google-gemini-ai-free-deep-research-tool/