💬│general

[heady jay]

You care enough to sling insults

GFY

[tame current]

thanks

anyways as i was saying

shits expensive

not as expensive as leasing a car

[orchid shore]

How do I apply the 2 months free if I'm already subbed?

It says for returning subscribers as well...

Oh nvm returning to perplexity

Sucks it's not for current subs as well...

[tame current]

where the dream channel at

[wraith crow]

Hi @tame current, the discord bots have been discontinued #📢│announcements message

[tame current]

oh

thx

[formal kayak]

I need to work 4hours to get 20$

API access is free i believe?

[rose sparrow]

Claude 2 has a 100k context even in perplexity. Correct? What exactly is the gpt4 context length model used in perplexity?

[gilded zealot]

Same question here. On #📢│announcements it was mentioned,

Offer ends Dec 3 2023 11:59pm, new/returning subscribers only.
Not sure what returning means here, but if it means people who were subscribed earlier and then discontinued, then this whole discount seems to be for signups only, not existing subscribers. Slightly incongruous as a thank you for people who continue to support, if I may say so - no offense intended, totally their call. Still happy to stick around.

[rose sparrow]

I’m a pre-existing subscriber. It worked for me by applying the code at the change plan section where i went from monthly to annual. It may also work if you cancel at monthly plan. And then re-subscribe by putting code in.

[gilded zealot]

I thought about trying that - but given the volatility of the current environment, wasn't willing to put it money for an entire year, even with a discount. Cancelling and re-subscribing might be worth trying. But I am lazy and will wait for someone else to try that an report back first 😛

[rose sparrow]

Lol smart move. For me, The claude 2 integration was something I still value so I decided to commit.

Do you know what GPT4’s context length is in perplexity?

[gilded zealot]

tried out. didnt work. cancelling the subscription only lets you renew, and renewing doesnt let you enter a discount code. I can try letting the subscription expire, and try then, but by that time the discount code would no longer be valid.

[gilded zealot]

not sure, tbh. This reddit post suggests 32k.
https://www.reddit.com/r/perplexity_ai/comments/16yfmt1/gpt4_context_window_for_perplexity_pro/

[rose sparrow]

Thats wack it doesn’t work.

[rose sparrow]

Yeah I’ve seen a few mentions like that. I do wonder though what exactly it is. Like for example Claude is 100K its just that perplexity keeps forgetting that unless i keep referring back to certain instructions.

[sweet jasper]

With openAI in a possible doomsday scenario

Gpt4 might go kaputt in the following weeks

[rose sparrow]

The whole saga seems strange to me. Apparently the new CEO who has been hired is the ex-ceo of twitch?

[gilded zealot]

I doubt gpt4 infra will go down - way too many corporate contracts hanging in balance. But improvements may stall - GPT5 will probably be delayed.

[gilded zealot]

he was the third person apparently to get the offer, and he accepted. there was also an attempt to merge openai with Anthropic over the weekend.
long term, I am hearing that switching to using Azure API for GPT4 seems safer atm. Microsoft has rights to the IP, and the hardware to give the assurance people need.

[wraith crow]

Perplexity use both GPT-4 8k and 32k depending on the prompt length

[rose sparrow]

Thank you for that. Do you know of anyway to keep perplexity from forgetting the topic we were on every 4 or 5 questions? Lol. I was trying to learn about a social science framework, and every so often perplexity would provide a response that says “when planning a trip to JAPAN you should”. It was all about Japan lol

[chilly forge]

This has been reported quite a few times now

[wraith crow]

That is currently a bug for some people, the team is aware of this issue. Afaik it’s only happening sometimes when copilot is enabled. So you could leave it disabled for now if it occurs to often. And be sure to report the wrong answer at the bottom by clicking on the flag icon.

[chilly forge]

But its quite wierd that people are getting Japan responses for some reason like why only Japan why not something else 👀

[rose sparrow]

Bro I actually don’t understand copilot at all. For example even with it off it still searches the web and uses the claude2 as the model of my choice. So what exactly is copilot doing? And if i dont use it, anytime i ask a question it still counts against my number of copilot uses.

[wraith crow]

Here is a great explanation of what copilot is https://blog.perplexity.ai/faq/what-is-copilot

[chilly forge]

#📘│faq

[wraith crow]

And the 600 copilot limit is actually the usage limit for Claude/GPT-4/Copilot.

[rose sparrow]

This was helpful. I didn’t see it on my perplexity web app. By the way, 3 seconds into the video it talks about planning a trip to Kyoto, Japan. Somebody at perplexity really loves Japan

[chilly forge]

Did it started with this ?

[rose sparrow]

Its very similar to that. Each time i would get a slight variation but it centered around that theme. Your photo is a perfect example.

[chilly forge]

Ah seems like its kinda fallback for something (either the model hallucinating or something else) my reason for this guess is since its revolving around Japan for the similar theme. This is my best guess

[upbeat turret]

It's interesting that they're using GPT4 and Claude 2 in the Discovery tab. The main question is why they don't use their own model. The first reason I have in mind is that it's not good enough yet. Or maybe they just forgot)

[celest stone]

Perplexity's own model is almost boringly neutral

makes it great at retrieving info

but the writing isn't captivating

[minor spindle]

did perplexity ai using chatgpt

ummm

[rose sparrow]

so if we wanted to specifically focus on coding tasks that day. Would it be better to set the model in settings tab to be perplexity and also have copilot on?

[celest stone]

thats what I do

[raven socket]

This is the best A.I. ever. Thanks for creating this.

[drowsy crescent]

Hi, i am wondering the AI model between perplexity, gpt4, copilot and claude 2?

[past otter]

Hey folks. How do I claim for perplexity pro 2 months free? Can I buy monthly twice or do I have to buy the annual plan?

[vocal cipher]

Applicable to Monthly too.
But since the offer ends on Dec 3, it means you can only use it for one month (-$20.00)

[obsidian gazelle]

Hey guys, does anyone know how to build generative AI applications?

[drowsy crescent]

I think another 20 $ will be applied for another month. This offer ends dec 3 but discount won't be ended

[past otter]

So, the best case to get 2 months free is to subscribe for a year and get a discount for 2 months?

[vocal cipher]

Hopefully. Hopefully the free promotion code is still valid when we subscribe to the monthly plan next month.

[vocal cipher]

seems so 🤔

[drowsy crescent]

I hope so

[past otter]

If I subscribe yearly, when will I be billed? At the end of the year or at the beginning?

[chilly forge]

beginning ofc

[primal bear]

Android app doesn't support file upload?

[drifting igloo]

What happened with the imagine channel?

Not visiblw for me anymore

[tame current]

bots dead

read annoucements

openAI won't just die from a CEO mishap

i really don't understand the nervousness

it ain't that deep lmfao

[sweet jasper]

My homies are already discussing mergers with anthropic

Its that bad

And its been like 4 days

But as sandipb said, too many business contracts and everything being powered on azure to begin with, so little to no downtime, AGI just pushed forward for a longer time perhaps

[tame current]

so Anthropic is the next openai

hilarious

[wooden portal]

Anyone else getting an error on log in?

[tame current]

nah

[formal kayak]

Disable adblocker

[wooden portal]

thanks

[tame current]

did that work 😭

that's horrifying lmfao

[tame current]

WHAT

😭

oh that's actually insane

[wooden portal]

i have my ad blocker to set trackers too

[tame current]

ahhh that's why

[wooden portal]

that might have been the problem

[tame current]

yeah that could've

no clue how to set it to block trackers for mine

[wooden portal]

what adblcker do you have

[tame current]

adguard

safari has this

[wooden portal]

get ublock origin

[tame current]

not avail

on safari

[wooden portal]

this is not enough

[tame current]

i know

[wooden portal]

check if duckduckgo privacy extension is available for safari

[tame current]

hmmm

it's fine honestly

i'm not that privacy sensitive

[formal kayak]

What you think about bot dead lol

[tame current]

never used discord that much

i'm fine with IRC at the end of the day

[formal kayak]

IRC has bot?

[tame current]

bots aren't important to me, unless they're utilized for moderation

[tame current]

IRC is internet relay chat or something

no bots

just messaging

[formal kayak]

It's discontinued due to low usage on discord... I love those bots and also help me to read old messages 🥲

[tame current]

fair fair

unfortunate but at the end of the day

we can't change much about it

[wooden portal]

i have news for you

[formal kayak]

Planned to create a bot... But API still in beta version and always im seeing unrelated answers on labs.perplexity.ai

[formal kayak]

@wooden portal news?

[wooden portal]

if the messages are gone they are gone

[tame current]

OH SHIT

i just finally understand what that meant

i thought he meant discord chat scraper 😭

[wooden portal]

⁉️

[tame current]

i thought he used discord bots to archive discord chat logs

damn, sorry for your loss bro

[formal kayak]

Huh ....? Chat gone what you are talkin abt 🙄

[wooden portal]

nothing

i am pulling your leg

[formal kayak]

🥲

I'm going to protest with the Perplexity CEO to bring back the Discord Bot

[tame current]

good luck

if you do then good job bro

[wooden portal]

https://www.perplexity.ai/search/7d424b15-8fc0-410f-b770-e4d1005757ac?s=c#7d424b15-8fc0-410f-b770-e4d1005757ac

qhar

[zenith atlas]

API will be paid per output and input tokens, they already published estimated cost IIRC, and there will be some amount of free usage to pro users

[tame current]

how much free?

i hope my ass doesn't go above it

[formal kayak]

As of now free for pro users right!?

[fathom ocean]

I have updated to Pro, I tried to leave and re-join via the Pro link but still can't access the Pro channels. Does anybody have any clue?

[formal kayak]

Pro channels are available?

[fathom ocean]

yes

[narrow raptor]

Please how do I apply the code to my account on the app?
Thanks.

[halcyon coral]

[formal kayak]

What's the difference between pro channel and public channel

[tame current]

nothing

if you have pro you can talk there

[halcyon coral]

You have more contact with the Perplexity team and usually get a faster response to a question or request.
There is also an event in the pro lounge from time to time.

[celest stone]

saw talk about adblock stuff stopping login, meanwhile on Firefox with uBlock+Disconnect+DDG no issues

the Safari blocker is really strong

[tame current]

nono

hes using ublock which fails the login

safari is fine

[celest stone]

Oh damn, RIP embeddings API...

[warped sequoia]

hey all, just got pro, im curious is there any reason to choose gpt-4 from this list?

i thought perplexity chooses best model for you based of query

[halcyon coral]

No, you have to choose your own model, Perplexity is a fine-tuned GPT-3.5 model, GPT-4 is the "smartest" and Claude has the most context (100k) and is often used for creative writing.

[warped sequoia]

i think gpt-4 has massive context now too in chatgpt i thought

or thats only via api

like i thought I can give a big pdf to chatgpt now but didnt try yet

so my thinking is that I will use perplexity for more real time needed info or things i need link references to

and gpt-4 via chatgpt for more complex queries i guess

[halcyon coral]

GPT-4's context on Perplexity is currently 32k

[halcyon coral]

GPT-4 on Perplexity should be the same 😉

[warped sequoia]

yea but its annoying to switch models

what i have now

is chatgpt in app via safari > add to dock

then perplexity in app too

bound hotkeys for each

just trying to understand when best to go to which app

also i dont fully get copilot

is it like agents?

[halcyon coral]

https://blog.perplexity.ai/faq/what-is-copilot

[warped sequoia]

curious how you use it

i see their examples but would be cool to hear more real cases

[halcyon coral]

Copilot can sometimes help with more complicated AI profiles, I almost always use Copilot except for two use cases, as Copilot often gives you interesting ideas with the queries.
Another plus point is that more sources are used

Fixed

[rain jacinth]

Thnx

[boreal delta]

I can't use this server to use perplexity anymore? Where is the channel to ask questions to the ai?

[halcyon coral]

#📢│announcements message

[halcyon coral]

👋

[wooden portal]

hello

[hearty quartz]

I'd like to try the 7-day, 1-month renewal option. What should I do?
I'm afraid I'll be charged $200 after 7 days.

[halcyon coral]

Your payment method will not be charged if you cancel within the deadline 🙂

Oh wait, monthly?

The trial is only available for the annual renewal

[dire stump]

Is claude 2 unlimited in pro or has limits?

per day?

[wraith crow]

600 uses a day this is shared with GPT-4/Claude/Copilot

[dire stump]

Is Phind model unlimited for free plan?

[halcyon coral]

🙂

[chrome jolt]

Thx!! 🤝

[halcyon coral]

This is Perplexity, not Phind 😉

[dire stump]

Sorry, I mean is perplexity model unlimited for free plan?

[halcyon coral]

Yeah, there is no limit.

[lethal onyx]

Guys, I need some help. So i just subscribed to Perplexity Pro using the Promo Code. but now there is a menu in the setting which i can choose what model to use in perplexity ai. does it only apply for the copilot? or it also apply for the quick search?

[halcyon coral]

Applies to all searches 😉

[lethal onyx]

Oh wow..

What is the difference between Perplexity model and GPT4 model tho?

[zenith atlas]

Perplexity is gpt 3.5 fine tuned

[lethal onyx]

Oh i seee thank you guys for your reply @halcyon coral @zenith atlas

[minor spindle]

thanks is better bing chat

bing chat just can asking for 30 question

[worn monolith]

Is something wrong with the Claude option?

https://www.perplexity.ai/search/Explain-to-me-X.OBesgCQUyxiFG8V1ym.Q?s=c

it doesn't give any answer at all...

[halcyon coral]

It is a known bug that some answers are empty. A solution is being worked on.

[worn monolith]

Ahh, cool.

[vernal dirge]

Nah I think it's their own model for the default version

[gritty sierra]

Hey, Just found about promo code for 2 months free, but is there other way to use it that don't use credit card on bank account data?

[halcyon coral]

Hey, that's not possible.

[gritty sierra]

OK, thanks

[tame current]

gpt 4 for the precision, Claude 2 for the bulk work

[crimson saddle]

Hello, just subbed to pro, is there any chat for pro support? or no?

[wraith crow]

You should have access to #👑│pro-lounge and #1111765634267742408 now

[crimson saddle]

ty

[tame current]

nevermind i see thr severity of openai

god that sounds so fucked

microsoft devours and then destroys the company and its products

[grim trellis]

What are examples you put in the “bulk work” category?

[tame current]

sorry bulk work as in high text input

[grim trellis]

Ah I see thx

[tame current]

Since it has a high context size

or they say

[grim trellis]

I haven’t played with Claude 2 much 2 mostly gpt4

[tame current]

100k tokens?

32k tokens max on gpt4 from perplexity pro

Its still 8192 tokens if it isnt needed

[grim trellis]

I see - do you know if plans for gpt4 turbo? @halcyon coral might know? Although OpenAI in such disarray who knows…

[tame current]

you may get daster responses in pro lounge i think

faster*

[halcyon coral]

#👑│pro-lounge message

[grim trellis]

Thanks! Sorry for repeat question 😊

[winged geode]

I rejoined discord after subscribing to Pro, but I still get “no access” how do I fix that?

[wraith crow]

You should have access to the channels now

[winged geode]

Awesome thanks for the quick response

[copper willow]

I guess I have the same question as @crimson saddle I didn't realise I was not seeing pro content

[gilded zealot]

What do you think are the reasons why Claude 2.1 and GPT4 gives dramatically different answers to the same question? The web pages being pulled in seems to be the same

[thorn moss]

it is not claude 2.1

[gilded zealot]

well, the ui says so:

[tame current]

WHAT

THAT FAst??

That’s actually insane

[thorn moss]

oh i have not yet checked and there was no announcement. Got to ask them first if that is legit though

@kind vault @neat elk

[tame current]

Still flaude 2 for me

either this is anotehr chatgpt plus thing where only a few get early access

or its a visual mod

[thorn moss]

nah its claude 2 still probably

maybe its a bug and there is no official announcement

[tame current]

nah its claude 2 for me

[tame current]

How

[thorn moss]

in web its claude 2.1

in Mobile its claude 2

so probably bug

[thorn moss]

i dont believe its claude 2.1

[tame current]

same

theyd make an announcement if they did

[thorn moss]

yeah

[tame current]

cuz perpleixty doesnt own claude so it wont jusy be special slow rollout

[gilded zealot]

might be a/b testing going on, and i won some lottery. But changing 2->2.1 in the ui cannot be not a deliberate change.

[tame current]

UI change definitely

no lottery of any sort

[gilded zealot]

probabilistic feature flag testing is kinda like a lottery? 😛

[tame current]

ui change isnt a lottery

unless web gets a quicker rollout

[thorn moss]

even rewrite in web is still claude 2

nah i dont believe it. not that fast

[tame current]

so its modded ok

it cant just be “lottery”

[thorn moss]

devs got no answer still so nah

[tame current]

fair

[gilded zealot]

ui changes are tested using feature flags, where you can decide what percentage of users get to see a feature protected behind a flag.

[tame current]

so youre special

okay

[gilded zealot]

nah, not saying that at all. just that it is not unusual, and you dont have to be "invited" to see features that other may not see before a GA release

[tame current]

literally chatgpt plus

Took me weeks to get gpt vision or image generation before others

openai can stick a nail in their servers

[gilded zealot]

lol, yeah, i got dalle-e 3 a week before my wife, while she got GPTV before me. even though both features were announced at the same time. the staggering of rollouts is not consistent.

[tame current]

yeah

i hope it isnt like that for perplexity

i just want some stability yk

[thorn moss]

but still its not claude 2.1 😂😂😂

not unless they announced it

[grand seal]

It’s showing up on my end too

[tame current]

[thorn moss]

yeah but how

[tame current]

wild

wild its wild this is rigged so rigged

[grand seal]

Only in the settings on the web

[thorn moss]

rewrite no, mobile no.

[tame current]

web yes

mobile not yet

i guess theyre changing it quite soon

[thorn moss]

oh boy if that is claude 2.1 and they put 2.1 in mobile for sure it will take weeks. 100 percent

[tame current]

i hope its a cissual change

yeah its a visual change

2 and 2.1 are almost the same

i hope its just a quick swap

ok goal today is to use all 600 requests

typically i only use 200-300

[thorn moss]

tf. nice goal 😂😂😂

[tame current]

i pay i must use

[strange kelp]

https://www.anthropic.com/index/claude-2-1

[sleek vortex]

perplexity api is free now??

[strange kelp]

https://www-files.anthropic.com/production/images/model_pricing_nov2023.pdf

[sweet jasper]

I got access to claude 2.1 now..

They are doing a progressive rollout

[sleek vortex]

Functions means they could bring copilot like user questioning though

[sweet jasper]

Uhm

200k context

Literally double context window

[sleek vortex]

200K context

Less hallucinations

[thorn moss]

there is no announcement guys chill

sometimes bug can happen so chill

[thorn moss]

only in web

[modern crown]

Has perplexity gotten better?

Before the search results were not the best

I have been using phind but wanted to check back

[sleek vortex]

how do i get pro role

[austere bridge]

I also have pro..

[sleek vortex]

Because i dont wanna leave the server to use the invite

But there’s nothing about how to get the role

[austere bridge]

the link linking to the pro role doesn't work-- I've done the same process

[sleek vortex]

Where’s that

oh

The link

yeah

thought you meant there’s some seperate linking thing

[sleek vortex]

but yeah

[austere bridge]

also still pissed I signed up at 1:30 pm and then 2 hours later 2 months free is announced for new subs..

[sleek vortex]

or is there no pro role and this just is the server

[sleek vortex]

It says returning subscribers

Maybe you could get a refund?

[modern crown]

Has perplexity gotten better ?

[worn monolith]

In what way?

Personally, I've found Perplexity to be fantastic since I purchased Pro a while back. I haven't seen it hallucinate a result in at least a month and a half, if not longer. The different modes allow a lot of flexibility with the type of response you want back, which I've found useful on occasion. I mostly use Perplexity as a research assistant, so...in that sense, yeah, I think it has gotten better.

[modern crown]

It used to give incorrect information a lot

Been using phind quite a bit

So was just seeing if it has gotten better

[sweet jasper]

Try it out

Its running gpt3.5 with search

And gpt4 or claude

With or without search

[modern crown]

Are they going to add the new GPT?

Gpt 4 turbo

[sweet jasper]

Probably when jts production stavle

Gpt4 turbo is ratelimited on api on openai

This is more on openai side than perplexity tbh

For those interested in claude 2.1 recall precision with 200k

Since this should be posted more often

[tame current]

gpt-4 is a lot more accurate huh

nice

[sleek vortex]

My theory is that

They’ve moved free perplexity to a in house model

“GPT4” might be gpt4 but it might even be gpt3.5-turbo

Remember when they tweeted about copilot using a gpt3 fine tune??

People say that when you ask gpt4 what model it is, it says gpt3

Asking gpt4 on openai site or places will say 4

And they can’t fine tune gpt4 so people saying it’s the training data being from gpt3 doesn’t make sense either

¯_(ツ)_/¯

[old kindle]

So is Claude 2.1 actually available on the website? Is it actually 2.1 or 2?

[sleek vortex]

We can’t really test since the file upload cap is 4 files

Somebody could try and upload a 200k token file but

We don’t know until they announce

Some people (including me) have “2.1” in the drop-down but we don’t really know if that actually is…

[spark phoenix]

Hey kinda off topic here. Can current subscribers use the two free months? How do I apply it to my account?

[old kindle]

[old kindle]

thanks

[neat elk]

yes, it is 2.1 already

[tame current]

HOLY COW

that’s cool thanks bro

[thorn moss]

when will 2.1 be ported to mobile?

[thorn moss]

nobody cared to question this one while sir denis is online and answering.

https://tenor.com/view/facepalm-really-stressed-mad-angry-gif-16109475

[grizzled canyon]

nah

[tame current]

bro theres no port 😭

mf this isnt a game

[steel igloo]

where can i find a decent resouce for promps ?

any will do or is there that one site/link?

[lime stump]

aweosme

[grizzled canyon]

Learn how to write your own. Many "prompts" are not very useful and are very fluffy

https://www.promptingguide.ai/ @steel igloo

[steel igloo]

Thank you!

https://tenor.com/view/thanks-fanren-xiu-xian-chuan-donghua-gif-532834689717725543

[neat elk]

mobile also uses 2.1, it just takes a bit longer to update the UI

[arctic plover]

Can users 'collaborate' on the chat with the uploaded file ?

[fleet olive]

Clinical and laboratory syndromes of disorders of the hemostatic system. 1. Normal parameters of vascular-platelet(thrombocyte) hemostasis 2. Laboratory diagnostics of thrombocytopenia, thrombocytopathies 3. Indicators of coagulation hemostasis, their changes in various pathological conditions. 4. Hereditary and acquired coagulopathies (laboratory diagnosis of hemophilia).

Clinical and laboratory syndromes of disorders of the hemostatic system. 1. Normal parameters of vascular-platelet(thrombocyte) hemostasis 2. Laboratory diagnostics of thrombocytopenia, thrombocytopathies 3. Indicators of coagulation hemostasis, their changes in various pathological conditions. 4. Hereditary and acquired coagulopathies (laboratory diagnosis of hemophilia).

[tame current]

Would I be able to use Perplexity to spin thousands of SEO articles to take over Google results?

[old kindle]

Thanks for quick updates and adaptation of new models on the website

[vast lark]

Just checked my mobile app on android (pixel 6) and don't have the option either.

[grizzled canyon]

that would be very slow lol, you would be better off using a smaller model instead of perplexity and search

[tame current]

Any recommendations?

[grizzled canyon]

like the dude said it is using 2.1 when making the request

[grizzled canyon]

do you know how to do SEO?

it's really multiple things

[tame current]

I think you just write a lot of articles

About a keyword

[grizzled canyon]

nah not just content

the crawler bots handle more than just content, they also have metrics for measuring the accessibility and speed of the site

they will also pick up on keywords and rank the site

[tawdry nova]

"spin articles" lol, use perplexity to find accurate information instead, thee. feed that Into another article generator maybe

[tame current]

nice

[grizzled canyon]

how else do sites say, for services such as plumbing, electrical or building get ranked high on google search? there are more metrics than just keywords

[grizzled canyon]

more efficient yeszx

work smart 😉

[tame current]

nice

thanks, i'll figure it out

[worldly field]

https://docs.anthropic.com/claude/docs/claude-2p1-guide#prompting-techniques-for-claude-21
how does this work on perplexity?

[hearty quartz]

How to switch back to Claude 2.0?🥺

[tame current]

I also wish there was "no censorship" mode. One of my use cases is generating effective insults and trolling users.

Especially combined with Twitter bots, I'm trying to troll everyone hard.

[tame current]

bro cant even spend time thinking about insults

wants insult
too lazy to do so

[sleek vortex]

Use the API or something

Mistral models aren’t “censored” if that appeals to you

or open source local LLM

[sleek vortex]

engagement, relation to users query so on

SEO is a whole field

And googles algorithms for search is massive

Why do you think they collect so much data everywhere?

[tame current]

disable adblocker

might work

[sleek vortex]

what’s the error?

[signal hamlet]

Hey @runic hatch! If you have some adblockers, turning them off or creating a rule for our site could help, as reported here: #👉│feedback-general message

[rancid spade]

What version of gpt-4 is perplexity pro using?

Is it the gpt 4 turbo?

[tame current]

gpt-4 turbo isn't public access

it's gpt-4 base, but enters 32k when needed @rancid spade

[tame current]

it reset

i can't burn thru 600 prompts today

argh why doesn't prompt reset at 0000 😭

[granite ocean]

I found a bug

[tame current]

then report it

lmfao

[weak plover]

Sorry newb question how many tokens in and out?

4k?

[outer raft]

What is the input token limit for search prompts and AI profile? Search - 2000 tokens | ~8000 characters AI profile - 1500 characters

#📘│faq
Doesn't say what output is

[cloud sequoia]

openai.error.InvalidRequestError: Invalid model 'pplx-70b-online'. Permitted models can be found in the documentation at https://docs.perplexity.ai/docs/model-cards.

how can I use pplx-70b-online via API ? @kind vault @tame current

[wooden portal]

it’s pplx-70b-chat-alpha

not -online

[halcyon coral]

It can be a bit more respectful @wooden portal ...but yes, the online models are not yet available, please also pay attention to https://docs.perplexity.ai

[cloud sequoia]

pplx-70b-chat-alpha only updated till September 11, 2021.

we need online version

[halcyon coral]

Yes, but it takes some time…

[cloud sequoia]

@halcyon coral openai.error.RateLimitError: Token rate limit exceeded, please try again later. ，
how can I apply to increase ratelimit..

I am evaluating the model of it’s pplx-70b-chat-alpha

[tame current]

ah so it's just that

so there's no "real" limit?

[halcyon coral]

There are rate limits https://docs.perplexity.ai/docs/rate-limits

[tame current]

ok thank

none on pplx

cool

[past galleon]

I just want to confirm, because maybe I'm misunderstanding. If you're a pro user, you can experiment with the API at no extra cost as long as you stay within the rate limit?

[halcyon coral]

Yes, if you reach the limit you can temporarily not make any requests, so you don't have to pay anything.

[past galleon]

cool, thanks for the clarification

[cloud sequoia]

how can I export the question and answer I asked in perplexity ?

[toxic venture]

top right share button?

[jovial haven]

why i can't ask or imagine now with perplexity?

i miss something?

[halcyon coral]

#📢│announcements message

[timid sleet]

hey! I can't buy Pro, can't log it 😦

[velvet flume]

i have problem logging in to the website
when i click on sign-in and enter my email or any other signin method it should error

[wraith crow]

If you just get the error message “Error“ please disable adblock and try again

[fierce beacon]

Thinking of moving to Perplexity instead of ChatGPT premium, any use cases where yuo still find ChatGPT quite useful?

I don't need dall-e, I don't need crazy plugins, just models trained on huge amounts of token which can give me basic insights on things I ask. Also coding buddy

[digital siren]

Advanced data analysis, whisper voice to text, phone calls with ai, dalle 3, and able to upload photos. I have both because both have tremendous value.

If I had to have one it would really depend on what my needs were. If you need very updated info for specific projects perplexity is faster. Otherwise ChatGPT has more features. Hopefully they’ll role out to perplexity at some point.

[zenith atlas]

I find chatgpt more easy to use when you dont need internet search - such as generating texts, drafts and other stuff. I find it harder to use writing mode on perplexity and that its not following instructions on the same level, probably because of the starting prompt.

I also use claude 2 on their official website and not on perplexity because perplexity is not feeding whole files to a model, and as such its impossible to work with big texts for me.

But when it comes to internet search and research there is no doubt I'll use perplexity as even free version with copilot is way better then other alternatives

[fierce beacon]

Thanks for your insights @digital siren, these are important differences I did not think of .

Thank you @zenith atlas as well, I will look into claude 2 and free perplexity then

[opal flame]

Hi, is pplx-70b-online based on llama2?

[toxic venture]

#⚡│ask-community message

he said the files uploaded are summarised by the models??
but i do kinda feel similar to you. the summaries feel weaker in perplexity than in GPT-4

[rapid plover]

is I'm the one who has issues with perplexity now? like all my threads are gone.... all setting are gone...

[muted anvil]

it's slow now

idk, everything loads very slowly

[halcyon coral]

Please… #🔔│status

[sweet jasper]

Yeah 500s

Some buggy push or infra going poopoo

[fringe patio]

mee too

[maiden elm]

When Perplexity is offline

[muted anvil]

just use MSFT Copilot as temporary substitution 💀

[maiden elm]

When it goes online

[rapid plover]

well it was not available but now it works with issues... if my thresds are gone for forever i cancel pro 😆

[zenith atlas]

https://cdn.discordapp.com/attachments/1111786888626438245/1175746860095389758/image0.jpg?ex=656c5a97&is=6559e597&hm=af2f68317040060ac9916108c92b8a268b476a74c5d61dd36d0619f489240b72&

[muted anvil]

Just save everything important to local

never rely on Perplexity Thread storage

[toxic venture]

Ah damn . It's written there .
Yeah that really limits the use of perplexity vs Claude directly

[halcyon coral]

Threads will be available soon

[old kindle]

🍉

[toxic venture]

Now I understand why some of my PDFs uploaded generated really terrible summaries

Or transcripts

[zenith atlas]

I tried same texts and original claude worked fine with summary while on perplexity there were always some issues. Although on pro chat some people made tests and confirmed that perplexity models sometimes can’t find specific things in documents while this same models on other platforms can do it

[old kindle]

When will perplexity be back

[rough haven]

I thought my wifi was bad haha, perplexity down? Always when I need it the most aarhh

[toxic venture]

Hmm so kinda feels that it's better to pay for gpt4 and Claude separately and use phind free 🤣 sorry to say that here

[zenith atlas]

This thing with documents is one of the 2 biggest issues for me that are not letting me to use only perplexity. Second one is not the best writing focus.

But I hope both of this things will get better

[halcyon coral]

Soon

[hearty quartz]

Is Claude 2.0 coming back?🥺

[toxic venture]

but i can totally understand why they chose to do it this way. coz of cost...
so i suspect its going to hard to change.
hmm was thinking of going long term with the black friday promo

[zenith atlas]

I prefer to have harsher limitations on file upload then to have this snippet stuff that is not even explained in full detail

Like what is big file? How it summarises this file?

[toxic venture]

yeah i know! i uploaded an S&P rating report, and all it did was tell me the disclaimer lol

lets see if things change. thank u!

[serene egret]

I used only around 30 co-pilot searches today, but somehow I have no co-pilot searches remaining for the day.

Can someone help?

[halcyon coral]

#🔔│status

[neat elk]

we are working on resolving the issue

[lean niche]

when trying to log in through google, this error popup

[tame current]

do you know whenabouts it'll be back?

[halcyon coral]

There is no ETA 🙏

[tame current]

okay

[rigid gull]

Clicking on Library is not bringing up my threads. Sometimes collections come up, sometimes it doesnt. I've tried it on 3 different computers and mobile to no avail. Just wondering if it's an issue you'all are working on with the back end?

[halcyon coral]

#💬│general message …..

[wraith crow]

please disable adblock and try again

[rigid gull]

oh man..duh. My eyes did not see the status channel. Thank you!

[humble night]

It is at this moment I realized how much I depended on perplexity🥲

[rigid gull]

Yea, so do I! Though I worry about becoming TOO dependent on using it for stupid simple things that I would remember instead of asking a question and just pasting it after small editing 😉

[humble night]

Using google to search stuff up is soo unefficient lol

[coarse jay]

@wise mesa hola

[lean niche]

Thanks that works!

[wicked blade]

Anyone else noticing slow access to Perplexity?

[worn condor]

Im noticing it

[worn monolith]

Me, too

[carmine fern]

lmao i didn't realise there was my ip my bad

[wicked blade]

I'm also unable to use Copilot.

[worn monolith]

Mine fought me to log into my google account, then showed 0 copilot uses once it did, then logged me out on a refresh.

[halcyon coral]

Guys please have a look at #🔔│status !!

[formal kayak]

Any plans in future for discord Bot 🥹?

[halcyon coral]

No, no plans

[eternal saddle]

Threads don't load for me

what's the cause?

[rigid gull]

#🔔│status message

[humble night]

it's back for me

[halcyon coral]

🆙 #🔔│status message

[neat elk]

should be fixed now

sorry for the troubles 😢

[rigid gull]

Yup. Seems to be working. Thanks to all over at perplexity for fixing!

[neat elk]

we run out of int32 in one of our tables 😭

[rigid gull]

Just out of curiosity, what does everybody here choose as their default? GPT-4, Copilot or Claud 2(2.1 now)?

[mint flower]

how do u get he pro role?

[wraith crow]

By having an active pro subscription. and rejoining the server with the invite link on the settings page

[mint flower]

um i just tried it

for some reason i didnt recieve pro

this right?

[wraith crow]

should be good now

[halcyon coral]

GPT-4

[worn monolith]

I definitely prefer GPT-4, too.

[rigid gull]

I'd say that too, but I am finding that there are occasions where claud-2 generates something better.

[mighty pebble]

what’s the usage of “copilot” button？

[halcyon coral]

https://blog.perplexity.ai/faq/what-is-copilot

[mighty pebble]

thanks！

[opal flame]

Claude 2.1 > GPT-4 imo

[fringe patio]

🤔

[zenith atlas]

gpt 4, sometimes with copilot and sometimes without, but mostly I use them together

[tame current]

eh claude 2.1 has more censoring and dumb refusals compared to claude 2, which already had much higher refusals than gpt-4

perplexity ai is so fucking infuriating

whenever i say "don't do (thing)" it goes out of it's way to do it

[granite ocean]

Ye

[grizzled bronze]

is this not meant for coding at all

asking it coding questions is not working lol

i tried claude 2.1 and gpt4

[viscid star]

I wonder which version of GPT-4 is used in perplexity?

[mint flower]

there are versions of gpt 4?

[toxic venture]

i see quite a few AI profile prompts for Claude in the #1124060874785685554 , are there any suggestions for GPT-4 ?

[viscid star]

I often find that just the standard response from the perplexity model is way better than GPT-4 for coding. ChatGPT is getting really bad, and I'm happy I unsubscribed from them

[viscid star]

Yeah there is

GPT-4 0314, GPT-4 0613, GPT-4 32k, GPT-4 Turbo

There are other versions but these are the main one's

For coding, you want to use GPT-4 0314. The newer one's just refuse to answer

[grizzled bronze]

its 32k

[gentle lily]

Does it support zip file?

[edgy raft]

How can I use collection feature on Android app?

[tame current]

is login broken?

i get an auth error when trying

[final nimbus]

I currently use ChatGPT Plus. If I switch to Perplexity Pro and use GPT, would the results be same if I ask same questions?
Thinking of buying Perplexity Pro since it has Claude etc too

[tame current]

chat yes

writing mode

but you wont get images or vision yet

[tame current]

disable adblocker

[tame current]

GPT-4 (base??) and only switches to 32k when needed

no clue if its 0314 or 0613

[toxic venture]

and its hard to process long pdfs..

[tame current]

0314 maybe

it doesnt even know that its gpt-4 rn

calls itself gpt-3

[sweet jasper]

Give us a prompt, we will replicate it.

[strange kelp]

Right now Perplexity Pro is available free for two month. Just visit this link https://www.perplexity.ai/pro?discount_code=THANKS23, the code will be applied automatically. You will have to select a plan (go for monthly) and enter your card details.
You won't be charged until two months. If you are satisfied with PP than ChatGPT, you can switch.

[crimson crypt]

if i had 10 apples, i ate 3 yesterday and i'm going to eat 6 tomorrow, how many apples do i have today?

keeps giving me the wrong answer

even with gpt 4

[strange kelp]

https://www.perplexity.ai/search/if-i-had-ta6BBJEuRz65nMSdHCDqNw?s=c

This is from GPT-4 (Co-Pilot enabled) in Writing mode

[strange kelp]

You need to change the focus

If you want to use GPT 4 without internet access

[crimson crypt]

how do i do that?

[crimson crypt]

oww, let me try it again

[strange kelp]

co-pilot helps when asking math questions even in GPT-4 writing mode

[crimson crypt]

couldnt find it on mine

it use to be there tho

[strange kelp]

Now enable co-pilot and start a new thread. Asking the same question in the same thread may confuse AI sometimes in Perplexity.

when creating a new thread, dont forget to change the Focus to writing mode

https://www.perplexity.ai/search/if-i-had-6SdrchJUTmePToGJpNIhmA?s=c

even without co-pilot it gave the right answer to me

[crimson crypt]

works, thanks

https://www.perplexity.ai/search/if-i-had-XM6F8gbnTLKbeR8a9SwohQ?s=c

[peak dawn]

Hi everyone, does anyone know if Perplexity chunks up the file you upload and if it uses a similarly embedding search and retrieve top similarly chunks before passing it to Perplexity to generate an answer. Just want to know so I know it is reading the entire document and not just 3 pages or hallucinating

[zenith atlas]

#💬│general message

[peak dawn]

@zenith atlas Hi, can we verify your experience. It should read the entire file if they create embeddings first and retrieve similarly text based on your query. I know claude does not need embeddings due to its massive token context

[peak dawn]

@zenith atlas Did you try switch to claude 2 in settings and then upload a file? Remember to switch focus to Writing so it does not use the internet, and instead use your file as a source

[zenith atlas]

It’s info from the FAQ, it doesn’t matter what focus or model you use

And yes I tried

[odd lava]

I hope if OpenAI bankrupted, Perplexity will still running the model lol, Perplexity sometimes hasn't worked several times this week

[thorn moss]

How bankrupt? they are supported by msft?

there are also many LLM that will emerge in the future. Perplexity is the best search engine they will be stronger more in time

i sense the force within

[sleek vortex]

I mean their free search isn’t even an OpenAI model anymore from my guesses

And openAI is not going bankrupt any time soon…

[tough turret]

Hello all, do you known any model provider that offers grounded internet search LLM via API? I am thinking in doing it myself but I thin someone must have done it before...

[sleek vortex]

Don’t believe anyone offers this via api

Isn’t too hard to do yourself

Manually you could scrape bing or DDG

Or use a library to do that

Or just look into something like langchain which can also handle splitting and embedding search if you want to do that

[paper viper]

https://youtu.be/4Ff2ZrhVkp0?si=myHtFaxbFNWnSbXD
Could this have an impact on perplexity?

[sleek vortex]

Probably not?

Considering openai’s growth and funding by microsoft i feel that their models and products will be available one way or another so like

[tame current]

Will gpt vision be implemented?

[tough turret]

thank you so much. valuable advice.

DDG is?

[peak dawn]

@tough turret Perplexity has an api for that

[peak dawn]

Perplexity has there own models also

[tough turret]

Hi Marco, the api doesent offer internet grounded search yet. ITs in the roadmap though. Am I missing smt?

[peak dawn]

My apologies, it is an upcoming feature

[dark solstice]

it is so weird how every-time i have perplexity a question, about history or anything going on around the world perplexity tends to say "it is very complex and muiltifaceted" without actually answering my question on what is going on

[craggy prairie]

Hey guys, I was wondering if anyone else had noticed a decrease in quality. Perplexity barely answers my answers these days. Even ChatGPT is giving me better—albeit slower—answers.

[craggy prairie]

literally

look at this, same question. perplexity vs chatgpt:
pplx: https://www.perplexity.ai/search/3edea595-bc43-497d-a340-d8d5eba8da40

[dark solstice]

im asking a question about israel- vs Palestine like "what is the history of it" and "why is it happening" and all it is saying is that "the situation is very complex and multifaceted" without giving me info on anything

[craggy prairie]

chatgpt:

for some reason i can’t share the link

perplexity didn’t even answer the question

I love perplexity, but I feel like it’s been giving me very low quality answers lately. The steerability is also worse, bc it forgets the context of the previous questions

[dark solstice]

my computer is fucking trash too

[craggy prairie]

wdym

[craggy prairie]

I don’t mind the clarifications like this, but they should be at the end of the answer. It should provide the most direct answer to the question possible first, and then give you all the context and whatnot

[odd lava]

Ikr, but for PRO users, the own model still shared in ChatGPT except they're replacing it with something besides Claude

https://www.perplexity.ai/search/4a587580-acfa-4574-83d5-b7e431cf809e

Lmao Sam

[crystal perch]

[sleek vortex]

DDG being duckduckgo

It’s slightly more scrapable but results are of lower quality in comparison to bing or google

Googles official search api is 100 queries per day after which you incur steep pricing

Bing you can probably scrape, I haven’t looked into their API though…

[tough turret]

yep

i use serpi

but i assumed I had to mix that with LLM

[sleek vortex]

Ah okay

[tough turret]

to have something that remoetelly resembles perplexity

[sleek vortex]

Might already be an intergration out there

[sleek vortex]

perplexity i’m guessing might work like this

at least on non copilot mode

• user input query
• some basic model or maybe a cache that can compare similarity to extract the keywords from your query
• searches the web or a cache alongside
• has those websites embedded in a vector db and then it queries it for snippets (otherwise how would they fit like 15 urls into 16k context?)
• passes compacted snippets and a prompt that tells it to reference sources and act like a search output

And then obviously they’ve fine tuned it with their large amount of user data

And their model for normal search might actually be based on llama/mistral and not gpt3 by my guesses

But that kind of slow probably is easy enough to replicate

You could get a preprocess LLM step to output a list of searches

Search and then use LLM again with context

And copilot might use ideas of function calling to ask the user questions?

Could actually link to whatever you want - a “local” perplexity-like LLM with code interpreter or something might be nice

Also embeddings docs search larger than 4 files…

[sleek vortex]

Have you seen the microsoft stock

Went down hugely for like a day

Now it’s back up

[odd lava]

I saw it lol

Almost hundreds dollars lost

[crystal perch]

[solemn trench]

im already subscribed to perplexity pro, can i use THANKS23?

[coral summit]

when the pplx-7b-online api be available?

[toxic venture]

Are you using Claude 2.1?

[toxic venture]

After all the discussions above and reading up , I kinda feel what they do is they summarise all the 15 url found , then feed it through the Claude/gpt
Same as how they summarise the PDF or txt file you upload before they send it to models
(Team, pls correct me if I'm wrong)

So your assumption about a vector might be wrong

[wraith crow]

you can when you switch from the monthly to yearly plan

[minor spindle]

bing chat i using it but is so limit

and i cannot text latex suit image with bing chat

[autumn cargo]

New GPT4 turbo or even just gpt4 or gpt4 32k from openrouter on typingmind api UI with talktweak google search and website scraping plugins works great to do web search and site scraping research like perplexity but i use both cause often perplexity pro (while has shorter different results) results that are different enough to add some diff value to query especially where info from many more sites is helpful and integrated into a response. Turbo is nice for research because its much more up to date and better generally. I still find it useful to use both. New claude should also be avail on openrouter soon im hoping if you cant get it directly like many

[toxic venture]

does it cost a lot to do each search?

im just thinking of being able to control my sources, like upload some reports/documents and insist it being part of the search process.
also dont like that Perplexity summarises my PDF/Text docs before sending to the model.

[craggy prairie]

no, gpt-4

[granite ocean]

Both are same? Perplexity pro also provides gpt4 and I compared it with chatgpt plus all the response are almost same.

[granite ocean]

Bing chat has a lot of layers of protection and modifications even though it is using gpt4, it doesn't look like it is using gpt4

[minor spindle]

is using gpt-4 is new but i hate is limit

[granite ocean]

yeah

[tame current]

free

[cloud sequoia]

is perplexity using bing api other than google API ?

[craggy prairie]

not really, the question was about the file size of the photos, not how many photos you can store on different capacity phones

[granite ocean]

ig

[distant grotto]

Hello, is perplexity down? it does not work on my side

[civic seal]

I can not use either

[dire stump]

Yes for me, it's down

[distant grotto]

thank you for you info, hopefully it will be working soon again....

[strange kelp]

Yes it is down i guess.

[woeful ember]

perplexity is lagging badly for me, anyone else? Whats going on

been slow for awhile, completely down now

[halcyon coral]

Sorry guys, should be fixed soon 🙏

[regal frigate]

。。

[halcyon coral]

@lean cairn

[lean cairn]

OUHH

[lean cairn]

thanks for the tag. Was worried ngl

[nimble rampart]

Perplexy = rocky balboa

[odd lava]

Everything should be okay now after OpenAI bossman comeback, its been fifth times fr

[tame current]

frfr

[lost bison]

Can Perplexity offer Claude 2.0?

And btw how do we add the discount code from the email

[sleek vortex]

Summarising seperately might also be a thing they do

Not sure myself

[sleek vortex]

they already do for pro users

[sleek vortex]

“sudo systemctl restart perplexity”

[halcyon coral]

It‘s Claude 2.1

[sleek vortex]

Oh he specifically wanted 2.0?

[halcyon coral]

Yeah ig

[valid iris]

I just pressed the button in the email, and it happened automatically

[warped sequoia]

i don't get why would one choose perplexity over gpt-4

isnt perplexity model fine tuned gpt-3

im trying to understand what is it its fine tuned on

why not fine tune gpt-4

so you get all the depth of training data of 4 + fine tuning

[halcyon coral]

Perplexity (FT GPT-3.5) model is faster

[gritty nova]

Is the GPT-4 used by Perplexity Pro users the GPT-4 Turbo version?

[halcyon coral]

No, the normal GPT-4

[zenith atlas]

OpenAI was letting to fine tune only 3.5 but on their last announcement they said fine tune of gpt4 is on the way, so perplexity probably will fine tune gpt4 when they’ll be able to do so

[earnest pebble]

How do you know which AI model should you choose?

[halcyon coral]

GPT-4 is the smartest, I almost always use that, if I know that I have a large document, I also like to use Claude because of the 100k context or now 200k context. Claude can also be very good for writing tasks.
I personally don't use Perplexity at all, but it's very good for getting information on a topic quickly.

[opal flame]

Is there a way to use perplexity on Apple Watch?

[halcyon coral]

Yeah, you can use a shortcut https://support.apple.com/en-gb/guide/watch/apd99050d435/watchos

[opal flame]

Starting the app is not possible so I use the shortcut to use perplexity with the browser (Safari), correct?

[halcyon coral]

A shortcut like this should work

[opal flame]

Thank you!

[near imp]

Does anyone know if there's plans to Upgrade Perplexity Plus with Claude 2.1?

[zenith atlas]

It is already, UI on mobile just wasn’t updated

#📢│announcements message

[dire stump]

does this work on android?

[inland axle]

Could someone explain to me what's the point of Perplexity Labs? As far as I understand, it's just for one time chats with the given AI models
Where does the API key come to play? Can I use the API to use Llama 2 or is that not the point of it?

I read the article Introducing pplx-api, but I don't understand much to be honest

[velvet anvil]

The main site is for your day to day usage for AI augmented search.

Perplexity Labs is like a playground for open source and fine tuned variants that you can experiment with. These are models above and beyond what you see on the main site (GPT and Claude variants). You cannot continue conversation once you close the session.

[inland axle]

So if I wanted to use Llama code, I'd still have to self host it somewhere?\

[velvet anvil]

Perplexity API is if you're a developer looking to use LLMs for your own ideas or workflows that save you time. The models available with APIs are predominantly available in labs too. Currently in beta and free to use with Pro. After general availability, it’ll have it’s own billing structure. You can use the API key to use LLAMA 2

[inland axle]

That's what I'm mainly looking for, coding assistance

[velvet anvil]

Do you have prior programming experience?

[inland axle]

Some, I've worked with some AI right now, I know enough to know when code it writes is unnecessary or if there's a better solution to it, but doing everything on my own wouldn't get me far

Mainly working with Kotlin (Also by extension, Java)

[velvet anvil]

Check this page out
https://docs.perplexity.ai/reference/post_chat_completions

Right side will show you how to make API request for different languages. There's also #🧪│api-general

[inland axle]

I also saw Replit in one of the Perplexity's blogs, looking into it rn, wondering if I should just look into that further or Llama code

[inland axle]

Don't have access to the channel

[velvet anvil]

Probably pro role has not been applied. If you have a pro account, check this out
#💬│general message

[inland axle]

Also, the reason why I'm looking into those alternatives is that I feel like recently I've had to do a lot more explaining to GPT models and they seem to miss more context
Not sure if it's just my issue, but I'm trying to see if I can find better alternatives that are more meant for coding

I joined back with the link, still didn't get the pro role

Got it now

[glass totem]

How can I get perplexity pro with claude 2.1 to only summarize go in further depth from a long youtube transcript?

[quaint wigeon]

Does perplexity pro have any special copilot related to programming knowledge or code learning?

[blazing dust]

hey all. i subscribe to perplexity pro. i also subscribe to replit and just got the 4 free months coupon. i also just got the coupon that was given to users from perplexity for 2 months free. can i use either of these coupons with my already active subscription? does the replit coupon expire ? thanks

[halcyon coral]

The 2 months free offer is only for new or returning subscribers, please contact support regarding the coupon from Replit via your settings

[rigid gull]

Does anybody know if there is an eta for the iOS app being able to switch between gpt-4 and Claude? (Rewrite) Thats the only thing that keeps me using the web version on my phone mostly.

[haughty bramble]

Do I need to cancel my existing subscription and sign up again to get the deal?

[halcyon coral]

If you currently subscribe monthly, you can switch to annual and enter the code 🙂

[lime lodge]

If I reg today and paid for first month 20$ can I get 2months free ? 😉
I'm a user of ChatGPT 4.5 for few months, but I thinking about change to perplexity bacause ChatGPT last time is to slow

[umbral mulch]

I love perplexity

[digital siren]

I’m curious if perplexity has future goals of adding more gpt 3.5+ level models to the core service? I’m not asking for confirmation of llama 3 or Gemini. I’m just curious if as a service perplexity sees itself sticking to two models or adding more as they become useful? I think it helps perplexity to stand out to have one subscription with access to multiple models.

[blazing dust]

ty

[hoary ravine]

Hi, everyone. I'm Victor, a Peruvian sociologist who wants to become a Data Scientist. I'm a content creator too. I've been using Perplexity as an academic assistant. To be honest, I'm not sure exactly what its difference with other AI Chats but I'd like to learn it. Currently, I'm working on a video about Microsoft's impact in AI and I'm going to use Perplexity to help me to find credible information.

[sleek vortex]

Perplexity looks at web sources and also files, ask you questions and then provides relevant bits to the model to augment and provide context to its answers

In comparsion to free chatgpt, you get web based context answers, gpt4 and other models if you’re a pro subscriber and you can upload files

in comparison to paid chatgpt you’re missing out on code interpreter and sending model images but arguably perplexity’s implementation of web augmentation is better than chatgpt’s

[hoary ravine]

Thanks for clarifying

[minor spindle]

are you india

[vivid kraken]

cannot update the profile picture no matter what, anyone else having this issue??

[dusky holly]

Hello, I just subscribed to perplexity pro and I was wondering on how I can apply the promo code.

[fallow glade]

is there a macos APP for perplexity?

[halcyon coral]

No, there is no macos app

[halcyon coral]

You can only enter the code before subscribing and when you switch from the monthly to the annual model 🙂

[dusky holly]

i didn’t see where I could have entered the code before subscribing

[dusky holly]

Thank you. Could I still do this even though I’ve already subscribed?

[halcyon coral]

You can contact [e-mail] maybe there is still a way to get the free months 😉

[gleaming frost]

Is it possible to subscribe to perplexity using Russian card?

[halcyon coral]

As long as the card is supported by Stripe, yes, you can just try it out 🙂

[gleaming frost]

I've tried and for some reason there is error. The card is usd and the purchase amount is 0$ (free two months), but it just doesn't work. From the app, the price is in rubbles, but google pay refuses to open.

[halcyon coral]

You can contact Stripe support about this, but I assume that Stripe does not support your card :/

[gleaming frost]

Okay, thank you. That's sad

[fast talon]

Hey everyone! I’m new here, I have questions about the API bata, is there someone here who can help me?

[halcyon coral]

Feel free to ask

[upbeat dawn]

is there a windows app?

[valid iris]

I had Claude 2.1 be my model, and it gave a response with no sources linked in the response, so then i edited my prompt to make it cite sources in the response, and it cited 10 sources in the response when it found 19 results.

[halcyon coral]

No

[upbeat dawn]

😔

[strange kelp]

It searches maximum 25 sources but uses only a few them to generate answer. So it find only things that matter to the query to made in the question.

[valid iris]

But how come no sources were cited in the response until i specifically asked for it ?

[strange kelp]

GPT-4 and Perplexity models always cite in the answer but Claude doesnt. Maybe Claude works that way.

[valid iris]

Ok so, i have tried to read about this, and i am still confused: does perplexity use google or bing to search the web?

And in the free version of perplexity, or when i use co-pilot, is chat-gpt 3 doing the writing, or chat- got 4, or bing, or perplexity’s own ai model?

[floral panther]

How many prompts from copilot is allowed in the free plan?

[halcyon coral]

#📘│faq

[ornate wharf]

I'm new here. Is there a website about the collection share links❔ Maybe like https://gptshuntai.com/

[fast talon]

Great thanks - can I use the perplexity Ai API key in a web application designed for OpenAi? I see on the perplexity website that their API framework is designed to work as an OpemAi API call. Is this just for custom code (Pyrhon)?

[halcyon coral]

Custom code? You can use the Pplx API just like the OpenAI API, a detailed example can be found at the bottom of the website https://blog.perplexity.ai/blog/introducing-pplx-api

[cerulean hare]

So GPT4 or Claude 2.1?

[halcyon coral]

No, I don't know of any website

[past stirrup]

Gpt4 is the better model overall
Use claude2 for files

[fast talon]

Claud2.1 is vastly improved over Claud2 so far it seems superior to GPT4 IMO - it’s a little too soon to tell though

[dusky holly]

There’s a Claude 2.1? is it available in perplexity pro?

[wraith crow]

It is, you can choose the model on your settings page.

[fast talon]

So just past in my Perplexity API key where an OpenAi key should go?!! That’s it? But there is a dialog box to select OpenAi models. Do I just select 3.5turbo? I get that my perplexity API key is model differentiated.

[halcyon coral]

Oh, now I understand what you mean...
No, unfortunately that's not possible, as you would also have to specify the endpoint or the model somewhere :/

[civic copper]

I wanted to subscribe to the annual plan using the promo code “THANKS23”, but I made a mistake in the operation and did not use this promo code.🤦‍♂️

[fast talon]

Too bad - thanks so much for clearing this up!👍 Could somebody help me to duct-tape the perplexity API call to my tools API using both keys?

[halcyon coral]

Feel free to contact [e-mail] 🙂

[civic copper]

Thank you
I wrote about my situation at "Pro Support", it says it takes two days to process. I haven't been able to sleep for two days now (=_=)

[halcyon coral]

Normally you should receive an answer more quickly 😉

[civic copper]

I must be too tired, that's why I made this mistake.🤦‍♂️

[meager hornet]

what’s a good perplexity model to use?

[lilac cipher]

Is there a way to have Perplexity AI limit its source information to a particular domain? For example, this is the result I am looking for... https://help-v3.phasereditor2d.com/scene-editor/working-with-parent-objects.html

But I want to ask the question in the form of...

+site:help-v3.phasereditor2d.com how do I work with parent and groups?

Presently, Perplexity AI will return information based on Phaser (the underlying framework) instead of instructions specific to the Phaser Editor 2D program.

[tame current]

https://youtube.com/shorts/W27zwbL6H3o?si=kwisKCK4zWCGusnD

[steel sonnet]

Why 3?

[halcyon coral]

https://discord.com/channels/1047197230748151888/1173623903596326912

[steel sonnet]

Then why in openai it says it is 4

?

[signal hamlet]

Hey @lilac cipher, here is an example for you - #1126524614449832146 message

[halcyon coral]

With ChatGPT it is in the system prompt.

[steel sonnet]

How can you proof that this use gpt4

[runic linden]

Loving the app so far, been using for a few days and even got pro.

I feel like the default perplexity model works the best, especially in longer conversations. I decided not to use the app as it burns the battery on my phone (iOS 17) and lacks web features. Looking forward to seeing this app grow.

[halcyon coral]

How can you prove it on ChatGPT?
Only GPT-4 can answer certain questions, GPT-3.5 cannot. Also, the answers are often written differently from the thought process.

[steel sonnet]

Tell me for first, why someone is kicking me third time and gpt 3.5 is answers better then this 4

Then show prove its gpt4

[tame current]

you cant

it doesnt know its gpt4

only chatgpt-4 knows its gpt-4

[steel sonnet]

Its bing gpt4?

[tame current]

its gpt4

not bing gpt4

not chatgpt4

its gpt4.

you get gpt4 from the API

you get chatgpt4 from chatgpt plus

you get bing gpt4 from bing AI

gpt-4 doesnt know its gpt-4

only chatgpt-4 knows its gpt-4

does it make sense?

if it says gpt-4 its gpt-4

perplexity (fine tuned gpt3.5) would not have any indication

ALSO

you dont have a pro subscription

you dont have copilot enabled

youre using perplexity, which IS GPT3.5!!!!!

[steel sonnet]

I have pro

[steel sonnet]

Should i enable this?

[wraith crow]

You don't have to, you can find some use cases for copilot here https://blog.perplexity.ai/faq/what-is-copilot
Also like IceLavaMan said you could ask some questions GPT-3 can't answer to check if it's really GPT-4. Here is an example:

Non GPT-4: https://www.perplexity.ai/search/a421d8db-f482-459c-b292-638c9ddd75e0?s=u
GPT-4: https://www.perplexity.ai/search/I-have-15-KeWA0OZhR5mliUh4FQ7v3g?s=c

[tame current]

if you have pro, its realky dependent

thought you were on free

mb

[sleek vortex]

well

it might just be fine tuned mistral/llama

[woeful ember]

Thank you. This is way more informative than top google search results (dumb ad blogs with 2500 words)

How else do you guys web search using AI? (Besides perplexity)

[woeful ember]

Googled and found https://www.geeksforgeeks.org/chatgpt-plus-vs-perplexity-which-is-the-better-ai-chatbot/amp/ says :

On the contrary, Perplexity AI uses an older language model- GPT-3.5. At first, you might think, this is it GPT Plus is better. It’s not that simple. Of all the models, GPT-3.5 is proved to be the fastest but GPT Plus with GPT-4 model appears to be beating its predecessor which uses GPT-3.5. Well, this can be considered as an exception.

[tame current]

nope

[heavy thistle]

Where is the API so I can add a perplexity bot to my discord server?

Can I create my own discord bot?

[viscid wyvern]

Do you have affiliate program to promote your ai service? I am a experienced affiliate marketer.

[fresh seal]

Any opp for dall-e to be offered anytime soon?

[sweet jasper]

As soon as perplexity 70b drops on api you can, right now you can use other models to power your discord bot, perplexity offers llama 70b with quite fast inference times.

[heavy thistle]

Alright thanks

[inner vault]

Hello. I'm quite new to Perplexity and I didn't find a way to select the focus or upload files through the Perplexity Android application (no problem with web version)... Did I miss something ?

[halcyon coral]

Yes will be added soon

[halcyon coral]

File upload is not possible atm

[halcyon coral]

No, there are only referral links.

[inner vault]

Got it. Many thanks

[steel sonnet]

i want generate articles for my site, should i use?

[tame current]

my advice is to write it yourself

but if you had to

copilot would be better, since it tailors to you

[frosty heart]

thats really a good example.. I tried with all the models in perplexity labs, chatgpt 3.5, claude 2.1, claude 2 and bard. Only gpt4 was able to pull that off correctly.

gpt4 with copilot gives wrong answer for some reason.

[steel sonnet]

why?

[fresh seal]

I’m confused. It was able to pull it off but what was incorrect?

[tame current]

it's your own website

put some effort

ykwim?

Is there anyway to display the math equation properly (subscript/superscript/symbol etc.). It is quite difficult to read when it is math heavy. I am using perplexity model as i am still considering pro subscription, would this be remedied if i upgrade? In chatgpt, even the 3.5 model (which perplexity based on for fine-tune) is showing equations correctly all the time. Thanks!

[steel sonnet]

Gpt4 is writing good so why i should write myself

[toxic venture]

i kinda feel that perplexity is good for researching information, and chatgpt or claude is better for writing
but pls correct me if im wrong, i cant get the perplexity prompts right

[near imp]

OMG FOR REAL ❤️ You guys are the best! I love Perplexity!

[tame current]

Is GPT4 vision now on perplexity?

@halcyon coral

Since the API is now available

[halcyon coral]

GPT-4 Turbo is currently only available as a preview model and is not yet suited for production traffic. The stable production-ready model will be released in the coming weeks or months.
Please be patient 🙏

[tame current]

Np thanks

[tame current]

Hi, how do I remove the annoying reference strings 1 2 3 4 when I copy the answers? It takes too much time to remove them manually. It's driving me nuts.

[halcyon coral]

#💬│general message

[heavy thistle]

I’ll ask perplexity a question like what are mortgage backed securities and use the answer it provides and the sources to learn and what ever I don’t u understand I’ll ask ChatGPT to elaborate more

[heavy thistle]

Exactly this

[toxic venture]

i have a question though...
so perplexity searches the web for and (1) summarises the articles
then it (2) summarises the results into what we see.

Lets say if we choose claude 2.1, does claude come in in (1) or (2)? or both?

[heavy thistle]

What’s Claude in comparison to perplexity

[halcyon coral]

The final answer is written by Claude, Perplexity is used for the routing process up to this point.

[toxic venture]

i didnt pay for claude pro, but i imagine same as chatgpt...
but chatgpt is more useful for now , due to the custom GPT.

For my needs, i spend 90% of chatgpt in summarising podcast transcripts (>10k tokens each) and reports, make it easier to understand the context and also determine if i shd bother reading.
i was hoping to inlcude the transcripts/reports into Perplexity search but doesnt seem to be working well.

i alos use chatgpt to write articles, by feeding them pointform and turning into proper articles.

But i do like perplexity for the really quick search function. very good for research

[toxic venture]

oic. so u feed the full websites to Claude (via perplexity) to generate the final answer?

[heavy thistle]

Try AIRPM extension it’s a library of prompts that could help you

[halcyon coral]

All relevant content

[toxic venture]

thanks! let me look that up

[heavy thistle]

You’re welcome, it’s pretty OP

What does Claude do in comparison to perplexity

https://blog.perplexity.ai/blog/introducing-pplx-api

If I paid for the pro to get an API how difficult would it be to make my own discord bot?

I don’t have a coding background

@toxic venture What do you think of the library

[tame current]

Will Perplexity have an image generator?

[halcyon coral]

DALLE3 will be integrated soon

[sweet jasper]

Ask any llm model to code the discord bot for you

All you really need is to know the url for perplexity api and what params it takes

[heavy thistle]

I've tried

error after error

with no coding background, not knowing how to fix them

[sleek vortex]

if you want i’ll do it for you

keep in mind that the perplexity api isn’t the perplexity website

its only open source LLMs on the API

No like normal perplexity or copilot or gpt-4 or claude

[heavy thistle]

Does it do what perplexity does?

[sleek vortex]

I believe not?

[heavy thistle]

does it pull sources and summarize those sources

[sleek vortex]

I dont think that’s available yet on the API

[heavy thistle]

So it wouldnt be any different from chatgpt's api when you ask questions and get general responses

[steel sonnet]

can i auto remove sources from gpt answer?

[halcyon coral]

#💬│general message

[steel sonnet]

there is no way to auto disable it?

[halcyon coral]

Only in writing mode

[sleek vortex]

Yes

[sleek vortex]

Can i get pro role? I can show you my subscription status

[halcyon coral]

sure

welcome to pro 🙂

[sleek vortex]

thank you 😊

[steel sonnet]

can i get too?

oh i am

[halcyon coral]

😉

[steel sonnet]

XD

[tame current]

alright then

[cedar bluff]

any word on when collections will come to ios? 🙂

[tender python]

I would like to get pro status or role too. Is that possible? 😄

[tame current]

i have an idea: write a bot that posts answers to stackoverflow but with errors, so that people were super miserable

[earnest nexus]

Has perplexity.ai gotten stupider? I remember few months back, it's output were so good and now it just blurts out whatever

[thorn moss]

can someone confirm this?

[wraith crow]

You can get that role when you have an active subscription and join the discord again with the link on the settings page

[sleek vortex]

I just posted a theory in the pro lounge but I believe (no proof) that free perplexity doesn't always use OpenAI models anymore

[teal oyster]

Might be chat gpt got stupider

[lean cairn]

Why is it that even with pro, i cant get prioritised for contact ? i tried messeging the chat thing on the settings page and sent an email, still nothing

[halcyon coral]

When did you send the message?

[lean cairn]

yesterday afternoon 😦 the message still isnt seen

it shows on the chat as "not seen yet" so i emailed in

[halcyon coral]

It's the weekend and also Thanksgiving (used by many for a four-day vacation), which means all users have to wait, you're not the only one 🙂
Sending messages plus emails only delays the whole process...

[lean cairn]

Ouhh my bad I didnt realise it was Thanksgiving ( I live under a rock )

[lean cairn]

Should i stick to the message or email ?

[glass swallow]

Can the chat completions API cite its sources at all?

[halcyon coral]

Where you receive a reply first, you can then simply write back that the email or message can be ignored 😉

[halcyon coral]

Chat completions does not currently support internet searches and is therefore unable to cite sources. However, there will soon be an online LLM, but more information will follow.

[glass swallow]

ah thanks!

[sinful plover]

howdy.. will claude 2 make its way into the android builds?

[quasi palm]

NO Code interpreter?

[thorn moss]

This is just what i have observed. Can someone tell me, why this happens like a never ending roller coaster, when perplexity improves the model or something to improve people noticed it and say it ofcourse in the channel, but after approximately for weeks suddenly someone says perplexity gets worst and it doesnt give good output anymore. Why is that? Does the system SELF revert itself from being bad or i dont know. I just noticed that like in a cycle. Is the improvements just like temporary? Hey people are you noticing that one? Doesnt make sense. Better if someone can give an answer to this or maybe a theory atleast. @halcyon coral

[feral hazel]

How do I get this 2 months free trial? Do I have to subscribe for at least 1 month to get 2 months free? I don't see an option to get 2 months free on the website https://www.perplexity.ai/pro?discount_code=THANKS23

[halcyon coral]

Claude is already available on the mobile apps

[sinful plover]

ahh sounds good.. needs to be selected in a browser I'm guessing?

[halcyon coral]

Can be selected in the settings

[sinful plover]

cool, I had to log out and back in

[heavy thistle]

Yeah perplexity shits on ChatGPT-3

💩

[tender python]

I tried it several times with the pro link and also reentered the server. I would be happy if you could take a look at it @signal hamlet.

[halcyon coral]

👀

[heavy thistle]

What’s better perplex Claude or chat4

For learning

[halcyon coral]

GPT-4, it's the smartest

[heavy thistle]

https://tenor.com/view/neil-de-grasse-tyson-mind-blown-mind-blowing-gif-4752028

I have my own personal AI professor

[rich widget]

Can anyone second the above info?

[heavy thistle]

Idk his name is gold and his name already explains itself

[fleet flume]

For the purpose of document review, Claude 2.1 can produce outputs with better context comprehension.

[heavy thistle]

Example?