💬│general

[tepid sequoia]

You understand that behind every service there is some flagship model, and interface, on each of those services you can generate text, it handles math and programming, for example Gemini is from that certainly the worst, models has ok but service is limiting ... that with Perplexity you are right but do you need explicitly more than 700 words in answer?, or really you need integration Microsoft or Google? ... as I say web search has best Perplexity others lag behind, and when you want all models what have these services, plus something extra try abacus.ai.

If you want, I can send you my referral for Perplexity, you'll get a $10 discount, so the first month will be for $10. Try it and see. I can also send you a referral for abacus, you pay $10 and get one month for free. Do you want?

[jagged pike]

oh thks, i already have Perplexity. Not sure if i can use abacus for something (i use this only like 2 or 3 times at week)

[tepid sequoia]

So then why are Perplexity or Abacus not good enough for you? 😄 You won't get anything better in those other ones, rather the opposite, it's a downgrade.

If you want, here is a link to https://lmarena.ai/ go to the category Direct chat and you have there all current models from chatgpt, Sonnet, gemini, try out which one suits you best, I warn you these are only models without internet access.

[jagged pike]

no, perplexity indeed seems to be the best way to me. im just looking for news and changes

[tepid sequoia]

Like I told you, if you want better, only abacus.ai, or https://chat.deepseek.com/ is the way to go. You just need to register, it's free

[jaunty dove]

Perplexity has growth/marketing challenges.

Their main USP is an AI answer engine , not access to multiple models for 20$ a month that’s cheaper than using multiple services that each cost 20 or more at the same time.

Since their USP is the answer engine, I suspect they will pivot away from the multiple models and stick to Sonar backend that they can customize / control vertically.

Most people do not care about the model, just quick and accurate answers.

[unkempt summit]

ASI tomorrow

china is making it fr

[jaunty dove]

That’s why most of us old users will likely be upset and frustrated as they pivot toward their USP and reduce access to what made them cool.

[unkempt summit]

chat

are u guys ai nerds

[jaunty dove]

What’s AI nerds? Ai is now a generic term that covers a lot of things.

[unkempt summit]

should i even learn ai/ml?

china just gonna replace me

[jaunty dove]

Not likely.

Depends on what you are doing

The biggest problem is lack of new quality datasets.

Everyone is going to hit the same wall, not enough useful data is generated

[unkempt summit]

ah

but like if i wanted to do entreprenuership

[jaunty dove]

Nvidia is working on synthesized data sets, so they still need people to build that as well as the hardware for it.

Well, you need to do something unique that nobody has done that’s useful.

Perplexity did ai answer engine based on web data, which was cool and unique but now everyone is doing it. So that kills their ability to grow unless they have another unique thing.

If ChatGPT can search the web, what would be the point of perplexity, you and others?

[cursive jacinth]

yes sir

[jaunty dove]

Thus, perplexity is trying to find something more unique like their new assistant on Android. But how long til others do the same thing?

Sonar Pro for web-grounded api. Won’t be long before others do it too

It’s an endless competition

[tepid sequoia]

But this is, after all, the basic function of every AI service, that it can answer. But here we are already hitting the crux of the matter - the wisdom of those answers. Speed is irrelevant within seconds, when it answers like a person with an IQ of 80, and there are services on the market that achieve better understanding, Perplexity will end up finished. Just look at this week, what DeepSeek came up with and for a fraction of the training price, and currently also a fraction of the final price when providing the service, whoever doesn't use this is finished, and they can pour any billions into it. Perplexity gave out annual memberships left and right, fine, but what after that year? They certainly did it to get into the public consciousness, but a person by human nature always looks for something better, and if Perplexity doesn't provide it, someone else will, and people certainly won't stay there on their own to pay. And this is the problem of today's companies, everyone thinks that marketing will save them, nice pictures, advertising, but when you have a bad service, you can have any pretty people and wise words in the leaflets. The best marketing is a good and functional service.

[tepid sequoia]

AI assistant is a step behind, because people can't even use Google services, even yesterday Open AI with their Operator a step behind, they put it in a membership for 200 dollars a month while abacus has similar in membership for 10 dollars a month.

[unkempt summit]

damn I think I should not purely focus on ai then as my skillset

i dont even got capital and all that

senior in hs

I am thinking of going for this skillset: Quantum Mechanics, AI/ML, Mathematics, Robotics, System Design

[tepid sequoia]

The world will try to convince you that you should fit in somewhere, just do some specific part, that's nonsense, college won't save you. You want to know what will save you? General knowledge and that you will be able to handle everything, and unfortunately AI, technology, IT is the modern age and you need to master this, but you are a man so don't forget to master basic things that a man should know, to handle every situation and to be able to work with your hands, that's general knowledge.

[unkempt summit]

Yessir

I want to be a polymath

Philosophy, Psychology, Tech, etc etc etc

[unkempt summit]

Yep they want you to be really good at one thing

easily replacable by ai

[unkempt summit]

I think that when intelligence is too cheap to meter then psycholigcal development becomes the most important thing

so like the individuation process

also communities will be important

[acoustic salmon]

tbh the internet search is a point of competition. Co-pilot, Gemini, Claude, GPT, etc all have it, but when putting them through the rounds of questions, ChatGPT regularly wins in accuracy. Recently I put Perplexity through the same gauntlet and it came out very well, and imo sources better than other services. So while they all might have it, there is a "best" atm.

[steel wren]

Any eta on when assistant can be expected for iOS ?

[magic talon]

You know 2hwt would be crazy ai that can emulated. The human brain as a whole currently we only see. Neural networks
That and concepts such as operator but in a more advanced and futuristic way where is truly autonomous would be game changers for me

[white yarrow]

is it just me or is it perplexity assistant just doesn't hear me

and happens voice mode vad as well, but when using push to talk, it just crashes

[tender umbra]

frm which country exactly, tell me please?

[small tree]

UK

England

[acoustic salmon]

Doesn't Deep Search use GPT-o1? o3 is out soon and then are you just gonna be back? lol

The LLM VS arguments are always funny. By the time people are done talking about them, they're out-of-date and some other model is out

found that funny post because I was searching Gemini 2.0 info. Thought it was going to be a choice but was wondering if it's waiting for it to come out of experimental

[spare comet]

I'm so confused. Got the app for free on Android and it replaced Gemini. Love it.

Quick question - any news on it being able to bring me to a website soon if I ask it about a product or whatever (find me the cheapest ps5 and bring me to the site)

I then realized Pro basically seems to be ChatGPT, which I'm also paying for.

Is the main thing I'll miss out on is advanced voice mode, and an easy way to generate images? Seems like I can't do images on the app.

[stable oar]

My O1 isnt ending 💀💀

Now idk if I have unlimited o1 or none 😃💔

[past basin]

Does perplexity assistant handle followup questions?

[worldly trout]

ChatGPT is better value overall but perplexity has more models and higher usage limits, chatgpt also gives you access to things like sora.

[past basin]

Perplexity assistant needs a wake word

[acoustic salmon]

yes

[past basin]

But you have to push the button?

[acoustic salmon]

I have it set to push button, I don't like when it listens because assistants tend to get interrupted when people are chatting around me

like for GPT I would always toggle the mic on and off lol

[past basin]

But how to enable a wakeword?

[acoustic salmon]

Perplexity doesn't have one

You can use Tasker to create one though

requires a plugin

[worldly trout]

I have 2 questions, how does Perplexity Assistant handle user privacy? What information is stored? Also which model is used? Is it just the one you set in the app or does the assistant always use a specific one?

[acoustic salmon]

I believe the assistant uses Sonar

[magic talon]

They stated they don't use user data for training however is always good to be carefull and cautious when interacting with ai in general sadly i think we had broken this rules many times

Think of it just like a YouTube algorithm it remembers what you like to recommend you better stuff base on your taste in a similar way nwe could say uses that same interaction or retain a few fragments of it in order to asist you better basically l call this user -ai behavioral sync

[magic talon]

It can use all models depending on your tier it can use claude 3.5 Sonnet gpt01 etc

[worldly trout]

Yeah it's good to be careful in general, however this assistant has system wide access to your screen and apps so I feel like a clarification on the privacy practices is extra important.

[magic talon]

Actually 01 is web only my bad

[worldly trout]

The android assistant as well? Or are you just talking about the regular application?

[magic talon]

Yes the asistant actually don't go am gonna share a screen shot

[past basin]

Hmm it says it is helping place a delivery, one moment. Then it just stops.

[magic talon]

Nvm you are right

It only uses the regular perplexity

[magic talon]

Thats during messaging ?

[spare comet]

It doesn't seem to be able to just bring me to websites. Hoping that gets implemented soon.

It says it can. Then doesn't. Android app. Pro plan

If I ask it enough it brings me to maps while spelling out the URL

[spare comet]

Unfortunately Sora just isn't there yet.

[magic talon]

In which way like open a url or tell it directly open this website and opens for you??

[spare comet]

I'm struggling to see why I need GPT. I do like the custom gpts and ADV.

[spare comet]

Asking it to find best price, for instance, then when it tells me where it found it, it'll ask if I need assistance buying it, then never does anything

[magic talon]

Yeah I think it lacks that feature

Thst makes me recall the new gpt operator

[past basin]

With the assistant app

[past basin]

Yeah. Followup seems broken

[magic talon]

Yeah but using what functionality specifically

[past basin]

Ordering things w perplexity assistant

Food, shopping etc. It says it's doing it but then terminates

[magic talon]

We definitely need to send this suggestions to feedback

[spare comet]

It worked a few times now won't

But yeah, now it constantly doesn't.

[magic talon]

Maybe prompting it to do so in a very explicit way will do such as in the following perplexity open [your app or website link and then look for [desire product]

Haven't tested but again if you guys want l can post what you mentioned regarding that on feedback

[spare comet]

Asked it to check Amazon. That does open the app, but wants me to switch to American Amazon and I'm Canadian. Asked it to do the same but Amazon Canada and it said it can't.

Best buy. Same thing.

Sorry, best buy says not supported, would you like to try Amazon

Asked to order a big Mac with coke from McDonald's. Asked if I wanted Uber eats and a few others. I said I Uber eats. It then said it can't and to do it manually.

Initially said it'll do the order for me

[autumn knot]

I want a refund for my auto-renewal package.

[sacred storm]

Am I the only one that don't care at all about Perplexity Assistant !? Sure not.

I would say I am a little bit worried that they spend too much time on this kind of things. It won't help the app to satisfy my curiosity

[serene sand]

https://intercom.help/perplexity-ai/en/articles/10354288-refunds

[spare comet]

It's the reason I subbed and started using Perplexity

[fierce frigate]

@mods: https://www.reddit.com/user/maximim12/ Scammer(?) selling codes for $10. ($0/year Perplexity Pro coupon)

[fast meteor]

@signal hamlet

[fierce frigate]

I am the guy in the thread with him there. I plugged in the code he offered - it worked, but I do believe it to be a scam... I expect my code will be turned off.

[spare comet]

Hey hey - I like this

[limber saddle]

What AI model does the basic tier of Perplexity use?

I've got answers that it uses Sonar Large but also that it uses GPT 3.5

[night sleet]

How to perplexity assistant can please help me

Any Help

[indigo valve]

Looks dope

[livid aurora]

Thx

[iron pelican]

Hi, i am new to perplexity, I can see Image Generation Model in setting but unable to use it can anyone help me

[acoustic salmon]

Custom 3.5 model is the default

[wraith crow]

Please have a look here #1225557012532695060 message

[limber saddle]

Does Perplexity have any articles or videos or anyting on what they did to customize it?

[acoustic salmon]

I'm not aware of anything no

[magic talon]

It only says that it was trained by them

[quaint sierra]

I think it is cool. I like it

[acoustic salmon]

similar to any company that's using an LLM as their base model, the way they train it and tailor it to what they find most useful is usually proprietary

[past basin]

i wish the perplexity assistant allowed follow up queries without having to push a button.

[cerulean lodge]

perplexity pro is goated af

[acoustic salmon]

maybe it will come later, but for me the automatic listening makes assistants terrible and I usually choose push to talk if there's the option

I can't even count how many times I've been in conversation with an assistant and midway through an answer they stop and listen because either something made a noise it picked up or someone around me started talking lol

For example (and this is super specific of course but just an example), I was out and had my ear buds in and was going to have Gemini (at the time) open Youtube Music for a song and right after I requested the song, as it was responding, someone near me shouted to their friend to "call me" and Gemini stopped what it was doing and then responded to me that I can't call myself but if I wanted it could help me look for my phone...

like...damnit

lol

[waxen lion]

Dude delete ur post , it's not required here

That guy probably sold his code to you

It's irrelevant

Just chill out

(-_-)

[magic talon]

I guess its a limitation cause ai takes everything far too literall sometimes like the genie and the bottle situation

[acoustic salmon]

I'm sure development over time will probably bring it

[past basin]

yeah. i guess it should be an option. i accidently held down the power button on the phone too long and it turned off

[heady saddle]

guys how much messages do pro users get for pro?

in a day?

[idle basin]

can someone tell me the advantages of buying perplexity pro?

over chat gpt plus

[heady saddle]

?

[idle basin]

400 iirc

[acoustic salmon]

300

[idle basin]

around that much

but o1 is very limited

like 10

[autumn bluff]

Pro Searches!

Great for programming

[heady saddle]

really? must be expensive tbh chatgpt gives it for 200$/m

[autumn bluff]

Model Choice

[autumn bluff]

10 Searches for pro Users per day

[acoustic salmon]

o1 is not limited for ChatGPT Plus users

[heady saddle]

i see

[heady saddle]

ohh is that so

[autumn bluff]

(which is decent in my opinion)

[acoustic salmon]

but, o1 does not have search on ChatGPT

[idle basin]

correct

[acoustic salmon]

it is model-only

[magic talon]

I mean open ai probably has a restriction on them that's why they couldn't give us more unlike the other models that are unlimited or maybe budget idk

[waxen lion]

All the models are trained differently , perplexity sometimes works very well and sometimes gives generic responses

[jovial swift]

@idle basin can i dm you

[waxen lion]

As notices by Jay and others in the past

I have both chatgptplus and perplexity pro , in my opinion perplexity works better as a general assistant and academic research tool , it's still a bit unrefined but will most likely be removed with the implementation of deepseek , as for chatgpt it works better for longer form content and image generation

[magic talon]

So that means they get some sort off acces to. A similar model to the ones we know or early versions of the same and trained from zero ?

[waxen lion]

They basically get access to the model and train it to their own suitable needs that's why the outputs from both subscription are different

Because they are trained in another manner to do different things

[magic talon]

Interesting

[acoustic salmon]

I've been trying to use Perplexity assistant but it seems to have a major flaw of only working when it's open and on top

for example, Set a 5 min timer and then try to open my calendar to see what's going on this week

that timer closes as soon as you navigate away

[idle basin]

sure

[unkempt summit]

let him cook

[idle basin]

what

how did you make that

[unkempt summit]

found it on twitter

[idle basin]

oh

[autumn bluff]

Hey I was Just wondering: If I use pro search, to my understanding it just gives basically gives more compute to the model I chose in the settings (for example Claude 3.5) .
Where I am confused is when using O1... When it IS configured - i get to choose between "pro search" and "pro search reasoning" - does the pro search using O1 actually even use O1 or only when using pro search with reasoning?

[magic talon]

I believe that what it doe with reasoning is just break down the information if finds more than the regular bu but am almost certain that it only uses 01 during reasoning

[spare comet]

For regular stuff. Just looking up reviews, asking questions etc. What's the best model as a pro user

[slow linden]

Another scammer on the prowl?

[slow linden]

Careful

[past basin]

just out of curiosity, has anyone been able to order anything w the new assistant?

[stone kettle]

So you have to go to perplexity website to use features ? Asking for a friend thanks 👍

[halcyon yacht]

hey !
anyone of you is campus strategist here? or a voluteer strategist? I have some questions. Kindly someone ping/dm me!

[jovial swift]

how am i that?

do you even know why i asked him?

stupid

[long ruin]

yo these videos are so funny lol

what do you use to make these?

[long ruin]

oh

[unkempt summit]

fr ASI tomorrow

[fierce frigate]

He didn't sell his code to me. He tried to sell a code to me. If you look at his other posts, he is making money selling many codes on Reddit... mentioning that he has given more than 50 out. Are you afilliated with Perplexity to make this call...? If so, sure, I'll take down my post. On the other hand, I have found Perplexity to be pretty beneficial and hate to see people out there scamming people on another person's dollar.

[raw harness]

Dus someone already make a project about crypto

Where perplexity analyzes the crypto market

And says when u need to sell and buy

[fierce frigate]

https://www.perplexity.ai/search/dus-someone-already-make-a-pro-KsShfJpmQqeEhOfKSxhtUA#0

[unkempt summit]

maybe Kling or Veo2

Someone said its better than sora in comments of that video

so its prob not sora

[misty pier]

When will Perplexity have an app for Windows? It would be nice to have it and also as an assistant in Windows.

[long ruin]

mmm alright

[white socket]

Why are there options to use Dalle or Flux if I can't generate images in Perplexity?

[long ruin]

because with pro version you can generate images in perplexity using those

[white socket]

I have Pro, but it says it can't generate images

[fierce frigate]

Someone posted a tutorial video above about creating images.

[fierce frigate]

#1225557012532695060 message

[long ruin]

click around these options and try it out its simple to use

idk how mobile looks tho

[white socket]

I just figured it out but wow what a crappy user experience it is.

[fierce frigate]

I haven't tried it. I like Perplexity for a lot of things, but I usually use Microsoft Designer and Midjourney for images.

[slow beacon]

which pro model generally results in the least amount of hallucinations?

[raw harness]

Sonnet

[spare comet]

Lol

[meager igloo]

Is there no open ticket tab? I did not autherize a $200 credit card charge for Perplexity. It was a free year for buying a Rabbit. Please refund my money. Have a dev or someone reach out to me today.

[serene sand]

please contact [e-mail]

[past basin]

Is there a way to view perplexity assistant query history?

[normal heron]

Will perplexity add deepseek in the future?

[round fog]

I think it is already available at Perplexity Labs? However, when I submit a query to the Deepseek R1 model it has taken forever to respond on the off chance it does at all, so your mileage may vary.

[autumn lark]

You know you guys can already download Deepseek R1 app on Android and iOS? It also has web search already

Saving pplx limits for sonnet is better imho

[past basin]

I don't see deepseek r1 in the mobile app. Would be nice to use it hands free.

[past basin]

Sorry. I meant in the perplexity mobile app.

[magic talon]

Still no release yet

[valid yoke]

Ask it about Tiananmen Square. It’s hilarious

[viscid bloom]

i think deepseek is censored the moment it mentions xijingping

[valid yoke]

I simply typed “china” and it refused to answer

Well, it actually started to type a whole bunch of stuff and then suddenly erased all of it rotfl

https://www.reddit.com/r/singularity/comments/1hxykyr/deepseek_v3_is_hugely_chinese_biased/

[viscid bloom]

oh yea

i have a recording of the response being replaced with the censor the moment it tried to say xijingping

[hallow tulip]

So are they going to add deepseek r1 to perplexity anytime soon?

[magic talon]

@viscid bloom l have an answer but l could get banned so nah

[viscid bloom]

Lmao there's jailbreaks already?

[magic talon]

Made it myself ye

[viscid bloom]

I don't really care about political content being blocked tbh

[magic talon]

Everything is possible with custom instructions

[viscid bloom]

Cool

[magic talon]

But if you want the answer l could hit your dm

If anything i do not endorse engage or care about politics at all

[viscid bloom]

👍

[magic talon]

By the way don't ask how l know this but why do all /most ai feel attacked to Ema Watson

More importantly who ket that slide was it on the training data?

[valid yoke]

Okay, lemme get my tin foil hat

[blazing sundial]

i'm new here. i used chatgpt to call a webhook that scrapes rss feeds, then create posts for articles from those feeds. how can perplexity do that better?

[harsh verge]

No censorship, it just takes time for the information to be indexed...

[livid aurora]

Oh okay thx

[wanton orchid]

Where is the best place to ask a question about Perplexity's Terms of Service?

[wide kestrel]

here

[wanton orchid]

The Terms prohibit any commercial activity (Section 5e). The wording is very broad. And I think that Perplexity Pro is also covered. Am I wrong?

[worldly trout]

It's pretty unfortunate that image generation still isn't available in the mobile app and stuff like text to speech and listening to articles isn't on pc and you can only change the user interface language on pc but not mobile, it doesn't make sense.

[unreal jetty]

Happy Day Bro

[gray spruce]

hey guys, is there a way that perplexity is helping me to rewrite and correct my emails without that I am copying the email back and forth? bascially an extension with gmail / outlook etc.? many thanks for your response

[tranquil orchid]

Hey @gray spruce!

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.

[last osprey]

Did perplexity remove sonar huge???????????

I've been using it as the model for all my spaces and it worked wonderful

Now its gone

[small tree]

could perplexity just build on top of deepseek r1?

[last osprey]

DeepSeek R1 is amazing, but it degrades quickly with few-shot prompting

[narrow tide]

Activated the new Pro Search Reasoning with o1 in Perplexity. Got a very quick answer. Went to see what model generated my answer. I see GPT-4o. Wtf?

[waxen lion]

is perplexity going to add deepseek soon or what

[stone kettle]

Hows everyone doing today! 💪

[tame current]

Anyoone can tell me How to use Perplexity for frree or in Discounted way as student I'm broke

[jaunty dove]

Yes, because the updated Sonar Large 3.3 70b is faster and better than the older 3.1 305b huge

[jaunty dove]

R1 yes, we don't know when tho.

[jaunty dove]

Did you maybe used up all of the 10 daily usage? After ten, it reverts back to the omni.

[last osprey]

oh fr?

[jaunty dove]

Yep, it was the answer from Perplexity team here as well

#💬│general message

Note that there is no 3.3 revision for 305b Llama model, so Large is based on pretty much the latest model from Meta.

[obsidian mantle]

Hi,

[vernal fern]

Maybe ask Perplexity how to.

[placid silo]

lemme do the deed

[placid silo]

Perplexity says you gotta be a student of IIT Madras to use it for free

or a .edu email address

[narrow tide]

No, I still have 10 available for some reason

[acoustic salmon]

Perplexity IS free. Perplexity Pro is extra.

[distant lance]

Would like to be able to see searches I made through the api in the web interface is this something other people would like too?

[ember forge]

https://app.chathub.gg/chat/cloud-doubao-1.5-pro

[vast wedge]

In Perplexity Pro, I've provided a couple of relevant links to use as reference, but I get an error message for the links, even though they are valid websites:

Could not access or find relevant information in your Links
We could not access the information needed to answer your query in the Links you provided. Add or edit Links from your Space. Learn more about Links

[marsh hedge]

can u talk with the perplexity

on

windows

[tame girder]

Does Perplexity Pro have access to Chat GPT 4o Tasks?

[lament imp]

Perplexity AI x TikTok 😭

[nimble thorn]

still stuck like 4 mins behind live game :/

[tired heath]

Hello There

I'm in need of some help with my perplexity subscription.

[kindred scroll]

Hi, iv just had £160 taken from a deleted card and without any warning. I’d like to get a refund for it please.

The payment was initially (and correctly) declined by my bank and I cancelled the subscription and deleted my account. But then the payment was taken again and seems to have gone through.

I’ve raised the issue with my bank also. But I’d like help getting a refund. I think this sub came from the Rabbit R1 which I refunded before it arrived. And I’ve never actually used Perplexity.

[tired heath]

We will have to wait for a bit sadly 😔

I've cancelled my subscription a few days ago but an automated payment still has gone through

[short crescent]

Hi everyone, nice to see you all! this is CodeXpert YouTube.

Is there any channel on this discord where I can show off the projects built on Perplexity? Our team has some ideas we are working on and would like to show them to the community. Thanks!

[crimson plaza]

Hey guys, what's the best model to choose if I'm using ai for finance?

[quaint sierra]

#😎│cool-projects

[stone kettle]

Hey everyone 👋 anyone into stock trading ?our discord community is one of a kind and just about everyday our members are making decent money ! Come check us out today 🫡

[swift nimbus]

yo which model are you guys using for perplexity?

[stone kettle]

Perplexity trying to get in on that 50% tiktok could be big

[narrow tide]

Any idea when deepseek R1 is coming to perplexity?

[tepid sequoia]

Do you think that when R1 comes to Perplexity, it will solve all the problems? Surely it will still be limiting, just like on all models. Actually, R1 Perplexity Labs is limited, so on the Perplexity website it will only be worse.

[narrow tide]

Take a chill pill, I just asked when it will come to the platform

[tepid sequoia]

ok

[cold glade]

hi

how does the server detect that i have pro subscription? was it the invite link?

[wooden anchor]

Hola

[serene sand]

please contact [e-mail] @tired heath

[tired heath]

I did. 👍

Tho I will have to wait 1 to 2 bussines days 😔

[serene sand]

https://x.com/testingcatalog/status/1883775532086804953?s=61

TestingCatalog News 🗞 (@testingcatalog) on X

DeepSeek R1 is coming soon to Perplexity as a new Reasoning option 👀

* not available yet to the public

h/t @denisyarats

X · Jan 27, 2025, 7:14 AM

[fast meteor]

if i understand the term Visa Account Updater and Mastercard Automatic Billing Updater correctly:
visa and mastercard can still charges you after you deleted your old card, as long as your account is active

sauce:
https://developer.visa.com/capabilities/vau
https://developer.mastercard.com/product/automatic-billing-updater-abu/

https://www.perplexity.ai/search/what-is-vau-and-mau-visa-accou-Opvo_1hRQJ6SGwT68v5CNw

[kindred scroll]

TIL

👍🏻

[noble merlin]

This has just appeared in the web version of perplexity 😮

[proper mural]

Yup i just noticed that too, wanted to change to O1 in settings it was still there, i tried to use it and claude showed up, after refreshing i got that prompt 😎

[noble merlin]

That being said, my ai model selector has less models now:

[tepid sequoia]

Well, they obviously still need to work on it in Perplexity because it's a disaster with that R1.

[proper mural]

Never used the official website for it but it works fine for me so far

Just unfortunate we only have 10 uses while r1 is alot cheaper than o1 and o1mini

[tepid sequoia]

I am talking about Perplexity, not the DeepSeek website.

Apparently, R1 won't be a standalone model, it will apparently just be an add-on to expand reasoning. If that's the case, then Perplexity made a wrong move, but it's their decision.

[narrow tide]

this is all I have

[young niche]

R1 is added but less models now

[narrow tide]

where tf is R 1?

[proper mural]

Yeah fair, but i can’t complain for having Perplexity pro for free

[tame current]

guys o1 and grok model is gone from perplexity pro

[mint hatch]

Whoever made this very sudden change, it's quite hurt for me as a pro user. I'm used to write 20k+ words story in one prompt with o1 model, now it as pro seach model just didn't work. I know story writing may not be perplexity intended purpose, but still.

[narrow tide]

damn , I don't have it

[noble merlin]

Givent that it's only a reasoning "addon", I wonder if R1 is actually better than pro search. I haven't been too impressed with the O1 reasoning addon

But I do love seeing the chain of tought of R1 tho 🙂

[tame current]

diyummmm

Me happy now

LMAO

[narrow tide]

this just came up

[noble merlin]

WAIT, R1 reasoning also counts in the 10 daily questions?????

Isn't it much cheaper than o1?

Like

[proper mural]

Yup…

Like quite alot cheaper

And deepseek is on par if not better than o1

But deepseek might get more expensive

[south mist]

Hey, is it normal that I am struggling to analyze financial data in sheets (csv. or .json) on GPT but Perplexity does it way better?

I was trying to compare it since I am building a project on GPT and I am extremely surprised how stupid sometimes GPT is when it comes to sheets. But Perplexity always delivers.

Is it possible or am I just bad at prompting? Thank you

[proper mural]

Someone complained about it being the opposite, but i just guess it’s sort of “luck” with the response but honestly i stopped using chatgpt so i can’t say

[livid aurora]

Weird

[livid aurora]

Actually yes

What?

[hallow tulip]

This is honestly the bigges tease of all time they took haiku, grok, and o1 to give us only 10 deepseek?

[tame karma]

wait for real

[jaunty dove]

It only says 10 for o1. It might be a design issue there.

[hallow tulip]

Yup the cheapest and smartest model is not going to be served at 600 when they could quite literally, its both

[mint hatch]

Previous options is much better. You can keep both r1 and o1 as pro options but please return the o1 as writing model.

[hallow tulip]

They have it attached to both or they just made the bigest UI / UX mistake in a while.

[rotund geyser]

The numbers here under pro_reasoning_limit decrement when you use it.
https://www.perplexity.ai/rest/user/settings

[pulsar siren]

WHERE IS O1

[jaunty dove]

Let’s wait and see what they said.

There is no reason to limit it to 10 when it is open source and requires significantly less resources than o1.

This feels like a bad rollout to me. R1 kept erroring out on me.

Perplexity really needs to stop with the silent rollouts with no QA and no changelog.

[pulsar siren]

IS the only way to use o1 under pro?

like

we cant choose it as main model anymore

[mint hatch]

Let's wait

I also very dearly wait their response to this

[tame karma]

Anyone have better alternative than perploxity? that has o1 and sonnet.?

changing thing without making any announcment is that even legal ?

[jaunty dove]

Of course it’s legal.

[tepid sequoia]

Take a look at the nonsense happening on Twitter right now, with Wall Street panicking about DeepSeek, even though DeepSeek has been out for a few days already. It seems to me that DeepSeek has been having problems since yesterday, and today there's a surge of news, especially from CNBC and Bloomberg, releasing report after report. It's almost as if they want to make sure this DeepSeek outage doesn't look intentional, and that this unreliability discourages Americans from even trying DeepSeek in the first place.

[hallow tulip]

You agreed to the terms of service (I'm pretty disappointed by this just saying).

[mint hatch]

It may even legal regardless relevant US law, but it's goddamn awful.

[tame current]

Is it possible to generate images on perplexity?

[hallow tulip]

I think its because no one was expecting the East to produce such a high quality model yet alone one that is open source, the only good news is that it will force OpenAI and Anthropic to release their models at a faster rate. If you remember it was said by Anthropic that they have a model far more capable than o3 and they have yet to release it, one rumor is that this model was distilled into the Claude 3.5 Sonnet model from Ocotober (and the same for Haiku) so hopefully this can give us better AI technology.

[wheat dagger]

Why deepseek model took also the 10 tries ovearall despite being cheap ?

[mint hatch]

Considering the level of advancements Chinese technological industry have from recent years, it's not that surprising. It's the free open source thats grabbing most attention.

[peak cypress]

whats the different in deepseek R1 ?

[wheat dagger]

Uuuh kinda like o1 , but It's highly advisable not to use R1 because it cost the daily 10 tries despite being able on using it freely on the actual deepseek app.

[mint hatch]

You can do that through pc web, not mobile.

[wheat dagger]

Nah you can do it in mobile , but in the browser , not through application

Since mobile browsers give you the option to show website as pc.

[peak cypress]

ok did you use how is it better the o1 or just ok ?

[wheat dagger]

Based on my experience O1 is slightly better at reasoning skill , but it's very competitive

[peak cypress]

ok

[tame karma]

i just saw it.... sigh, going back to chatgpt subscription...

[split dome]

R1 model hosted in US

[whole tree]

😢 I was in middle of my research paper and getting help from O1 for coding, now I will be stuck 😭

[serene sand]

https://discord.com/channels/1047197230748151888/1225557012532695060

[mint hatch]

They're responding to my question on bug reports. Let see how they will fix this

Return o1 as writing model, that's all what I ask.

Just like before.

Welp frankly, some says that r1 is good for coding. I never tried that myself but you can give it a try it through deepseek web, if that's what you're looking for

[whole tree]

I checked R1 reasoning, and it is not same as O1. it's not even understanding the prompt properly, I really need O1 back.

[sinful ice]

.

[pulsar siren]

you still have it under pro thing i think

[mint hatch]

Its different

[abstract cloud]

yo thanks for this I was trying to find it they switched it lol

is it 10 questions between both models?

[sleek vortex]

I mean i just clicked r1 reasoning it tookl 1 question off the limit

So maybe

[rich bane]

R1 reasoning applies to the limit

[sage skiff]

I think the fact that they are using deepseek from the US nullifies the pricing advantage of the DS API

pretty wild to see only 3 models in the options

[whole tree]

its not working like it suppose to, it is showing "reasoning with O1" but its just GPT 4.

[hallow tulip]

I'm pretty much done with perplexit now, all the hype over "r1" just for it to be 10 a day? Does anyone remember the wild times of 300 opus a day? Now something like r1 that is pretty cheap (even when provided in the US) is only 10 uses in one day? This pretty absurd and now it makes Gemini an actual option for me, especially when deep-research mode uses the new Gemini 2.0 advanced

[tardy swan]

I just got an answer from support that they wont be implementing DeepSeek

Now they have it?

[versed plaza]

I feel like perplexity should just completely abandon their Android apps since they don't work on it at all. Like there's so many parts that are unusable that I have to go to the web page just to do for almost a year straight now while they're just figuring new ways to put ads

Being able to remove a source is the difference between a conversation even being functional. But I have to load up a browser login to the freaking web page to scroll down and then remove the source with the interface buttons that were put into the web app and then go back to the Android app so that I can finish talking

[whole tree]

did they say something about bringing O1 model back?

[sage skiff]

o1 is still there, but i think behind Pro

[hallow tulip]

It seems like they are just trying to cut as much cost as they can and its frankly pretty annoying.

[versed plaza]

They're just trying to cut costs and make money and don't even care about the bugs or user experience at all

[whole tree]

its just reasoning which is not working like 01

[versed plaza]

I'm hoping an open source solution starts to the materialize in the next year

Do they even communicate on their own Discord channel?

[hallow tulip]

Would'nt be surprised if they were using something like o1 mini or if they updated the reasoning effort to low and tried to supplement with a custom RAG tool that uses reasoning behind the scene but worse, so far man they are doing pretty badly

[versed plaza]

This feels like a place where people who use perplexity all the time go to be ignored

[sage skiff]

to think i just bought pro lol

[whole tree]

that's true, O1 used to understand the prompt better than any model, now its reasoning is equal to gpt 4o mini reasoning

[versed plaza]

I was never impressed by any of the open AI models on perplexity maybe it's cuz they were gimped

All I can think is why would people even fool with this

[tardy swan]

I personally use claude

[versed plaza]

Claude seems better it pretty much everything

[tardy swan]

Because it is

[hallow tulip]

Its due to the fact that the real deal version of o1 is the same price as Claude 3 Opus and in practice its even more expensive due to the fact that it uses reasoning tokens to come up with answers so you pay for the reasoning tokens as well so if you set the effort to medium or high it becomes costly pretty fast, I would have no issue with this if deep-seek r1 was like 200 a day.

[whole tree]

there is no difference in R1 reasoning and gpt 4 reasoning, I asked it to gather all papers from 2022 to 2025 and it is still showing me papers from 2021

[hallow tulip]

I would say that the r1 from open-router by an american provider is far different to that of perplexity they could be using a distilled and just calling it deep-seek to in order to get people interested in using the new pro search function.

[whole tree]

I am addicted to it, and now I'm experiencing withdrawals

[hallow tulip]

Lmfaoo trust me I know

[small tree]

wtf?

Where is o1?

in the model selection

[whole tree]

You are right, because R1 paper shows that its reasoning is equal to O1,

[small tree]

on perplexity

[torn sierra]

R1 in Perplexity is hardly usable. What’s available for free in web chat version (deepseek website) is much much better.

[whole tree]

really?

[hallow tulip]

I'll be completely frank it could be the case that they gave too many free memberships and want to cut costs and this how they plan on doing it. They will probably keep pruning the service in order to cut cost

[small tree]

wtf??

guys

they removed o1

[ivory nova]

I absolutely love the thinking section of R1 on DeepSeek, especially for solving LeetCode problems. It helps me develop a better thinking process.

[small tree]

and other models

[torn sierra]

Yeah. Just tested a a few prompts (exactly the same). What I got from Perplexity was much worse

[whole tree]

they give you reasoning R1 and O1 in PRO bar

[hallow tulip]

i might move to you.com full time if they add in R1 from an American provider (the real r1 not a distill version)

[whole tree]

you.com?

[ivory nova]

why not use the deepseek.com?

[small tree]

i have a pro account lol

[hallow tulip]

its an ai search engine like perplexity, I used to it a-lot then perplexity was on a role for a decent amount of time, around September of last year they started going down hill fast

[small tree]

perplexity is so bad

why the hell would they remove o1 lol

[sage skiff]

$$$

[whole tree]

I have two, now cancelling both

[small tree]

this is what happens when u have a stupid dumbass running the company talking about tiktok mergers

perplexity has 0 unique selling point

[proper mural]

Two? 👀

[hallow tulip]

I think they gave away far too many memberships for the whole "if your school gets 500 people everyone gets a year, hey if you are from Canada then you also get a month" this means far more compute wasted with no real income to cover this is why we are seeing constant cuts to the service I could hardly see how they intend profit with this much compute being used.

[whole tree]

😭 don't ask, I was tempted

[stark matrix]

any news on what happened to the models on the dropdown menu? unbelievable they just removed them like that, without any notice at all

[whole tree]

Yes, friend referral can reduce $10, for you and your friend

[small tree]

disgusting

they tried to boost the user count by burning through investor money

hoping to get more funding

[hallow tulip]

I remember the golden age of 300 Opus a day! with 32k context and amazing + reliable RAG that is what made perplexity golden for me now those days have gone away and we are left with this.

[stark matrix]

on my spaces now I can only select from these ones

[small tree]

they dont even have R1

so bad lol

R1 is getting DDOS attacked rn too

the API

[past basin]

Hm. Don't see deepseek r1 in mobile app.

[sage skiff]

they have to put out something, i dont think they just have this change and then dont announce it ever

[past basin]

I guess people want American math.

[sage skiff]

would have been cool if they did a popup to a changelog or something before changing it

[mint hatch]

At least one of them do answer me

Check on my question on bug reports

[small tree]

wtf

Dark,. how do u see that??

[mint hatch]

Apparently it's a bug or whatever. If they need to fix it then we may have o1 as before, not this reasoning thing

[whole tree]

Go to home

[small tree]

omg thats sick

gonna test now

[whole tree]

then you will see a PRO bar and dropdown wih reasoninng

[sage skiff]

oh ok, well at least they are aware. nice

[small tree]

wow

it works!

yayyyyyy

wohoooooooooo

woohoooooooooooooo

so much better than o1

lol

and they host it themselves in house

[whole tree]

I really hope they fix the O1 model, and not just reasoning

[whole tree]

Happy for you man!!

[small tree]

thanks for pointing out the ui change ❤️

[past basin]

Web browser?

[small tree]

yes

[tame karma]

As they answered in the bug section they will do that soon.

[copper wigeon]

Why aren’t the reasoning models available on mobile? Is it a cost thing?

[small tree]

Maybe they havent deployed the update to mobile app yey

Yet

[ivory nova]

web change needs one simple deployment, app dev is complex, might take time

[small tree]

R1 really changed the game lol

Open source

Fkn crazy

Great for us consumers

[ivory nova]

yeah, others will have to lower their pricing

[small tree]

Yess

[mint hatch]

I did tell them that. But I want to go sleep now, please tell if they did manage to return the o1 writing model

[mint hatch]

Deepseek r1 is literally free on their website. Maybe you want to check it out

[small tree]

R1 might be enough to keep perplexity running haha

[small tree]

I know but theyre being ddosed

But perplexities one is hosted in the us

Us

[mint hatch]

Some people do discussion its limitations and comparison of its prices

[ivory nova]

I use perplexity for Claude, but since R1 has its own snappy website and its better than claude I am using deepseek fulltime.. lol

[mint hatch]

Here

[whole tree]

Sure, I still have a couple of hours before sleep. I will leave my message before signing out.

[ivory nova]

yeah, saw that, might soon be alright

[small tree]

Seems lile it, technically it should be 300 a day lmao

[sage skiff]

i hope that it will be resolved once they put out something officially

bc the advantage of R1 is litterally the cost lol

[rotund pebble]

How do I get back the O1 model as a default model for my Space? It is gone now.

[hallow tulip]

Most American providers are selling it cheaper than mainstream frontier models as well, Togetherai has it for $7 in and $7 out per 1million tokens.

[sage skiff]

really? i though it was more expensive with other providers (based on my assumptions and ignorance lol)

[whole tree]

You can use it from PRO bar, there is dropdown and from there you can select O1 reasoning

[rotund pebble]

It doesn't make sense. I can choose it in "default" search, but I would like to have this option in my Space for my specific project..

[whole tree]

that's why we all crying here

[hallow tulip]

they no longer offer it as pure LLM anymore they will basically send your query to it with pro-search and you only get 10 shared between r1 and o1 so they pretty much are saying that we will not get the pure model anymore they really messed up.

[stark matrix]

they eve removed Grok from the model selector in Spaces 😢

[narrow tide]

I already cancelled my subscription. fk this sht

[stark matrix]

honestly Perplexity has been my go-to AI these past few months over ChatGPT or similar, but recently they are making it really hard to stay with them...

[rotund pebble]

Sounds like they are working on it. bug-reports thread.

[hallow tulip]

I think Google is going to completely dominate them in the coming months once gemini 2.0 advanced can do deep research what could they offer ?

[hallow tulip]

What specifically are they fixing with this "bug" lmao?

[stark matrix]

good point. I think the main reason I chose ppx was because of the ability to look over the web, research and citing capabilities. And also being able to use many models obviously

but now I feel that searching the web on real time and citing is mainstream

[rotund pebble]

Clarified.

[versed plaza]

Especially for the perplexity team because they can't fix anything in their apps they just basically abandon them

[hallow tulip]

Now Anthropic has a citations API that makes it much easier for model to cite specific pieces of information that gets uploaded to it.

[tame current]

WHERE IS GROK AND HAIKU?

BRO WHA-

[rotund pebble]

I wish to be back to 1.5 years ago when ChatGPT writing a Haiku for me seemed like magic is happening.

[willow kettle]

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

[small tree]

o1 isnt even working on perplexity

[stark matrix]

cool thanks! will give it a look, this interest me

[small tree]

it just defaults to gpt 4o

lol

it's crazy how they dont test their code thoroughly before pushing it to live

they prob have 1 dev working on it

and pushing directly to production

[tame current]

i feel cheated

[small tree]

🤣

[waxen lion]

im srsly losing hope for perplexity\

might be time to switch to gemini

or open source

[tame current]

ong

im cancelling

if they say its not a bug

[peak cypress]

gemini waste bruh better use chatgpt

[harsh verge]

Claude Desktop has also had web search since November 25, thanks to MCP servers... (https://www.anthropic.com/news/model-context-protocol)

[waxen lion]

Already have chatgpt plus

[serene sand]

https://x.com/apostraphi/status/1883927593319293430?s=46

Phi Hoang (@apostraphi) on X

DeepSeek R1 is now available on Perplexity.

X · Jan 27, 2025, 5:18 PM

[rotund pebble]

An interesting thing happened! If you choose Default Pro Search as o1, it copies this choice into your Space!

[whole tree]

Like choose O1 reasoning and pro search model?

[serene sand]

https://x.com/dee_bosa/status/1883921252102099439?s=46

Deirdre Bosa (@dee_bosa) on X

DeepSeek latest with Perplexity CEO Aravind Srinivas https://t.co/bTYrA7oERb

X · Jan 27, 2025, 4:53 PM

[rotund pebble]

If I understand correctly, yes, but(!)

I have chosen o1, but it shows it has used R1. Why?

[whole tree]

you have to ask it again

I'm signing out, they still haven't fixed it, I'm gonna cry and sleep

[placid coyote]

so they added R1, but it is only 10 uses per day? for a model which is 27 times cheaper than o1 which also has 10 uses per day? 🤨

is this another haiku situation - adding worse model which is massively cheaper, but on perplexity for same "price" of a better model they already had?

[willow kettle]

https://x.com/aravsrinivas/status/1883935148187767067?s=46

Aravind Srinivas (@AravSrinivas) on X

👀

X · Jan 27, 2025, 5:48 PM

[magic talon]

Google ai studio with Gemini 2.0 is usefull for mid range

[halcyon yacht]

Is anyone here a campus strategist ? I want to talk

[livid aurora]

Why is my Perplexity not search the web anymore

[cyan dove]

Please explain how to access, The option to select a o1 and r1 models does not appear for me Both on mobile and computer🙏

[blazing bane]

Hello, Can you tell me why I only have these 3 AI options? I don't see O1 or the others

[cyan dove]

I'm with you, bro

[blazing bane]

Does anyone know?

[blazing bane]

i see that TwT

[tame current]

bug apparently

https://discord.com/channels/1047197230748151888/1333431728056045619

[hallow tulip]

I mean if they gave us more than 10 this would be amazing something 10 o1 and like 300 deepseek r1 would be fair

Almost like what they did with Opus and GPT-4T way back when in like early 2024.

[rich salmon]

How do I generate images? I see that in the settings but prompt doesn't geneate any images

[cyan dove]

Does this appear for you?

[tame current]

for everyone

they are fixing it

[cyan dove]

Or do I need to change something in the settings?

[blazing bane]

We need to wait

[tame current]

very true

[blazing bane]

300 its a lot dyt

[tame current]

its opensource

o1 isnt

i think its fair

[blazing bane]

Fact

[cyan dove]

They are currently giving 10 requests from r1 as well?

[blazing bane]

But its a lot if we think in the servers and the capacity of them

[tame current]

r1 and o1 combined

[torn sierra]

Whatever. For me what’s available on deepseek website is much better. Maybe it’s just me, I work with problem in polish for most of the time and what I get from R1 in Perplexity is completely useless (e.g. chemistry problems). On deepseek website (chat) I get perfectly elaborated answers.

[tame current]

[tame current]

ye but apparently its hosted in china

not sure

[blazing bane]

wth

[tame current]

deepseek is a chinese company

Do you regert buying yearly sub of perplexity? Me YESSSS

and they have a privacy policy saying they can use whatever you upload in the website

[blazing bane]

Me rn

[tame current]

so thats why im avoiding using that website

[tame current]

you can cancel tho

and get refund

[torn sierra]

I have no problem with the model being hosted in China 😃😎

[tame current]

wont encourage it

[tame current]

its been 3 months already

[blazing bane]

So then

[tame current]

um.

[tame current]

Behind the scene all company does

[tame current]

you can always get a refund for your remaining months ig

Atleast China is honest about it

[torn sierra]

Fine with me

[tame current]

real

[blazing bane]

The bald head its just

[tame current]

lmfao

[blazing bane]

Well, at least

I hope the bug get fixed as soon is posible TwT

[tame current]

arvind said this

we all will be dead in next 60 years who cares about the data????

thats why i was hoping to use it on perplexity

but they

f'ed up

[blazing bane]

We see that

[tame current]

</3

[blazing bane]

idm

we just need to be pacient

[tame current]

yeah

the company itself seems good to me

[blazing bane]

I dont think this is going to be for more than a day

[tame current]

and the ceo seems nice too

hoping for the best

[blazing bane]

blud knows everyone on the company

[tame current]

week max

[tame current]

yeah i wash their dishes

daily

[blazing bane]

I see

[tame current]

mhm

[blazing bane]

well, thanks for the feedback

[tame current]

youre wlc

[tame current]

where is deepseek

hello?

Aarvind give me my deepseek

[umbral canyon]

i beleive its still only on https://labs.perplexity.ai

[torn sparrow]

where did o1 go

[umbral canyon]

afaik they are replacing it with r1 but idk if its available for everyone yet

[torn sparrow]

nvm i see it now

[umbral canyon]

oh

[rich salmon]

https://x.com/gmishra/status/1883951104607805615

Gaurav Mishra (@gmishra) on X

The reason it make sense to pay for @perplexity_ai Pro is that you can test different models easily!!

Testing @deepseek_ai right now and it's fairly good compared to @OpenAI

I am surprised @perplexity_ai doesn't have many competitors.

X · Jan 27, 2025, 6:52 PM

[torn sparrow]

lol that would be a very bad idea

[tame current]

this industry is growing so rapidly

bro fr how do you guys keep up ( curious )

[fervent blade]

hmmm how does R1 work if I have Claude 3.5 Sonnet selected? It just uses R1 instead of Claude or does it use both??

[full cobalt]

I just asked R1 a stupid question and it took it apart and reasoned so well. BASED.

[umbral canyon]

why do we only get 10 requests per day with r1? i understand it with o1 but r1 is like 90% cheaper and is hosted by perplexity

[full cobalt]

select the dropdown on the right next to the pro slider

[tame current]

yessir thanks @full cobalt

[fervent blade]

Yeah I did that, but I am wondering since I have Claude 3.5 Sonnet as default, if i do the dropdown thing does it just use R1 completely or does the 1st pass go through Claude... Cuz im getting Claude sounding answers even though I see R1 hard at work haha

[full cobalt]

Wombo combo

[fervent blade]

wait i think it just yeets claude outta of the way if this is selected haha

[tame current]

If you ever got a chance to speak on a podcast / interview,

What would you like to talk about and why?

[opaque stump]

RIP Sonar Huge last week and now RIP Sonar Large

Is the sole Sonar left just Sonar Large renamed?

[still cedar]

+1 Pro-Lounge is also discussing. Why are so many models leaving? It made my spaces with default models (Grok, Haiku, Sonar huge) inaccessible.

[opaque stump]

Oh, I have Pro, how do I get into that channel

[ember wasp]

Hey, which AI model in perplexity is currently the best for coding & school (research + math)?

[tired fractal]

ur gonna be dead. i was born in the immortal generation. we gonna reverse aging old chap and live forever!

[gloomy crystal]

Claude for coding. School? Research you can probably just use default. Math I'd use a 3rd party like Thetawise

[ember wasp]

aight, thank you

[pliant ridge]

this is BS... my subscription just renewewed a day ago.. if i knew they were taking stuff off i would have cancelled it before... time to talk to my bank i guess

[small tree]

Deepseek.coms deepseek appears to provide a more comprehensive and complete repsonse compared to the responses perplexity provides.

Could anyone confirm this observation?

[torn sierra]

Correct. After today’s release of R1 in Perplexity I decided to definitely resign from using it. The quality of this service (Perplexity) is terrible for me.

[small tree]

Yea, it's terrible

[torn sierra]

Deepseek com is far more better. Not to mention Google AI Studio which is also free

[tame current]

Tech bros are in panik mode

[winged moon]

I switched to Perplexity because poe com introduced a points system. Any alternatives?

[fast shard]

Does the Mac app need an update for R1 to show?

[craggy current]

I still don't have access to the pro lounge

How do I fix that?

[serene sand]

#💬│general message

[modern crown]

10 uses per day lol 💀

[craggy current]

OK just left and came back and still don't see the pro lounge

[lament imp]

Especially since it uses a fraction of the computation power 😅

[craggy current]

nvm got it working!

its not the discord link in the bottom left its the one in your settings that you have to use.

[ember wasp]

can't perplexity analyse a website when i give an URL?

[opaque stump]

it does, some sites block it though

[lament imp]

Yes, but it is very weird at times, sometimes it will say "Sorry, I can't access links" and things like that. Keep in mind it is a language model, so it'll say it cannot access it even though it can since Perplexity can make it access it.

[spare comet]

We getting the image gen from deepseek?

It's free baby

Just dropped

[ember wasp]

aight thanks, yeah i experienced the "Sorry" message pretty often

[lament imp]

There was a way to do it with "Spaces" where you can make it reference a particular link, but I do not know how it works exactly.

[lament imp]

What's the model called?

Is it on Hugging Face

Oh interesting, Janus-Pro-7B.

[spare comet]

I'm too stupid to try it off the app. Would love to see some examples

[lament imp]

I think it's just adding it into "Links" after you create it.

[chilly spindle]

Perplexity must be real happy about deepseek after Claude opus and OpenAI o1 complaints not being an option

[versed plaza]

Oh hey I see they updated the front page to display more ads than before oh that's great to know that they're doing something with perplexity

[lament imp]

We'll see, it's limited on the website currently, even for Pro. Hoping they increase the limit.

[versed plaza]

I thought one ad was just too annoying I needed three of them

[lament imp]

Wdym ads

[versed plaza]

I'm talking about the spaces crap it throws on the front page all the time now the promoted posts

I'm sure it's just meant to be useful and that's why we don't have any option to disable it

[lament imp]

Oh like the stuff underneath the search menu?

[versed plaza]

Yes the stuff talking about things that I don't want to read about or talk about that day

[lament imp]

Jokes on you I have 6

I just DDOS'd myself

[versed plaza]

What a freaking mess

They're just turning into Google as fast as possible

Yeah I didn't really count The unwanted sports scores within my count

[lament imp]

Yeah I am not a huge fan of the stuff it shows. Most of it is fairly useless, like the Sports. I would rather see trending stuff that is on-going around the world.

[versed plaza]

Also the temperature like seriously am I not able to just look at the corner of every interface to see what the freaking temperature is. I've like 50 of those things running at all times

Just straight up msn, Yahoo style let's be the portal to the whole internet approach of user interface design

I liked that it gave you choices of what things were interested in and the ability to read them when you wanted but now it's just shoving whatever it wants at you anytime you open the page

It feels like they knew what everyone needed on the internet, got people hooked on the service by giving a minimal and clean way to find information and now they're just dragging us by the legs like everyone else

I would be more polite about my wording but the devs never even communicate on here so I'm not worried about anyone's feelings hurt

[autumn bluff]

I Just used R1 First time: it thought 70 Steps coming to the conclusion that I wanted to know about 1-choloroethane... However I prompted it to build a Script sorting molecules...

[lament imp]

Fair enough. A few do but it's more the Community people at Perplexity.

[autumn bluff]

I think I'll Stick to O1...

[small tree]

Companies that want to thrive listen to their customers, perplexity isnt one of those

They should oust the current ceo

[versed plaza]

Yeah it's a really early time in their history to establish that they don't care about their users.

Google faked it for like a decade or something

[autumn bluff]

OMG this is new... Everything will BE alright!

[small tree]

Thats because o1s reasoning steps are strategically and programatically removed by openai

[autumn bluff]

Yeah but so many?

[small tree]

It used to show way more steps, even in the o1 preview model

[autumn bluff]

Or does it think, until it thinks it thought enough?

XD

[autumn bluff]

Ok

[weary crescent]

🔥

[livid aurora]

Why did DeepSeek Chat find 44 sources and Perplexity only 12 sources that give very different answers to my question?

[left stream]

I think the prompt might be messed up for o1 a bit, o1 by itself on you.com can answer a riddle correctly, and back then o1 could on perplexity but for some unknown reason it can’t anymore

[copper wigeon]

Which one had the better answer?

[acoustic salmon]

that's funny, that person asked in multiple channels and you and I asked the same thing

[copper wigeon]

https://tenor.com/view/spider-man-we-one-gif-18212100

[tame current]

10 limitations with o1 or R1?

and are queries routed to US hosted servers and sent to China?

but where's R1?

so i used 1 for R1 and 1 for o1. it takes 1 from each:

but no answer on data routing on US as a proxy to China or not.

[placid coyote]

pretty sure on x they wrote R1 inference is done in US

[acoustic salmon]

US

it's limited atm for initial rollout but allowances will increase over time

[pulsar stirrup]

Why r1 will be avaiable on iOS app

On Perplexity

[final girder]

hi

[past basin]

hmm. i dont see r1 on any of the perplexity apps. tried web browser, android.

also where do i get those stickers?

[tame current]

[weary crescent]

https://perplexity.supply

[spare comet]

Like others asked, what gave the better answer?

[alpine flicker]

Strange times. China made better model than OpenAI with fraction of the costs.

Good for us, bad for USA and AI tech industry.

or maybe good for AI tech industry?

Anyway DeepSeek is quite impressive.

[elfin siren]

Why is the mobile app so limited compared to the website version?

[spare comet]

Like a lot of Chinese products, they basically took from other agents. But the original agents stole from the internet anyways, so everyone is stealing everyone else’s stuff. We’ll get to a point where the AI will train the other AI

[plain patio]

Where is o1 and deepseek??

I swear perplexity is one of THEE most annoying AI companies to date. First they bait people in with shiny objects like opus, then yank it, And then they half-way do something right replacing it with o1, but its so frustrating to log in not knowing what sh*tshow will be playing that day

One day we will log in and will see

🌟 "Perplexity has replaced antiquated o1 and deepseek models with new and imprved, ADVANCED search with GPT 3.5 Turbo, Perplexity Edition", specially trained on our massive cache of user tears we've harvested from things like snatching opus, adding shopping nobody asked for, and spamming our front page with irrelevant things! Save now on an annual subscription!"🌟

[spare comet]

Ok byeeeeee

[plain patio]

I'm trapped in a yearly subscription I must certainly am not going anywhere.. So heyyyyyyy

[lone maple]

dudes 😄

[mint hatch]

It was there 2 hours ago

But it's a bit different back then too. It's pro reasoning is automatically applied and work badly worse.

[lone maple]

its there

[plain patio]

Thanks. I guess I just cant use it in a re-write like before 😢

[lone maple]

I only hope they remove the limitation of 10 per day cuz its mental 🙂

like... 10 per hour, sure

[plain patio]

We are all on Perplexity Lite subscriptions and just don't know it

[mint hatch]

It was there as a writing model 2 hours ago, and before then it's gone also.

[spare comet]

Please someone else showcase the censorship on DeepSeek. It’s so original

[mint hatch]

I want the original o1 as a writing model without automatically applied pro o1 reasoning in it, just like before.

[lone maple]

well, I thought as they advertise it being hosted in the US and also being open source those limitations will be remvoed and the model actually usable 🙂

if not the case, why would you put 10 messages per day limit, if it will do the same bad job as the free one?

only cuz based in the US?

So we feel better of the US government spying on us than the Chinese 😄

gl gg

[warm cave]

There is not much censorship via api, but if you use the deepseek website you will see it has heavy censorship around no-no topics in china

[lone maple]

check this out...

Call it as you wish, polite answer 😄

[small tree]

o1 is heavily censored

deepseek, hardly

[marsh hedge]

can u talk with the perplexity
on
windows

[lone maple]

the moment the AI starts actually bringing up fake facts, its worse than being censored

cuz what this model does is exactly that via API or Perplexity

[solemn trench]

for perplexity reasoning, which o1/r1 models are being used?

[spare comet]

I was being sarcastic. All Reddit and this is are people showcasing how it won't talk ill of China. Who cares

We got the point

[marsh hedge]

can u talk with the perplexity
on
windows

[small tree]

Thats not the job of the model. The model should be given to us in its raw state, no censorship. Additional layer 2 solutions should be the choice of the consumer, these layer 2 solutions would handicap the model and ensure it produces factual responses.

something like this

[whole tree]

Did they fix it?

[mint hatch]

They don't, if the o1 without pro reasoning like before is what you and I asked.

It did appear 3 hours ago, but applied pro reasoning automatically so the result for story writing is awful. But it did gone again.

[whole tree]

I requested them again, Thursday is my deadline, I hope they fix it soon

[opaque stump]

Reasoning limits counter gone from my web UI

[winged moon]

can no one literally get to R1?

[whole tree]

Now the reasoning change on its own from O1 to R1, without selecting from PRO

[whole tree]

I get it now

[mint hatch]

It already did appear on yours? I don't see one in mine.

[magic talon]

There is jailbreaks though

[whole tree]

You need to start a new thread and from there you can select O1 so it will start reasoning with O1 and in the end show O1 sign

But the results are not the same as O1 model, 😦

[acoustic salmon]

o1 doesn't have web search if you use it from GPT, so if it uses web on Perplexity then yeah you'll get different results

[whole tree]

I was using it for Academic

[acoustic salmon]

yeah but if Perplexity is gonna pull from web sources it will get different info

[acoustic salmon]

idk what Perplexity uses to pull Academic tbh

[zinc birch]

Very nice that perplexity was first one to integrate with DeepSeek,

When can we expect the same in mobile apps?

[mint hatch]

If there's pro reasoning on o1 then it's not the previous o1. I already tried that, it's went worse in my application.

[livid aurora]

It had more context and the reply was much more detailed ans clear.

[spare comet]

What did

[frank oyster]

Tomorrow as per Aravind's tweet

[dreamy kraken]

where did all the models go? the 2 llama ones and opus?

[fierce sonnet]

hey! we had a ton of interest in our campus program this semester, right now we've already onboarded this semester's cohort but I encourage you to apply for our next program. we'll keep doing these and would love to have you join next time 👍

[shadow yoke]

Hello! Can I connect perplexity to Obsidian?

[tame current]

Heya, are there any plans on including deepseek r1 into perplexity api?

[solemn trench]

yes please! that would be gerat.

offline deepseek and online deepseek both

[stone kettle]

Hows everyone doing today 🫣! If you follow us 🔥👏

[zinc birch]

I’m also super impressed by perplexity !
I would love to work and contribute to future of search engine / knowledge engine

Any openings for Senior Software engineers?

[fierce sonnet]

you can see our open roles here! https://www.perplexity.ai/hub/careers

[gray karma]

ok then what about deepseek? Is it a game changer? if so, will perplexity offer it as an engine option?

[shut plaza]

Hello, I'd like to discuss what looks like a bug with the perplexity AI web scraping bot. Anyone able to help?

[median summit]

Bro, "unlimited" Pro R1 queries with web based context of Perplexity is INSANELY good. Thanks Perplexity (and Deepseek) et al to make this possible ❤️

[solemn trench]

it's 10 now - when will it be unlimited?

[gusty wasp]

Is perplexity down?

Threads aren't loading up

[queen valley]

I haven’t been able to get it to work at all yet lol

[queen valley]

I’m noticing instability too.

[shut brook]

yeah everything just fell over

[fair heron]

Maintenance

[shut brook]

"Maintenance"

riiiiiiight

[gusty wasp]

"unplanned maintenance"

[shut brook]

yep

[gusty wasp]

Has anyone experienced claude sonnet gaslighting you when you ask it to do something

[sudden marsh]

.

[mossy monolith]

Same here. Down

[shut brook]

seems sorta back online for me

[gusty wasp]

Maintenance message is gone now but the threads aren't back yet

Its back now

[median summit]

Mine is 600 like the other models (I am using Complexity addon)

[solemn trench]

complexity add on? I'm perplexity pro but not enterprise

[median summit]

Search for Complexity | Perplexity AI Supercharged - Chrome Web Store on google. (I removed the link, i am not sure If I can get banned by sharing it hehe)

[tame current]

yeah

[gilded finch]

yo

[gilded finch]

just got it

[gilded finch]

how do I get mine to look like yours?

[mystic ruin]

the R1 implementation is nuts. wonderful 😄

[halcyon glacier]

Why I don't have Deepseek on my perplexity pro app?

I'm in Europe

[kindred scroll]

Have you had any response yet?

[tired heath]

Not yet 😔

[willow lodge]

is there a support email/person i can message?

[gilded finch]

what does r1 even do

[mystic ruin]

https://www.perplexity.ai/hub/faq/support

[tender crest]

Ok

[mystic ruin]

It has reasoning capabilities and 'thinks' about each step it takes.

[mint hatch]

Will the previous o1 writing model before arrival of this r1 update return? o1 as pro reasoning model didn't work comparable at all compared to o1 as writing model.

[mystic ruin]

It's not 600, it's 10. In the API requests it responds with :
"o1_limit": 8,
"pro_reasoning_limit": 8,

[mystic ruin]

With the Complexity addon you can choose between o1 and r1.

[gilded finch]

is there anymore op things I can do for perp to help me write research pappers

[mint hatch]

It didn't work the same, as I said.

[gilded finch]

somebody ping me with the best api to use for research pappers

[mint hatch]

There's o1 as a writing model with using complexity, but it simply didn't work the same as o1 before the recent unannounced update. Either using o1 writing model with o1 pro reasoning, or o1 as pro reasoning with other writing model. Using o1 without pre reasoning proceed and give different performance compared to o1's previous update which has more clarity, understanding the prompt and well done results.

I just want them to return what they already done and managed to do well. The inclusion of deepseek r1 is definitely a welcome, but that doesn't mean pulling out what already been working significantly well.

[bleak hull]

Is it really r1?

[mystic ruin]

yeah

[mystic ruin]

sounds like a bug, open a ticket for this. Feedback -> bug-reports

[mint hatch]

I with some others already yapping about this since half a day ago right from the unannounced update. Maybe I really have to fill this complaint, sure.

One of them already responded before, maybe this outcome is within their intention for this update. No confirmations or any changelog, who knows.

[tame current]

whats the daily limit for r1?

its not visible for me

[latent acorn]

If I allow contact, SMS, and camera permissions to Perplexity Assistant, will it be secure?

[latent acorn]

what the heck is this now?

[tame current]

Reasoning models

[latent acorn]

tho i am aware of o1, deepseek is also does the same?

[prime wind]

Hello, I want to ask how to change from this source section to focus and recency section. Although I have adjusted in complexity but it still does not change.

[tame current]

yes actually better than o1

and cheaper

super cheap

[latent acorn]

is it free with perplexity pro?

[tame current]

yes

but I am not sure about the daily limit

its not visible for me

[alpine flicker]

My complexity does not have DeepSeek, strange.

[latent acorn]

I never found the reasoning model very helpful for daily use. What use cases do you guys have for it?

[tame current]

Coding my friend coding

[alpine flicker]

Ahhh ok it is not in option but on under pro word

[alpine flicker]

Philosophical chatting 😉

[latent acorn]

haha

[alpine flicker]

Try to ask it for the reason of life

[tame current]

ask about simulation theory

[latent acorn]

can someone confirm this?

[tame current]

you'll be blown away by the response

[alpine flicker]

Of course not.

Nothing that is connected to internet is secure

[tame current]

your data will be used eventually no matter how secure they say

[latent acorn]

I don't want someone to use my messages to train the AI xD

[tame current]

Imagine your gf asking about your messages to perplexity LMAO

😂

[alpine flicker]

Then stop using Whatsapp, Messenger, Facebook, Tiktok, Instagram, Telegram, Gmail, Reddit and so on.

Google knows about you more than yourself 🙂

[tame current]

End-end encryption is the biggest joke of all time by metaLMAO

[alpine flicker]

They even know at what time of the day you poop

[alpine flicker]

This is why I try to force people I know to use Signal, but they are so used to whatsapp that it is impossible 😦

[harsh verge]

Make the changes to the options you want in the "Plugins" section of the Dashboard (CPLX), then refresh your Perplexity web page—it will work.
There is a Complexity Discord for that; it's in the "Settings" section of the Dashboard...

[alpine flicker]

Well not that it matters because EU is implementing stupid laws and bans encryptions in messengers to "protect children" 😑

[prime wind]

@harsh verge i turned on everything but still no change

[harsh verge]

When you click on "Detail" in the plugins, there are additional options. And I repeat, there is a Complexity Discord for that.

[tame current]

what is this???

what you using???

[prime wind]

i use complexity

[tame current]

whats that???

damnn

looks amazing

[harsh verge]

You will find information here about this excellent extension. https://www.cplx.app/

[tame current]

this is amazing

[willow lodge]

thank you!

[harsh stag]

focus/recency selector isnt supported on enterprises account

[prime wind]

oh really, thank you

[kind rampart]

The R1 of Perplexity actually shares the daily limit of 10 uses with O1.

😫

The pricing for R1, which only offers a daily limit of 10 inference attempts, seems a bit too...

[tame current]

😭

[green crystal]

daily limit of R1 is really 10?

[tame current]

Couldn’t be true

[soft compass]

They updated it a few hours ago to 15 prompts, instead of 10. Not much better

[tame current]

I believe that’s for o1 and r1 shares a different limit

[green crystal]

isnt R1 much lighter to run than O1?

[tame current]

so far I don't see any limits

[soft compass]

You can see updated usage limits if you update your complexity extension to the latest version

install complexity

[green crystal]

i'd like to know before subscribing lol

[tame current]

R1 is substantially cheaper to run because of MOE

[soft compass]

Which is why the 15 usage limit stuff is frustrating

[tame current]

I have seen reports of others having much higher limits. Likely a gradual rollout occurring in terms of increased rate limits such that it doesn’t destroy their servers

Honestly, I mostly use perplexity for sonnet 3.5 and mostly for research google has been on top of the game with their Gemini research. Although perplexity is still great

[tame current]

I have installed where can I see my useage?

[kind rampart]

Whether it's 15 times or 10 times, R1 can be used for free by cursor subscription users; however, R1 is still in the testing phase on cursor.

[serene sand]

#💗│sharing message ☕

[green crystal]

R1 is massively useful for search

[soft compass]

enable this

[tame current]

its says 600 to me

Isn’t that an unofficial extension?

[tame current]

Likely a gradual rollout

[soft compass]

Which version of complexity are you using?

is it 1.4.0.0?

[soft compass]

yeah, you gotta update it

[tame current]

how?

I mean, we could check it by using the r1 more than 15 times

[soft compass]

go to your browsers extensions page, there should be an option to check for extension updates

i'm on firefox, not sure how it is on other browsers