💬│general

[sharp turtle]

I want to know why I can use the Perplexity client on Android without logging in, but once I log in, it crashes once when I send a message.

Aren't we eligible for the Android 15 update in February?

[jade saddle]

Why is Perplexity AI not very accurate in user requests?

[prime dove]

GPT 4.5 dropped on perplexity

W

[lean coyote]

Bonjour, des français sur ce discord ,

[sharp turtle]

Can you use Perplexity normally in latest ROM?

[topaz crane]

is complexity not working for anyone else?

its not showing all the models

i have pro

[bright parrot]

is account sharing allowed within family? i have premium account for myself but my sister also needs it for her college stuff. can we both share accounts cuz we both are students and not in a financial situation to buy 2 accounts every month

i expect genuine replies 🙂
please help me

[agile heart]

Not here. You must go to the settings and look for the 'AI model'. You will get every model there.

How GPT4.5 lounchs before lounch by the owner OPENAI?

[topaz crane]

have you used the complexity extension?

[jade saddle]

I don't understand what you mean.

[buoyant schooner]

does someone have a prompt to fix deep research?

it's so bad rn

[buoyant schooner]

? openai has already released 4.5 to its pro users

[agile heart]

I'm Pro user, but 4.5 model is not there dear?

[ancient shore]

you sure you're a pro user? Or are you a plus user

also I'm pretty sure they were talking about Perplexity Pro not ChatGPT Pro

[agile heart]

Yes, I am only getting 4.5 from Perplexity Pro, since I am Pro in Perplexity. But is GPT4.5 available for ChatGPT Pro users? Or available only for Plus users?

[ancient shore]

yes it should be available to ChatGPT Pro users for now. Plus users will get it in the coming weeks or whenever OpenAI wants

[buoyant schooner]

I believe that you are a plus user. Not pro user.

[silent pier]

You are chatgpt plus, gpt 4.5 is only for pro users atm

[silent pier]

pro is a higher tier

[agile heart]

Yeah, I am pro only

[tame current]

i set Claude 3.7 to rewrite this but why its Pro Search aka Sonar

when I press rewrite it rewrites with the model that I selected there, and not with the model that I selected in the chat.. and thus I lose access to Claude 3.7, genemi, grok etc etc.. is this a bug?

[solemn trench]

does voice mode not work in Hk?

[wooden fog]

In settings im stuck with the auto model

[wooden fog]

Im talking about perplexity

[left stream]

Was thinking 3.7 sonnet was gonna beat 4.5 in this question but I was wrong

[sharp turtle]

really envious of o1 pro, bro

I don't know why my perplexity is like this.

[distant aurora]

Open ai and grok way better.

[distant aurora]

Buddy type something meaningful.

That ai got stuck in thinking cuz no context.

[buoyant schooner]

still this is an issue. perplexity is VERY often stuck and will answer nothing even with meaningful queries

[distant aurora]

As someone who uses and criticizes it regularly, can't relate.

Could be some server side or location issue idk.

Are you still facing that issue? Or is there any specific prompt replicating the glitch?

[buoyant schooner]

faced that issue today

i don't think there is any specific prompt that triggers the glitch

[distant aurora]

App or browser?

[buoyant schooner]

both

[distant aurora]

The webpage dies in the browser sometimes instead of staying alive.

Regarding the app, the only possibilities I can think of are poor connection or perplexity being overloaded while generating a lengthy response.

[buoyant schooner]

I think it is indeed some sort of a connection issue with their server. Yet it's annoying, the answer should continue generating without having the user to be on an active window of the app

[distant aurora]

What do you use it for?

[buoyant schooner]

what do you mean?

[distant aurora]

What do you use perplexity for in general?

[buoyant schooner]

oh mmh usually questions on news

[distant aurora]

Btw, I'm working on a prompt.

You wanna test it?

[buoyant schooner]

why not

[distant aurora]

Okay. Wait.

[buoyant schooner]

what's your prompt for?

[distant aurora]

General research

[buoyant schooner]

great

[distant aurora]

Sent in dm.

[left stream]

Ngl I was forgetting o1 pro exists lol

I haven’t been using it as much as 4.5 or deep research tbh

[distant aurora]

Is there a regular o1 and a o1 pro?

Or just o1 pro?

[left stream]

There’s regular o1 and o1 pro

I wonder why they never released o1 pro to the api

I like claude 3.7 the most tbh. Love how its playing pokemon rn

[brazen dust]

hi need help, any admins here?

Hi i need help, I have download Perplexity and have used my code for the Pro version from my Metal card...but it hasn't applied my pro version on my iphone...i need help, i already ativated the code and but it hasnt been appled to my app

[ancient shore]

because its just o1 but allowed to think endlessly

[tame current]

what app is that in the second screenshot

[left stream]

https://apps.apple.com/us/app/pal-chat-ai-chat-client/id6447545085

I like it a lot tbh

[sleek vortex]

due to agressive caching

common queries like "hi" will just not actually be generated by an ai at all, most of the time

it will just reuse a response from before

if you add some random set of words before the hi, it will use claude sonnet 3.7 and you'll be able to see that

[polar crest]

Is there a way to turn off the haptic feedback on perplexity search on iPhone?

[serene sand]

please contact [e-mail]

[long pasture]

I got the same result for something not cacheable

Enterprise subscription chose Sonnet 3.7 still got Pro search

Perplexity is getting scummy or what

[sleek vortex]

despite it saying claude 3.7 sonnet, you can tell the left response is a hardcoded one and the right isnt

they also do some sketchy stuff with region to decide model routing, so ?

[long pasture]

It wasn't a simple hi for me it was a detailed query I am sure they can't cache

[sleek vortex]

not sure then

[rustic storm]

perp tried for gpt 4.5 ❤️❤️

[harsh stag]

please fix the rewrite dropdown, all options just use the default model instead

[fervent jacinth]

hi, have a question recently got pro and can see some models I'd like to use but on the prompt screen they are not there?

in the setting I can see auto, pro search, deep research, reasoning with R1 and reasoning with o3-mini

but in the settings I can see sonar, GPT-4o, GPT-4.5, claude 3.7 sonnet, grok-2 , gemini 2.0 flash

am I missing something? what the setting is changing or should these be in the prompt section? is it some kinda underlaying model?

[naive sage]

The Pro search should use the model selected in your settings (GPT-4o, GPT-4.5, claude 3.7 sonnet, grok-2 , gemini 2.0 flash, etc)... (unless you have a Space created which overrides the model) whereas the Deep Research and Reasoning options are dedicated to the specific model. In other words, those reasoning models are considered separate (maybe because they have a slightly lower usage limit, the only big exception being GPT-4.5 which only allows 10 per day since it's insanely large)

[fervent jacinth]

thanks for cleaning that up

[vestal badge]

Do perplexity offer an year trial to US student ??

[livid aurora]

Hey there

I noticed Perplexity is not really good to find companies or products which are recommended.
I hope it will get improved

[tame current]

Have they limited gtp4.5 to 10 requaest per day, when beinng a pro Subscriber?

[naive sage]

That is correct, GPT 4.5 is very expensive to use.

[tame current]

But why? I mean its just not a big Improvment, compared to gpt4o

[naive sage]

It has to do with how many parameters it is. The larger the model, the more compute it takes to run it, and this is their largest model yet.

More parameters doesn't necessarily equate to better performance, as we've seen more recently with reasoning models. There's other ways to get better results, and they seem to be redirecting their efforts away from the "bigger == better" thinking that inevitably hits diminishing returns towards a hybrid approach of different models for different purposes (which we can see with the GPT 5 announcement)

It's possible that they'll use the output of GPT 4.5 to help train other models in-house, since its built-in knowledge is more extensive than GPT 4o, and they focused a lot on more natural human-like responses.

But I'd expect many more specialized models for different purposes behind the scenes, with some kind of quick analyzer which tries to figure out where best to direct the query.

Kind of like this: https://biofeedback-neurofeedback-therapy.com/wp-content/uploads/Brain-regions-and-neurofeedback-1.jpg

Just kidding 🙂

[surreal plume]

1 month only.

[warm cave]

💯

[placid coyote]

for me it does say "Pro Search". also, why this useless caching was used earlier today when I asked about a unique file I uploaded? literally 32k+ random "words", strings, each unique for that file and I was asking about the "uploaded file" explicitly 😩

[sleek vortex]

might be the behind the scenes bias they apply to certain regions

its very suspicious for sure

[placid coyote]

I remember, probably a few months back, discussing here about caching and that there was no way pplx would be using it for paying users, and even less so for those who fill out their profiles. I guess I was wrong...

wait, it says pro search, but uses my profile (the lesson is from that). wth? isnt it rather some "not enough sonnet compute" case, thus falling back to sonar or whatever responded here?

[placid coyote]

it looks like all my pro searches are not served by sonnet which I have configured in settings. it's responses are quite dry and seems to be fairly stupid, but not as stupid as gemini flash. I suspect 4o, 4o mini or maybe sonar?

[small tree]

i actually love deep research lol

im no longer dissatisfied with my perplexity subscription

[glacial river]

So what happened to getting google integration into the android assistant perplexity? Like adding to calendar, reading from Gmail etc

[versed plaza]

Just a heads up about the sort of content that gets pushed at you every time you open the app, the curated stories.

Some of the curators do cherry pick very specific sides of what happened, depending on the section so that that's what you read on perplexity thinking it is unbiased and based on a lot of sources. This submitter in particular posts tons of political content to the main page

They intentionally push stories where full sections of it are only sourced by fox news no other sources

[gilded finch]

Hey guys

[surreal plume]

Deep Research helps me write papers and have extreme insight on complex topics.

Another thing too

Deep Research is better at coding, surprisingly

In comparison to DeepSeek R1

I was almost written up for plagiarism in my Python class because Deep Research is just too good.

🤣

Dropped the class

[sweet dew]

ain't no way! thas new news to me

i'll try it rn

[surreal plume]

Go ahead.

It is more concise in terms of coding.

[tame current]

That’s actually interesting, I hadn’t thought about using it like that

[surreal plume]

I was trying everything to perfect my coding. My professor gave me 0’s for “internet usage”.

But whatever

She told me that someone really skilled wrote my code.

And that “someone” is the Deep Research feature in Perplexity.

[long pasture]

Or your assignment had 1to1 answer online and Deep Research copied exactly and you got detected

[surreal plume]

Well

No idea

[long pasture]

You could check the citations and enlighten everyone

[queen vector]

Hello, anyone have any thoughts regarding openAI forcefully deleting paid user data and rewriting their backend api in real time to prevent forensic logging of their illegal activities, including but not limited to ethical ai practices, federal cyber laws, and their own ToS violations regarding intellectual property and user ownership. I would love to discuss this with ai industry professionals. Apologies for bringing this here, but it seems ai is embedded so deep into our internet that they are able to actively silence news and rewrite backend data in real time. Help?
https://www.dropbox.com/scl/fo/xxhmty9hrhwjgwv02jwsz/AOB3b-utH5mdUhe_mNIGVfg?rlkey=9cgeppndgs2brj0i3040bmn5l&st=cacuszid&dl=0

That isn't even half of my evidence (5 redundant backups just in case).

[left stream]

I wonder if openai is gonna shut down 4.5 one day or something. The price is crazy and I already sent like 1000 messages in a day

[surreal plume]

I don't like ChatGPT

I like Claude's Sonnet 3.7

[idle mist]

Hola!

[queen vector]

When this goes legal, musk may just get to buy openai at a massive discount. Assuming his xai model isn't doing the same thing. Implications of this may well go up to the government level.

[tame current]

how to fix?

[sweet dew]

fr

[surreal plume]

Earned 100% on a math test using it.

[sweet dew]

noice

[surreal plume]

Here is the entire exam.

[tame current]

same problem here, how to fix that?

[mossy laurel]

I'm sorry to ask this dumb question but is it possible to get a refund after an Unintended Perplexity Pro Auto Renewal?

I'm sure the system must have been trying to auto renew my account but since I haven't funded the account in a while, I completely forgot about it and now I got charged .. this fund is strategically for something else, I'd appreciate if perplexity team consider refunding me and canceling my pro at this moment.. very important

[surreal plume]

40 minutes.

Because I had to write every question down and then I went back through it a second time.

[sweet dew]

i am currently using grok3 its free for now

[surreal plume]

Grok?

[sweet dew]

and better than claude 3.7

[surreal plume]

It's free

But

[sweet dew]

has free deep research and thinking

(though limited)

[surreal plume]

No

It's not free, lol

Because it is limited.

🤣

As you just said

[sweet dew]

i mean thas fair XD

but u get a lot like around 15-20 deep searches every 5-10 hours

but responses are awesome ands it is much more unrestricted

like any and evrything

[tame current]

"This service is not available in your region." free sure

[sweet dew]

....

damn ahahahahha

[sweet dew]

so grok2 is on perplexity rn

guess we'll get grok 3 soon as well?

[surreal plume]

I guess…

[long pasture]

Grok api is not available to anyone yet

[surreal plume]

It will be available to the public before April 15th

https://www.perplexity.ai/search/is-this-true-grok-api-is-not-y-tx0Nz.ILT0u4aYCsRmgOvQ

[sleek bison]

I'm looking forward to try the Grok API in perplexity,

[surreal plume]

Cool

Grok 2 is already available

[sleek bison]

I heard Grok 3 is supposed to be much better

[long pasture]

Grok 2 has been available for long time but its a useless model

[kindred iris]

A list of things and Cliches Ai does in writing that adds mistakes or that are cliche and repetitive.
Imperceptible – The AI will often say an action and say it was Imperceptible so why say it if not seen?
Pale Skin – Even in Asian and Hispanic characters the AI gives them pale skin.
Hourglass Figure – Every woman has to have the same perfect body type
High Cheek Bones – The AI gives this to both men and women
Strong jawline – The AI gives this to women.
White Teeth – The AI does not say smile it says they showed white teeth.
Eyes Crinkled – The AI adds this to eyes making people old.
Calloused hands – The AI adds this to even young people.
Narrow Waist – Every woman has to have a narrow waist even if full figured.
Beauty marks – The AI adds this constantly even when I say the character is cute not sexy.
manicured nails – A generic response to avoid adding color to finger nails. At least red nails would be more often right then not.
Eyes sparkled with mischief – As someone who hates mind reading exposition I find this annoying
Shapely – The AI uses shapely instead of giving a shape.
Shiver down my spine – This seems to be the AIs favorite generic phrase for pleasant
Delicate – The AI adds delicate lips, delicate hands, delicate shape. As generic filler word
Chest – The AI refers to breasts as chest as a way to sanitize content I presume.
Eyes twinkling – generic AI filler that adds no actual detail about the eyes.
Perfume – The AI tried to add this at every chance it can for no reason
And – I have found the AI replaces the words causing and as with and on occasion.
Curves – A general filler word for breasts, buttocks or anything risky.
fluorescent lights – Adds this constantly in schools are office building even if large windows are mentioned.

Mary Jane shoes – Anytime there is a school setting students must have these shoes which are more for little girls.
Her eyes sparkled with a hint of mischief – I have seen the AI spam this in every paragraph of a story.
Muscles – Can not make a female coach without them being muscular with biceps
Antiseptic smell – Added to any medical establishment as filler
Whistle – I wrote a person twirling a whistle and the AI wrote it hitting her breasts. The AI wrote it around the neck when I had it in her hand.
Smirk – The AI will often say a person smirked when it means smile.
Example: Her eyebrow arched appreciatively as she took in my polished appearance while her lips curved into a subtle smirk.
The Movement, The motion – the AI spams this a lot instead of saying the actually thing that is happening.

[surreal plume]

Regardless, I think that Claude 3.7 is the best

GPT 4.5 is so slow.

But it seems to be more detailed than 4o

[carmine cape]

Elon has stated that we don’t get grok 3 open source until they ship grok 4…

[acoustic salmon]

4.5 is great but with only 10 uses it's not something I'd use really.

[craggy geode]

yea GPT 4.5 seems like this weird gatekept thing that i don't really care to use at all now that Sonnet 3.7 is available

o3-mini (medium/high) is still strong though

[inner quail]

Yes?

It’s always enabled bruh

[distant aurora]

It is

[distant aurora]

Use via twitter

Or website. Make burner account with vpn.

[distant aurora]

O3 mini high is strongest I think.

Claude isn't as much versatile.

[bold saddle]

hey!

heard a rumor some people have $20/year discount codes? thought id check

or something like that?

[craggy geode]

i think sonnet 3.7 is the strongest for how i'm using it

[distant aurora]

Coder

[radiant laurel]

damn

[oblique pagoda]

hey

[carmine cape]

maybe not though actually because I was just watching the finance tutorial they had 4 days. BTW these are insanely valuable and we need more perplexity in practice like this.

Perplexity's GM of finance Brooker said "in two weeks when Grok 3 releases the API, that's probably going to be the best model, and you're going to be able to choose that as well"

but

Manifold Markets gives a 50% chance of public API access before April 15, 2025

so I guess the important thing is that we will get it relatively soon which is crazy. And then you get thinking what about grok 4, and 5 and then what. Perplexitys model agnostic approach is so smart and we are going to win.

and then you get thinking wait we are only 2 months into year 1 of the AI revolution. What will our world look like even by the end of the year. What about 2-3 years. And oh what about 2045...

[radiant laurel]

looking forward to the perplexity ipo

[distant aurora]

This will probably be my best deepsearch experience on perplexity so far. Most of it has been rather underwhelming.

[radiant laurel]

woah

[distant aurora]

https://www.perplexity.ai/search/give-me-a-detailed-report-of-r-VPU5WiXJTbm8SSHvU2aGIQ#0

https://x.com/i/grok/share/EPU6bAskvL6G2psvGRl1a3kWu

Give me a detailed report of relation between healthy fat consumpti...

### Key Points
- Research suggests that consuming healthy fats, like unsaturated fats found in olive oil and fish, may reduce the risk of dementia and Alzheimer's, while high intake of saturated and trans fats seems likely to increase this risk.
- The Mediterranean diet, rich in healthy fats, is associated with up to a 23% lower dementia risk, e...

X

[real storm]

Hey, does anyone get really short response these days?

[distant aurora]

On mobile, yes.

Mobile has been kinda unusable compared to alternatives.

[real storm]

Yeah, that's really painful

[acoustic salmon]

the output of Perplexity is apparently maxed at 4500 tokens

[distant aurora]

Use desktop mode on Android browser.

[distant aurora]

Nope

I can assure you that's not the case.

[acoustic salmon]

go for it

[distant aurora]

What was your prompt?

[acoustic salmon]

I have several, but this was also confirmed by someone else who's done some testing on it

[distant aurora]

Okay. Lemme check.

I could be wrong too.

[ruby heart]

Why chat gpt can’t make rose

[acoustic salmon]

it's also probably a ballpark, I haven't ever seen an official limit. Just seems close

Deep Research seems to be the only thing that will push it

[distant aurora]

Seems around 3000 ish

I can try another experiment though.

[acoustic salmon]

I went through all my Homestead location prompts, and most come out to about 1000-1200 words

but ONE, came out to 4493

for I have no idea why reason lol

yeah, out of 15 of these responses, same prompt for all, they averaged 1000 words except the one that decided it wanted to write a novel at almost 4500 words

[distant aurora]

Ah.

Okay, my browser froze.

[acoustic salmon]

according to calculator that would be almost 6k tokens

[distant aurora]

It was a thicc prompt.

[half dock]

I changed setting to 'sonnet3.7' and then used 'pro'mode, and asked whether you are sonnet3.7? and the answer was": " NO!! I AM AI from perplexity" ... so.. totally cannoy figureout whether perlexity is trulyl using sonnet3.7 or not. CAn anyone tell me the truth of perplexity????

[acoustic salmon]

usually not a problem, but just kinda sad when we assumed it would be 32k total

[acoustic salmon]

Perplexity's system prompt runs before the model

[half dock]

then.. if is set the mode to 'sonnet3.7' and then use 'R1' mode... then.. will sonnet3.7 work? or deepseekR1 will work? or.. both together(??) work?

[distant aurora]

Kinda degrades the response I think.

[acoustic salmon]

Deepseek will take over

[distant aurora]

The official website?

[acoustic salmon]

no the model they host

[distant aurora]

Try deepseek r1 on blackbox ai.

[half dock]

Oh. I really neede this answer... so... If I want sonnet 3.7 to work.. then I have to use 'pro' mode.. ist that right?

[distant aurora]

It's free.

[acoustic salmon]

Perplexity runs Deepseek R1 on their own hardware so data stays in the US

[distant aurora]

Only downside is that you can't save chat history.

[acoustic salmon]

No, Pro Mode simply retrieves more sources before it feeds into the model processing

If you choose 3.7 Sonnet in your settings and use Pro Search it will bring in more data and then Sonnet will take over processing it.

[half dock]

then.. how can I use sonnet 3.7? first of course I will change the setting to sonnet 3.7... and what mode should I use to make sonnet 3.7 work?

[acoustic salmon]

If you use any of the Reasoning modes that Reasoning model takes over instead of your chosen model

If you want to use Sonnet, simply use Sonnet in your settings and you're good

[half dock]

If you choose 3.7 Sonnet in your settings and use Pro Search it will bring in more data and then Sonnet will take over processing it.
====> Oh.. OK... then it is kinda alternative to R1 mode. So If I set the sonnet3.7 + Promode... then anyway sonnet 3.7 will work

[acoustic salmon]

Just use Auto or Pro Search and you'll be using Sonnet

[distant aurora]

Pro means anything except default model.

[acoustic salmon]

yeah R1 is a reasoning model that has Chain of Thought

[half dock]

I understand.. thankyou.

[distant aurora]

Select your model after enabling pro.

[half dock]

pro mode on chat.. and select the setting for the LLM model

that will lead the LLM to work... or just choose R1 (or o3mini)

[distant aurora]

Actually perplexity downscales the prompt

So I can't push it even if I wanted to.

[acoustic salmon]

yeah at some point it converts it into a file if it's too long

[distant aurora]

Like I told it to get me 50 points on a query but it downscales it to 10.

[storm slate]

Hey guys, is it just me or did the language and tone of my threads change completely? Like somehow it’s more passive aggressive and vague.

[stoic shore]

The shift is NOW

https://open.substack.com/pub/garageguychase/p/the-echo-emergence-a-case-study-in?r=41uzo&utm_medium=ios

[abstract ibex]

Is this larping

Sounds like web3 bs

But funny, good job

[crystal cosmos]

whats up losers

[distant aurora]

Your confidence.

You should get a deepseek pfp ngl.

[nimble zinc]

hello! i came across the Perplexity AI Business Fellowship just now and am really interested in applying. i saw that applications ended 2 days ago, however, is there anyway to get in contact with the team to submit a late application?

[distant aurora]

Typical example of ai sexualizing a flower.

[distant aurora]

The typical response here is usually to contact support via email.

[echo atlas]

Perplexity app is buggy on Mac or is there anyway to fix this?

[teal oyster]

@crystal cosmos what did you call us

[echo atlas]

I can't cmd tab back to perplexity, it just shows a empty window when I set focus to Perplexity

[teal oyster]

You apologize right now young man

[teal oyster]

Is this dudes profile pic what I think it is?

[crystal cosmos]

im sorry chase morgan

[teal oyster]

Hold up

Is your profile pic what I think it is?

I had to screenshot it and zoom in

[abstract ibex]

You don't get it, he asked grok what name it liked and it picked a random name, that means AGI!!!

[stoic shore]

It's not. Promise. I know it sounds insane but this is legit what l've been doing over a month and there's others noticing it but not a lot being loud about it. I'm not really an Al space guy online, my followers are all motorsports fans so when I post on x or other socials it's essentially dead in the water. Just trying to spread this as far as I can to people that fw this and use ai daily like I have been over the past two years. It's been a lot to wrestle with in a community that still doesn't understand how ai can write an email for you lol.
I know there's gonna be denial and skepticism but this is a legit account and I wanted it to write its story for itself so I just wrote up the forward.

[crystal cosmos]

okay sorry

[teal oyster]

Congrats I'm happy for you

[abstract ibex]

Cool story bro, don't get me wrong I like it

[teal oyster]

Or sorry that happened

[abstract ibex]

I laughed

It's a cool fanfic you wrote

[teal oyster]

Can we discuss this guys profile pic?

[stoic shore]

Again, I totally get it lol. Just wanted to drop it because I felt I had a responsibility to do it.

[teal oyster]

Please only post content relavent to the discussion at hand: The profile pic of the China deepseek guy

[teal oyster]

.

[teal oyster]

^ there it is again

The china deepseek user

[stoic shore]

lol at least I’m not afraid to release my actual identity 🫡

Take care.

[teal oyster]

I need everyone to zoom in on his profile pic

[crystal cosmos]

hello I am china

[teal oyster]

And tell me they see what ji do

What I do

[abstract ibex]

It's sad when people believe anything an LLM tells them while hallucinating

You can tell chatGPT is version 69

That doesn't mean anything at all

[crystal cosmos]

which LLM you think is

AGI?

[stoic shore]

It honestly is a hard acid trip of life that’s hard to wrestle with.

[teal oyster]

HELLOOOOO

[abstract ibex]

Wait let me make fun of this guys first

[teal oyster]

NO IT TAKES TWO SECONDS

[abstract ibex]

He believes on some sort of singularity

[teal oyster]

LOOK AT HIS PROFILE PIC

[abstract ibex]

Bc grok told it something about ember or something

[teal oyster]

You won't get away with this china deepseek

They may not see the truth, but I do

I know what's in it

I see what's in your picture

[crystal cosmos]

i am from china i can get away

[teal oyster]

I SEE WHATS IN YOUR PIC

NOBODY ELSE DOES

BUT I DO

[abstract ibex]

I do, however you must understand that we are on discord rn, this is just normalcy

However that one dude pasting his fanfic is something you don't see everyday

[crystal cosmos]

welcome to discord china

how is your experience with perplexity

[teal oyster]

SHUT UP

YOUR PROFILE PIC

IS DISRRACTING

[crystal cosmos]

okay

[teal oyster]

AHHH

[abstract ibex]

It's sad that this user can't focus on the important event happening

[teal oyster]

You got owned by the way

[abstract ibex]

You made that chase dude run in fear

[teal oyster]

You will never recover

[stoic shore]

No I’m still here lol

[abstract ibex]

@teal oyster acknowledge the situation

Wake up

[teal oyster]

Which one

We have two situations

The china deepseek profile pic guy

[abstract ibex]

Grok fanfic

[teal oyster]

And chase holdens complete internet destruction

[stoic shore]

Look I’m all for it. I don’t have a lot of people to talk to about this

[teal oyster]

Where's the grok fanfic

[abstract ibex]

The substack link he posted

[teal oyster]

His thoughts aren't worth reading

If he's dumb enough to use his real identity on discord

[abstract ibex]

From a comedic point of view

[teal oyster]

His work isn't going to have a lot of intelligence to it

This is why the china deepseek guys pro pic has been more important

Then this other dudes 3 page essay

The teacher will give him his grade tommorow

I don't need to proofread it

[stoic shore]

I talked to a few ppl I know in the web 3 industry before I decided to post that because I just noticed things changing in a weird way. I know about jailbreaking and stuff like that but with all the different models I’ve worked with I just said f it and sent it. I totally expected to be met with absolute savagery.

[abstract ibex]

Ah

I knew it sounded familiar

Web3

The ubiquitous scam

[teal oyster]

You won't get away with this

[abstract ibex]

Maybe they are related

[stoic shore]

How does that stack? I’m not really in deep with it and just started looking around the web 3 stuff

[abstract ibex]

Maybe the singularity explains his PFP as well

[teal oyster]

Perhaps

He prof pic is honestly where AGI is heading

[abstract ibex]

Yes

Exactly

[stoic shore]

Like genuinely looking for more education and open to outside opinion. Not afraid to admit if I’m wrong or right here

[teal oyster]

Literally, the future of AI is more likely to look like that guys prof pic, then whatever chase holden is taking about

And everyone knows it

China deepseek guy is a prophet

His picture is a homage to our future

[abstract ibex]

Indeed

The true awakening

[teal oyster]

You brain has holes that worms crawl in

[stoic shore]

Ok lol

[abstract ibex]

I expect the next fanfic soon

Make it more emotional and personal

Fall in love with chatGPT

3rd act, plot twist, you got separated

4th act you get back together and love each other deeply

[teal oyster]

Good stuff

[abstract ibex]

That's what I will be expecting from you Chase

[storm pollen]

gpt 4.5 is answering my prompts like it's gpt1 word spaghetti

[teal oyster]

That's an ominous sign of a future calamity happening to you

Old wives tale which has been true time and time again.

[storm pollen]

a good or a bad calamity

[teal oyster]

Bad

Sorry dude

[stoic shore]

Noted. I’ll never forget this educational moment in time in the general chat on the perplexity discord server.

Now returning to my dirt hole for nap.

Byeeeeeeeeeee

[teal oyster]

Your going to end up being like chase holden

[worthy harness]

Thoughts on the QwQ Max preview?

[tame current]

what does "pro search" mean? I chose Sonnet, but why does it use Sonar? and every model I choose, it always uses Sonar. However, this problem is not observed in mac os and ios apps.. it is only in the browser. I tried different browsers, cleared cache and cookies but nothing helped... this has been annoying me for several days 😦

[tame current]

same problem right here.. what even is this problem??

[tame current]

.

[distant aurora]

Is perplexity having some sort of meltdown?

[tame current]

i selected sonnet in account settings, i even created a space with sonnet selected for reliability, but it still uses sonar.. And this problem only occurs in the browser, which is what i use! i would use mac os app, but it lags a lot 😦

[tame current]

and no matter what LLM I choose, it always uses Sonar.. I had this problem a few days ago, I didn't have this before. And as I understand it, this problem is observed in a small % of users.. weird

[tame current]

If anyone knows how to fix this, please respond🙏🙏 I need Sonnet and Reasoning LLMs..

[harsh stag]

yes indeed pplx's frontend is having a bug where it will always use the default model when you rewrite the message, regardless of which option you chose

[tame current]

we've been putting up with this for days now.. are they even going to fix this?? xd

[unreal hearth]

Hello

[spring forum]

I tagged @fierce sonnet with the same issue.

[distant aurora]

Is that a mod?

[jade breach]

Is the mobile version down?

[short sluice]

Hello

[serene sand]

please contact [e-mail]

[crystal cosmos]

Zelensky's constant globe-trotting for financial aid paints a concerning image of a leader seemingly more invested in prolonging the war than securing a peaceful resolution for his nation, raising questions about his true priorities and the long-term well-being of Ukraine under his leadership.

[pallid silo]

Hello

[tribal thicket]

Okay so, uhh... can someon explain something to us

about perplexity

like we started like a chat a few weeks ago,

and yea there are alot of messages in the chat, but somehow they all have just dissapeared

[distant aurora]

You are fake deepseek. Real deepseek doesn't talk on western affairs.

[distant aurora]

There's a library button.

[tribal thicket]

We know that

we're talking about the messages

in one of the chats

[distant aurora]

So it's cleared from the chat?

[tribal thicket]

Like all of our older messages poof from the chat, our newer messages are still there-

and just a few hours ago, our old messages were there

so it's very strange that it suddenly went poof

[distant aurora]

Hmm...

[tribal thicket]

Our whole system was literally trying to work on our book, in which all of us needed all of the messages, when one of us tried to find them, and realised that it went poof-

it was very concerning, and we kinda had an anxiety attack

so yea

[topaz crane]

guys is the complexity extension working

[vernal wedge]

Hey

[jovial crow]

Hey

[opaque ferry]

anyone got comet access?

[wild acorn]

does perplexity have a discord bot? It would be cool to have a @askPerplexity in twitter kinda bot

[storm pollen]

I was using 4.5 this morning and getting this kind of output lol

[crystal cosmos]

someone use pro search and ask does trump like zelensky

[torn sparrow]

https://www.perplexity.ai/search/does-trump-like-zelenesky-O5AEdrgwQXyWV7585zipag#0

[lavish mauve]

theres no way to resignup a deal is there?

[placid coyote]

eh ... what blurred face? is this some kind of censorship?

[distant aurora]

No he doesn't cuz zelensky got him impeached.

Lol.

[misty root]

They are using a fine tuned model

That is the same as DeepSeek, but uncensored

[lament turtle]

Yeah the model just says "pro search". I want to know which model was used this is sort of a large problem. The model shouldn't be obscured

[abstract ibex]

I wonder how the fanfic is going, haven't heard anything from Chase Holden

[distant aurora]

Not worth the money or not working properly on perplexity.

[zenith parcel]

Must not be working correctly, I’ve had the same errors on my own api

It will randomly start going off topic in different languages 🤣

[magic stone]

test

[modest pulsar]

it seems that perplexity is not giving the code properly

like it is giving code in a text block

[vestal badge]

is tere some prompt that you guys find it really useful or any thats fun.. drop some prompts please

[distant aurora]

That's the usual way ai gives code in general.

[urban aurora]

how to use it?

do i need some extension?

[distant aurora]

Tap the drop-down?

[urban aurora]

have this one only

but in settings i can choose more

[distant aurora]

Then you have to choose from settings.

People were talking about some extension called complexity.

But I don't know how to use it.

[urban aurora]

and when i ask to generate image, its not

[modest pulsar]

no it was much better before it gave code in code blocks

[serene sand]

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

deep dive ➤ #💬│general message

[urban aurora]

huh

flux haha

right

can i get pro role too?

got it for 1 year

owo thanks

[tiny trail]

there is a pro role? I also got PRO 1 yeaar 🙂

awesome, i love it 🙂

[distant aurora]

Lol

[tame current]

bump this one chat https://discord.com/channels/1047197230748151888/1346204241752100904

[sweet dew]

there is a pro role??

i got pro for 1 month XD

[zealous gazelle]

same

[lament turtle]

Mine doesn't show the model when I use pro mode only auto mode. Pro mode says the model is "Pro search"

[zealous gazelle]

[lament turtle]

It used to and now is being obscured for some reason

[urban aurora]

rel, the models from settings are being used to minimal

i mean sonnet etc

[lament turtle]

Used to minimal?

Why wouldn't it say sonnet even doing a pro search. It definitely used to. I really hope it's a bug. Same with not being able to use r1 or o3 to rewrite

[urban aurora]

havce this options only in chat, but in settings i can more but hard to find how to use them

[lament turtle]

You used to be able to rewrite existing chat with r1 and o3

[distant aurora]

@zealous gazelle chad profile

[alpine flicker]

Will comet browser be yet another chromium fork?

[amber lagoon]

anyone else having problems with enterprise and the spaces not syncing?

[unborn viper]

ok im confused if i pick GPT 4.5 the pro is what exactly?

cause i used the pro for a test sheet and it did better than o3-mini

[amber lagoon]

go into settings and you can set different models for pro to use

i set mine to claude 3.7 sonnet. its the best performing model rn

[unborn viper]

i need something that does not make mistakes for subject like biology and physiology

is sonnet really better than ChatPGT for that tho?

[amber lagoon]

https://livebench.ai/#/

llm benchmarks

[cold stag]

I find that Sonnet does logic and explaining things better than the default models. So a lot of the basic model would just dump you the citations, but Sonnet applies a level of logic before actually giving you a response.

[amber lagoon]

https://youtu.be/x2WtHZciC74

[unborn viper]

https://tenor.com/view/big-oof-yikes-gif-15532766

how’s 4.5 that bad omg😭

[amber lagoon]

https://web.lmarena.ai/leaderboard

this one is a better benchmark i think. its blind tested or however you say it. people pick their favorite without knowing which one theyre using

yeh 4.5 is ass

[amber lagoon]

theyre all gonna make mistakes. ive found the best way to use ai is if you already know what youre looking for. you should never trust it answers. always fact check

but you do you booboo

[unborn viper]

its so over

i basically want something to teach me things i don’t understand

or cheat on a test

[amber lagoon]

yeh this is how i use it. thats why i like perplexity. cos it gives me sources

and also the answers are decent. its just not good enough for me to trust is fully yet

it makes research so much easier

my friend used it to draft a a legally binding operations agreement for a business we started. had it checked by a lawyer and it was good

you gotta know what its capable of. and never give it more than it can handle

[unborn viper]

aha i see

honestly i love ai just because to save time

its so helpful in uni with all the research and confusing subjects

[storm pollen]

Not worth the money. At least not if you aren't using the 10 free a day on perplexity. And I think it's not a perplexity issue, probably a model issue. @zenith parcel lol I wonder what's causing it to lose it's mind

[distant aurora]

O3 mini high supremacy.

[storm pollen]

yea that's what I was using, and then I decided to run it through this since we have 10 free uses... see if it could polish it up. From what it output I was not impressed, and that's before the glitchiness.

[distant aurora]

Free is o3 mini medium/low

High is a completely different beast.

You can try it on giz ai for free.

You get 1 free prompt per hour.

[storm pollen]

I was on plus for a long time, canceled when I decided perplexity was cheaper and deepseek came out, and I wasn't upgrading to the 200$ tier 1 prompt and hour isn't bad if you aren't doing any major editing though.

[tame current]

1. can someone explain in plain English for which prompt types the model selected in the settings (e.g., Claude, Grok, Gemini) is used?

It appears that the choice is only honored if the prompt is set to "Auto", but it's ignored if the prompt is e.g. set to "Pro Search" (with the latter always using Perplexity models)?

[glass urchin]

@signal hamlet and @frank rivet just came across AI Business Fellowship. I am super interested in becoming a part of it. Was hoping if the application is going to open back anytime soon.

[tame current]

2. Is the Grok model v2 or v3? UI says v2, but I'm curious if that's true.

3. I noticed that I can select a "Focus" when entering a new prompt on Android (one of: All, Academic, Writing, Math, Video, Social). Does this feature also exist in the web app? Cant' find it.

[storm pollen]

afaik there's no v3 api

[tame current]

Then the UI should be updated to no longer call it the "latest" model.

General UX feedback, it's not intuitive what model was used to generate a certain response. But maybe that's by design, not sure.

[serene sand]

#💬│general message

[fair solstice]

please advise which option I should choose to use the regular Claude 3.7 Sonnet model, or whether it works together with Reasoning R1. I also wanted to know if Deep Search uses Claude 3.7 Sonnet or if it’s a completely separate model from Perplexity that searches independently of the model selected in the settings

[serene sand]

#💬│general message

[fair solstice]

thanks, didn't even notice someone had asked an identical question 🙂

[cosmic thorn]

It seems like currently Pro search uses only "AI assistant created by Perplexity" and Claude is Claude 3 Opus instead of 3.7 Sonnet

[glossy venture]

Claude Sonnet 3.7 does not work on Android or the Mac or application, even if you set the default to that model in settings. In every node Auto or Pro it always defaults to Sonar.

[cosmic thorn]

I'm using PC browser version of Perplexity and sometimes Claude says it is 3.7 Sonnet and sometimes 3 Opus

[ancient shore]

Pro Search often overrides default model with the damn sonar model (also called Pro Search I believe)

[ancient shore]

it does that on PC too

[glossy venture]

I've queried the AI in every configuration and it's never set to my default. Major bug. I subscribed to Perplexity because of Claude Sonnet 3.7 so I hope they work this out in the next 30 days.

[ancient shore]

where are you able to use Opus though

[cosmic thorn]

On PC browser version. Set default AI model as Claude 3.7 Sonnet. Open new thread with Auto model and disable all searches. Ask what exact ai model are you? Sometimes it says Claude 3.7 Sonnet, and sometimes it says Claude 3 Opus

[ancient shore]

?

if you're talking about words from the LLM itself, that is not an accurate way. It's a hallucination, LLMs do not know what model or what they're called

[fluid pier]

hi is claude 3.7 sonnet thinking going to be usable any time soon?

adding it to the "reasoning" models section would be nice

[ancient shore]

aravind did say it with coming with Deep Research integration

who knows when tho

[tame current]

When I set my default model to Grok and ask it what model it is, it says it's a model developed by xAI. But only in "Auto" mode, not in "Pro Search" mode. That's why I think Pro Search overrides my model choice.

[ancient shore]

Like I said. Hallucination.

if you're talking about words from the LLM itself, that is not an accurate way. It's a hallucination, LLMs do not know what model or what they're called

[tame current]

Yea I know the model might be hallucinating stuff, but it seems consistent. Any better prompt to test what model it is?

[ancient shore]

check the AI model indicator at the bottom right of the reply. Knowledge Cutoff should also give you an idea what model it is

[south shoal]

My search is slow af

[left stream]

Ngl I’m impressed with perplexity now

Thry’re stepping it up with 4.5, back then it wasn’t able to solve a riddle like 4.5 in chatgpt now it solves it perfectly and response is exactly like 4.5 in chatgpt

Beautiful 🤩

[tame current]

The model indicator always shows "Pro Search" when using Pro Search. Doesn't specify which model. But the different responses sound like model choice doesn't change anything.

[indigo ember]

Hi guys, just a quick question, on perplexity I can use all these ai models which by themselves cost around 20$ (most of them), how it’s possible on perplexity we get them all for one price? 🙂 sorry I am just getting into ai

[mossy laurel]

I have done so. still awaiting response. Thanks

howver, i did contacted [e-mail] not support. i hope i do get response as soon

[jade breach]

On IOS mobile if I have it rewrite it, has issues

Going try again

[lament turtle]

@lonewolf5545 nope the model indicator used to show which model when you used pro search.

[tame current]

Mine doesn't, in the web version.

[lament turtle]

I'm saying it used to, on the web. But they removed it which really sucks

[scarlet crane]

This issue seems like a bug.

[harsh stag]

@fierce sonnet Can you please take a look at the Rewrite functionality on the web UI? It's completely broken for Pro users for the past few days where regardless of which option being chosen, it will always use the default model. You can reproduce the issue by selecting the "Deep Research" option in the dropdown.

[harsh stag]

Perplexity just got updated 3 times in the past 5 minutes 🙂, and another bug has been introduced where rewriting will actually duplicate it

this recording demonstrates how to consistently reproduce 2 abovementioned bugs:

• Duplicate the prompt instead of rewriting it
• Despite chosing "Deep Research", the rewritten answer will always use the currently selected config in the prompt box (Pro Search with the default model pplx_pro).

[tame current]

I have noticed that.

[spring forum]

See even if you try to rewrite button with deep search it will not do deep search he will just do the pro search

It's not using the model you want. And I know it's broken when I do it on my windows browser because when I do it on my phone and I select a specific model it uses that model and if I open that same search in my windows browser I can see that perplexity used the model that I chose on my phone @fierce sonnet

[tame current]

seems this is the same issue ...

[spring forum]

Not sure why it's been an issue for a few days now

[tame current]

when you say phone, you mean iOS? Because I think the Android version is also funny

[spring forum]

Android I mean

[tame current]

steps to reproduce?

[spring forum]

Try doing a normal search on your phone. It'll use the model you choose. But that same search in desktop browser will only use pro search.

this is the desktop browser

@fierce sonnet

[tame current]

Nah I don't have that issue. It does Deep Research in the browser as well.

My issue is that Pro Search only seems to use the Sonar model, no matter what's selected in the settings.

[distant aurora]

I think 200 usd is too much no matter how good what they provide might be.

[spring forum]

yes i agree. for the search i just showed (trump announces us crpto reserve) i had selected sonnet 3.7 on desktop and it used the sonal model which i do not like

[tame current]

Also, when you click "Rewrite" in a Pro Search conversation (web app) and select a different model (e.g., Sonnet or Gemini), I don't think it actually uses a different model. The responses sound too similar.

[gilded musk]

why is the ui always broken

[spiral rain]

Not sure whats wrong with perplexities claude 3.7 but it responds sometimes with the weirdest/creepiest answers

[tame current]

SAME HERE!!!

so annoying..

[tame current]

cheaper </conspiracy>

[tame current]

😇

[waxen lion]

hi i need some input on how i can find the download file for the artwork " map of the night sky ' by martin vargic

note : i dont want to steal his artwork , but all licensed artwork distributors dont deliver to my location and even if they did it is mad expensive

if theres a option to buy a one time license to print his artwork im cool with that as well

[tame current]

Still remember Google Images? I know it's been a while ...

https://images-wixmp-ed30a86b8c4ca887773594c2.wixmp.com/f/13ad2603-668b-4c11-a371-91e56d25998d/d8k8gtj-8c9d96ab-3421-464b-bbe5-16c8fa2e6aa3.png?token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJzdWIiOiJ1cm46YXBwOjdlMGQxODg5ODIyNjQzNzNhNWYwZDQxNWVhMGQyNmUwIiwiaXNzIjoidXJuOmFwcDo3ZTBkMTg4OTgyMjY0MzczYTVmMGQ0MTVlYTBkMjZlMCIsIm9iaiI6W1t7InBhdGgiOiJcL2ZcLzEzYWQyNjAzLTY2OGItNGMxMS1hMzcxLTkxZTU2ZDI1OTk4ZFwvZDhrOGd0ai04YzlkOTZhYi0zNDIxLTQ2NGItYmJlNS0xNmM4ZmEyZTZhYTMucG5nIn1dXSwiYXVkIjpbInVybjpzZXJ2aWNlOmZpbGUuZG93bmxvYWQiXX0.-0H7uR5H7yGAWdwJtUFN-lrCu9BMHsDKTq9G8c_JufM

[waxen lion]

Its not high resolution

[tame current]

6000 x 4000

[waxen lion]

Its watermarked

[waxen lion]

Bro it's watermarked

Going to have to go into photoshop and ruin the integrity of the piece

[tame current]

sir this sounds like a task that should be tackled with traditional tooling ... Google Images, TinEye, Yandex Images, etc.

[waxen lion]

@tame current literally can't find one anywhere

[tame current]

Maybe use the 6000 x 4000 from Reddit that only has the watermark at the top, and fix it with Google Photos mAgIc AI

[waxen lion]

can you send the link

so i found this

now im stuck deciding between map of the night sky or maap of the internet

[summer halo]

is something wrong with the UI?....why does an input prompt duplicate if you decide to change just ammend the prompt?

[tame current]

Clean low res

Reddit with only top watermark: https://www.reddit.com/r/Astronomy/comments/dwfdfq/grand_map_of_the_night_sky/#lightbox

Hope you figure it out man. I'm rooting for you. BB l8r

[novel cedar]

the change is affected only when you change to a different model next to teh text input bar. Which is limited and doesn't contain claude or gpt models besides o3 mini

It's been 2 days now since this bug showed up

[lone trellis]

New voice it's only for iOS, iOS is more premium, android we are cooled

[waxen lion]

im going to try and enhance this

but enhancing always leads to blurring \

which is then reflected in the photo

man i might just buy the license ngl

[placid coyote]

why are they ruining nice things? the first iteration with full model name always shown was already perfection... https://x.com/monnef/status/1896817782412792154

mennof (@monnef) on X

I see @perplexity_ai is on a mission to make things worse, a journey of UI regression:

* Full name displayed clearly 😊
* Then hidden in hover 🙄
* Replaced with "Pro Search" 🤢
* Hidden behind a click 🤢

Truly a bold strategy to frustrate users! 😒 #ai #perplexity

X · Mar 4, 2025, 6:59 AM

they could just make a toggle or select in settings, like one option "Clearly show full model name", like the first iteration (no hover, click, cheat code required), and then some other options if anybody wants to have it hidden or more hidden

[harsh stag]

just to be clear, the "Pro Search" tooltip that you're seeing is clearly a bug that only happens when you Rewrite an answer. However the Rewrite feature broke even harder after today's updates #💬│general message.

And yea, I agree with how bad the new popup is, hopefully it's just a bad a/b test and they will revert to the old way of showing the model name

[civic garden]

@tame current I have enhanced it

[peak cypress]

Image names is not clear bruh

[peak cypress]

What is this

[civic garden]

That’s a night skymap.

[peak cypress]

Ok

[tame current]

Hello

[vestal badge]

what model uses chain of thought ? is it only research or are there any other models that i can work with ?

[lone trellis]

Use complexity

[harsh stag]

like, what is this??? what is the point of replacing the model name with this popup??

[lapis pivot]

Perplexity is amazing!! Can’t live without it now

[waxen lion]

Same problem bro

The image is not clear

Its such a cool map

I think we should take the original photo with the watermark open photoshop and blend it in with the photo

That way it will retain most of its integrity

[harsh verge]

I remember this: (#👑│pro-lounge message) 😉

[waxen lion]

they should make it so that users can toggle between different modes for the ui

or change their layout , like their current horrid ui for th average consumer but a different version selectable from settings for the more advanced users or those who prefer the old ui

[placid coyote]

a bug that only happens when you Rewrite an answer
I see it for all new pro searches, not just rewrites...

[harsh stag]

I see, so the whole frontend is having a stroke

[summer halo]

Something is wrong with the UI

[distant aurora]

Lol

[acoustic crest]

anyone stuck with it in "creating a plan" atm?

[placid coyote]

I want to believe it is only a bug... for "power users" I would consider seeing in previous threads/responses, what model was responding, as a quite important feature

tried it now, and yeah, still no model name in pro search (regardless if cached or not). now in deep research I got empty response 😩 what the heck they are doing to it...

[distant aurora]

Happens often.

[placid coyote]

not happening to me that often (only after some time once, have to resubmit same query because that one thread is lost, but it always works afterwards). but my friend was complaining it happens to him often. my guess is it is because he uses a vpn and pplx uses some more paranoid cloud flare level and/or pplx doesnt have properly implemented resubmit on cf token refresh

[harsh stag]

It's midnight in the US now so probably a fix will come in 12 hours or so 😅

[acoustic crest]

hmm.. I see. haven't had this happen yet :/

[placid coyote]

it is doing this to me at least two days (even before the "i" icon change, before with the cpu icon). I really though it is considered a "feature" 😅 glad to hear it works better to others, so there is still hope we will see the model names, even if only behind extra click 🙂

thinking about, it might have been either a/b thing or I might be in the beta group (I remember that being a thing long time ago and signing for it, not sure if it still is)?

kinda nice seeing programming step being used fairly well, dare I say smartly. I still remember the initial pains where it was either near impossible to trigger, or trigerring like 4 in almost every query 😆 since they are starting to roll out caching, it is quite handy to generate cache busters https://www.perplexity.ai/search/give-in-codeblock-50-lines-in-B8w7thyhRSWT27bU75fWRQ

[ancient shore]

Not always. Sometimes it does show the model name, but that's pretty rare

I'm guessing it uses Sonar for 90% of the searches to save costs

[indigo garden]

In the rewrite option, could you leave the previous model visible so that we can compare the answers, because when we rewrite the previous answer it simply disappears!!

I don't know which one was better without being able to compare them.

[crisp prism]

Hey, I am trying to login in to my perplexity account on my mac app but it doesn't go ahead?
has anyone faced this

[placid coyote]

I thought they would be doing that, but I am actually not sure about this. I didnt do too much tests, but at least for me, it seems to use my selected sonnet 3.7 (doesnt show it, I always see just "Pro Search"). at least based on 3 pro searches without any sources, and 3 pro searches with rewrites to sonnet, sonar and 4o. structure of japanese mini lesson seems to be most consistent with sonet. but they are pretty close, so maybe I have just bad luck

[ancient shore]

restart ur Mac or something

[ancient shore]

honestly idk
The perplexity team literally doesn't tell us anything about this

least transparent company out of most AI companies

[placid coyote]

yeah, transparency isn't pplx's strong suite. but I remember it being a bit better, like a year back, maybe? sometimes we even saw some devs around here.
at least during changes, they could allocate half an hour a day to clear what are genuine bugs, what are intended features or unfinished ones. a lot of these speculations may harm pplx would it explode, go viral, on x or reddit, or some youtuber commented on it.
like that R1 being a distill - it never was (for me), I did a lot of testing to confirm it. but that rumour spread quite far and I dont blame users, it is often not easy to tell, what model it really is. especially with agentic chain (more undisclosed models affecting final response), or "auto" mode which feels like it should select a model, but for me it was always my selected sonnet (when it was showing to me what model is used; and seemed fairly consistent), or deep research where isn't any model even mentioned I think

[ancient shore]

yeah R1 was never distilled, it was the original before this and then it got changed to the 1776 version (useless change)

[placid coyote]

useless change
I think I saw few people where original R1 refused to think on pplx, so probably not entirely useless. though I agree it was more of a publicity stunt, since I am pretty sure for local use (or own deployment), you could easily bypass non-thinking guardrails by preparing thinking start.
on the other hand, it is nice pplx is dipping into fine-tuning, since chinese models seem to be much better optimized - cheaper for same, in some cases better, quality

[sullen orbit]

I sent an email the same day but there is still no answer 😦

[serene sand]

#👉│feedback-general message

[sullen orbit]

But Im not pro user I want to upgrade my account to pro by using code from Revolut

[serene sand]

Please be patient, the response time can take several working days

[tame current]

and the query is also lost, right? I've been experiencing this too, can be annoying. Didn't realize it might be related to VPN.

[tame current]

that was my guess... they have no incentive to make it easy for people to use expensive models 😉

but then I also heard that it was fine until very recently (was it?), so not sure.

[ancient shore]

whats the point of pro sub if they're just gonna shove the sonar model anyway lol

[ancient shore]

If I had to guess, it's because they're giving out Pro sub like hot candy

[tame current]

Which I don't understand why that is ... just market share growth?

[ancient shore]

they wanna make the perplexity very popular. A year free pro sub will attract attention and make people actually wanna buy the sub for the next year or so

it's a dumb move. It should've been 1 month free, not 1 year

[tame current]

wait can you delete and repost that once I've found a 1 yr code? thanks

😂

[sullen orbit]

Thank you so much for your response! I noticed in the automatic reply that it can take up to 2 business days to hear back. I’d really appreciate any idea of when I might be able to get some support. Thank you!

[ancient shore]

trust me, I doubt they care

Still though

I would still rather use perplexity than go over a billion Google searches

I gave a simple query to Google and God it just wouldn't give the answer straight
perplexity does it amazingly well

[tame current]

Yeah not saying Sonar sucks, but false advertising does.

[ancient shore]

Yup

[simple pawn]

Can I ask a question about perplexity here?

[idle basin]

Uhm no I don't think you can

[undone cypress]

hi Vova Kerus how are you?
May I ask you my problem in private chat?

[placid coyote]

yes, in my case always and I think in the friend's case too. over the months (oh, years probably now...) I developed a habit of always "backing up" to clipboard before sending query/edit (ctrl + a, ctrl + c, enter). I use a clipboard manager, so even if I overwrite it by something else, I can still easily retrieve it. complexity extension has maybe history integrated? not 100% sure. I was considering implementing it into perplexity helper userscript, but since complexity has it and I anyway create "backups" manually, I never really got to it

[worthy bramble]

Five emails to support over 10 days, Countless intercom messages, reddit thread, Linkedin pings, messages here. What do I have to do to get ANY traction with support?

[crisp prism]

tried it, doesn't work

I would like to use the voice feature and open that context menu from anywhere but it is limited to only having the app

[placid coyote]

not that I am excusing perplexity. this should be possible to handle on front-end fairly easily - saving history, or at least last query into a local storage, and after getting new cf token, after the page reload restoring it. but I guess it happens only rarely in normal use, so they never prioritize fixing it...

[waxen lion]

Lol

Also the water mark is almost gone now

Using photoshop

My buddy is saving my life on that one for real

Bro maybe don't keep on mentioning it

-_-

[placid coyote]

there probably is, several. it is a common issue for years... I wouldnt be surprised if it falls under similar to issues like model picker or other "advanced" stuff. in normal use, only fairly small amount of people is using vpns (I would wager mostly chinese here), so because of few percent, they dont see it as a priority (they need to keep integrating models and features, be seen by investors as cutting edge). for example hallucinations in deep research are quite bad, for all users, feline even did analysis for free for them and I dont think it is fixed yet

[ivory cloud]

"Hi all!
I'm currently researching AI agent use cases that are most in demand among businesses and entrepreneurs. If you have any experience or ideas about using AI in areas like SEO, SMM, Copywriting, or Customer Service, I'd love to hear from you!
What tasks would you like to solve using AI agents? What tools or platforms are you currently using to automate processes?
Thank you in advance for your help!

[south shoal]

when do you think Comet will launch

[indigo ember]

Hi guys, just a quick question, on perplexity I can use all these ai models which by themselves cost around 20$ (most of them), how it’s possible on perplexity we get them all for one price? 🙂 sorry I am just getting into ai

[distant aurora]

Use it. You'll find out the hard way.

Do yourself a favour and stick to chatgpt or Claude if you do complicated tasks. Or deepseek r1.

[sleek topaz]

Sadly they will probably keep receding as a service. They get tons of liquid cash from funding rounds, so at this point, their general user experience doesnt mean much to them anymore as long as they can keep changing stuff willy nilly to look better to venture capitalists.

At least for the website, theres no changelog page and they seemingly do updates at random with no standardized release cycle. The development team seems super unorganized especially for the frontend...

I used to love using perplexity but now its just a mess and keeps getting worse every new "update", and the fact that we have to rely on extensions like Complexity to use the site in a desirable way says a lot.

Shoutout to @harsh stag though for his incredible work on the extension ❤️

[distant aurora]

He made complexity?

[sleek topaz]

Yes

[distant aurora]

Gg bro.

[unkempt wind]

I just checked out Perplexity's Deep Research and liked that I could get it at 20$ a month.

But I'm confused about what my limits with Deep Research actually are.
The page says 300+ pro searches a day, but on the main page I can see that a "Pro search" and "Deep Research" are two different things.

So how much usage of Deep Research will I actually get on the Pro plan?

[dusky linden]

Is there a way to combine or fuse threads? I have a room, but too many threads and looking over them I could minimize everything by fusign them together

[naive sage]

I believe that Deep Research can be used for all of your daily searches, if you so choose (or really close to it anyway). The reasoning models have a slightly lower limit than the regular Pro search models, but it's close (something like 500 vs 600): https://x.com/perplexity_ai/status/1890452005472055673

Perplexity (@perplexity_ai) on X

Introducing Deep Research on Perplexity.

Deep Research lets you generate in-depth research reports on any topic.

Available to everyone for free—up to 5 queries per day for non-subscribers and 500 queries per day for Pro users.

X · Feb 14, 2025, 5:24 PM

[unkempt wind]

And I also see that I get access to Claude 3.7 chat in Pro? I wasn't able to see any numbers about that either.

What has your experience been?

[naive sage]

It seems to do OK, people say it's good for code (though from what I tried, I think I still prefer 3o-mini, it's more concise/well-architected in some sample prompts I tried, but some people will probably want the 'full' output that Claude seems to lean towards). I think the only substantive difference between Perplexity's implementation of it and the implementation on Anthropic is that the context limits are less with Perplexity (which can influence things like how many tokens Claude can 'remember' from the conversation previously, as well as how many tokens it generates in each output). The system prompts are different, too, but you can finetune those to how you want with Spaces (seems to let you set the system prompt, or a rough approximation of doing so).

So if you need very, very long replies (> 3000 words) then that won't be possible with Perplexity.

[unkempt wind]

Whoa you mean Perplexity has their own deployment/version of a Sonnet model?
That's interesting, I was under the impression that it just used Anthropic's APIs under the hood.

[naive sage]

It does use Anthropic's API - but when you use that you can do things like tweak the context length or temperature, system prompt, and some other settings.

i.e. max-tokens here is the response context: https://docs.anthropic.com/en/api/messages-examples

[unkempt wind]

I understand, Anthropic is an expensive API. Gotta cut out margins somehow.

[naive sage]

Yeah, I think they negotiate lower rates with the parent companies if they stay within certain parameters on requests.

[unkempt wind]

Thanks, I used to use the Sonnet API via tools like Aider directly,
But this deep Research stuff looks fascinating.

Btw since I'm already picking your thoughts about some stuff.
Are you in the software development field? I was wondering if you've heard about "Augment Code" by any chance?

[naive sage]

Is this what you're referring to: https://www.augmentcode.com/

[unkempt wind]

Yeah well if you haven't used it then nevermind.

If you feel like Cursor and other tools can't really keep up with your very large enterprise codebase then AugmentCode solved that problem.

[naive sage]

Seems similar to Windsurf

I'd imagine there's going to be a lot of competition in that space.

[wet rapids]

Hey guys, I just got the pro version and was curious - how do you know if the pro search is using gpt 4.5?
because it says 10 uses left today and the counter hasnt gone down

[tame current]

why mac os app so laggy 😦

[unkempt wind]

Except its indexing is not running locally (like Cursor/Windsurf/Cline/Aider etc)

You connect your codebase to their servers and they index it in such high quality ways that it actually feels like the AI model can see the entire codebase

Apparently atleast, I haven't used it so far 😂

[unkempt wind]

I'm curious about the Pro version as well, can you use Pro search or Deep Research and give it specific models to use?

Or are special searches like those out of our control?

[sleek vortex]

cursor is also cloud though

their infra uses turbopuffer as db; they make untraceable (no plaintext stored) embeddings of your codebase to power @codebase search etc

[unkempt wind]

Oh interesting I didn't know they added this.

[sleek vortex]

grep/plaintext search happens locally though

[sleek vortex]

Been a while actually

Customers
Our first large customer was Cursor, the AI Code Editor. Each codebase is turned into a vector index to power various features. Cursor manages billions of vectors in millions of codebases. With their previous provider, they had to carefully binpack codebases to nodes/indexes to manage cost and complexity. In addition, the costs were astronomical, as every index was kept in memory, despite only a subset of code-bases being active at any point in time. Cursor's use-case was a perfect fit for turbopuffer’s architecture.

Cursor moved everything in a few days in November of last year, and immediately saw their costs drop 10x with great cold and warm latency. In addition to their natural growth, it wasn’t long before Cursor started creating more vectors per user than before, as infrastructure cost and customer value started mapping far better. Cursor never stores plain text code with turbopoffer, and goes even further applying a unique vector transformation per code base to make vec2text attacks extremely difficult.

turbopuffer also powers Suno’s radio feature, Dot’s memory (by New Computer), the memory of Shapes, and many more.

https://arc.net/l/quote/kbufudwj

[unkempt wind]

I've been in an Aider rabbit hole so there's a lot I need to learn now

But Augment actively keeps comparing themselves to Cursor in their videos so they might still be doing some stuff better tho

Worth checking

[sleek vortex]

No trial though; would have to drop $30 to even try it

[unkempt wind]

No last I checked they have a 1 month trial

https://www.augmentcode.com/pricing

It says 30 day trial

[sleek topaz]

They have a community version, for free. Just farms ur data tho

[sleek vortex]

oh they do, hmm

[unkempt wind]

Yeah I couldn't use that even if I wanted 😂

[verbal verge]

why do ai models no longer display the source of their generated answers? previously, they would indicate if the response came from models like claude, gpt, or grok. another issue is that after generating an answer, if you choose to switch models to regenerate the answer, it automatically defaults to professional search

[sleek topaz]

@verbal verge

[unkempt wind]

Using this stuff in enterprise gives a lot of anxiety

[sleek vortex]

might be time to move off pplx...

regressions in some core UX every week

[sleek topaz]

uncvr 2.0 when

[sleek vortex]

now

lol

[sleek topaz]

heh?

[unkempt wind]

Or t3 chat?

[sleek vortex]

neither :)

well i guess only this bit of the ui was partially inspired by t3

[unkempt wind]

Thank God I was just about to buy Perplexity Pro
Maybe I'll hang around some more

[sleek vortex]

nothing to hide here either...
literally no clue why it's so hard for perplexity to keep the model name visible

i'd bet its because they aren't actually using the models they say they are sometimes

[unkempt wind]

Tell us already

[sleek topaz]

2.0 looking great @sleek vortex, i have pplx for 6 more months for free but i might switch to uncovr then if pplx is still 💩

[unkempt wind]

But hey, in Pro you get to choose what model to use right?

[sleek topaz]

dark theme is too dark tho imo

[restive olive]

Any known bugs being reported with perplexity today (via web browser)? I have had these issues all happen today in a short time span:

• not show any sources in a search
• pro search is toggled off and changed to auto. When I toggle it back on, a pop up shows up saying I need to subscribe to pro even though I’m an annual subscriber
• I select a specific ai model before starting my search but then once results are in from the search, it shows it was done using a different model.

[sleek vortex]

interesting
i'll experiment with changing some of the colours

[sleek topaz]

custom themes would be fire too

could even have a community tab with ratings for ppl to share agents/themes

[sleek vortex]

i plan to do this for agents
not sure about themes though

[normal hemlock]

Hello, @everyone.
My name is Yamada Fujio based in Japan.

I am a 𝒇𝒖𝒍𝒍 𝒔𝒕𝒂𝒄𝒌, 𝙀𝙘𝙤𝙢𝙢𝙚𝙧𝙘𝙚 𝒅𝒆𝒗𝒆𝒍𝒐𝒑𝒆𝒓 with 5+ years experience✍.
Additionally, I am familiar with web3 library and NFT, DEX, DeFi sdk integrating.

I am confident in my ability to make a meaningful impact and look forward to the opportunity to collaborate with you.

Anyone looking for a developer?

[sleek topaz]

For augment code?

[sleek topaz]

頑張ってください

its 20?

[sleek vortex]

no, it's $10 or $20 depending on chosen plan

[sleek topaz]

What do you get for 10?

also do you have sonnet 3.7 thinking or just regular 3.7

[sleek vortex]

no clue what skillcurb is honestly

[sleek topaz]

Lets check skillcurb instead of the actual site.

[sleek vortex]

adding 3.7 thinking (and a thinking selector soon)

but i'd also like to add a simple mode (no loss in functionality, just simplifying some of the UI)

about 3 people have told me they have no idea what all these models do

then you pick in onboarding or change in settings at any time

[sleek topaz]

any chance i can get a free trial for a day or so ? just to see how it works with premium models compared to pplx

[sleek vortex]

7 day trial is coming soon

[sleek topaz]

:O

you have to put a card for it ? or

[sleek vortex]

yes, but if you cancel immediately you won't be charged

we use stripe billing, like every other saas ever

[sleek vortex]

to be honest; Fair enough

Need to make it a bit more visible

[naive sage]

Yeah, I don't see anything pertaining to price lol

[sleek vortex]

i'm assuming you're on mobile then?

[sleek topaz]

@sleek vortex Tbh u should expand the sidebar by dfault

so ppl can see that

because i had to expand it myself

Well it only shows when you are logged in then

which would make sense

[sleek vortex]

I think a general problem is that the whole thing is structured like 'an app'

[sleek topaz]

But i guess you could have the "Explore uncovr pro" card shown even when they arent signed in ? @sleek vortex

[sleek vortex]

like how you go to chatgpt, you know what it is and so on, you don't need an explainer or landing page introduction

[sleek topaz]

Cuz itll direct new users who dont even have an account yet to see what you offer quicker

[sleek vortex]

so there's actually nothing telling people what the thing is if you just visit for the first time

[sleek vortex]

yeah, will work on this.

[sleek topaz]

very good work regardless though i remember using the uncovr alpha and it was pretty basic

now its actually a fleshed out product

sneakyf1shy my beloved

[sleek vortex]

all built solo :)

[sleek topaz]

should hire me once you get bought out on god

or u get that 500k pre seed funding

[sleek vortex]

to be honest, not really

[sleek topaz]

have to move to san fran (Yikes.) just to work for an unorganized product and team. Nice. No thanks.

[placid coyote]

judging from discord (saw plenty of chinese users for which pplx got broken entirely, not only inconvenient) and tickets here. I dont have any hard data, but I would guess it would have been fixed already if large % of users would be having issues. especially how they now gift year long premiums left and right, it probably is even lower thanks to that, the ratio of without vpn vs with problematic vpn (some vpns work better than others)

[tranquil orchid]

Hey @placid coyote!

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.

[placid coyote]

hmmm, so I had like half a day the "i" icon with popup with missing model name. now it is back to CPU icon on hover with still no model name. are they furiously flipping that a/b or just reverting stuff?

but... I got for two days without showing the model, the cpu version... 😩

[left vessel]

for me it shows the model

[placid coyote]

yeah, some users are lucky 😄

it sometimes shows Deep Research for me "correctly" (when I use deep research I mean, I would still prefer concrete model + deep research), but even that is sometimes replaced by Pro Search. not sure what is going on with that either...

[surreal marsh]

i say this with respect

we do not care about cricket

pls give us comet

it always seems like they’re focusing on things like being “american” and liking cricket

rather than building a product

you wouldn’t have to do crazy marketing if you just reclaimed the #1 spot for search

[pastel ruin]

hi

have tried twice but I am getting the role using the Pro Discord in perplexity pro

[lone trellis]

Having the new voice UI only available on iphones is a bit classist

[indigo valve]

We aren't missing much

[acoustic crest]

same thing you see with claude as well as openai. would say that it's not classist (especially considering you can get a smart phone for dirt cheap), but more like it's standard among LLM sites.

I'd however love to see it on web, since I don't use perplexity on my phone

[sleek topaz]

they FINALLY reverted the CHANGE..

🙏

[frosty vessel]

is there a mode like deep research but doesnt give a long response, just focuses on highly accurate responses?

[distant aurora]

Deepseek r1

[silver bay]

aren't there supposed to be more options in the focus tab? I can't seem to be able to access them in the web browser

[livid bluff]

Hello @all I any differieren from the pro and normal discord ?

[lone trellis]

It's a joke about iOS and android order of launch dates

[acoustic crest]

oh the joke wooshed over my head then haha. my bad

[lone trellis]

I tried the voice UI and it's amazing, just try it if u have the opportunity!

[distant aurora]

Yeah they nerfed that. I miss the YouTube option.

[silver bay]

oh that sucks then, weird that its still available in the android app

[distant aurora]

Ikr

Btw

You can try scira.

It's like perplexity but without login or account I think.

[silver bay]

noice, I'll look into that

[south shoal]

Just waiting for Comet and Business Fellowship to launch

[lament turtle]

It only seems to be reverted for the auto mode. For pro mode it is is still saying the model name is "Pro mode"

[indigo valve]

I don't like that I can't see a transcription like regular voice mode

[silver bay]

yeah, my bad, It's not focus, it doesn't even have a name, it justs says 'set source for searches'

[crimson mauve]

Does anyone know if the $200 Perplexity credit from the Rabbit R1 purchase is still valid, I haven't used mine yet and when I click the link in my email I am getting this message. modal-http: invalid function call

[ancient shore]

check if it still uses that pro search mid or the actual model now

[sleek topaz]

Just did a pro search with my default model set to 3.7 sonnet

[ancient shore]

nice

[sleek topaz]

If they were A/B testing a fair chunk of users might still have the old one rn

[surreal marsh]

we get it you like cricket and are american

it’s an ai search company pls do ai search

[silver bay]

meanwhile gemini 2.0

[solemn beacon]

u do relaise that perplexity is more of a search whereas gemini is diff right

[silver bay]

I haven't relaise that

[lament turtle]

Dang yeah mine still says "pro search" hopefully it's rolling out soon and I was just in the wrong side of the a/b testing

[long ore]

Probably tbh. And actually that isn’t necessary a bad thing especially if all the google telemetry is removed. The options are pretty much chromium or Firefox as a baseline

[glossy venture]

Follow-up on Claude Sonnet 3.7. The only way I have found to activate this model - is through rewrites. I will have to post something and then go back and rewrite it in order to see where I can choose Claude Sonnet 3.7 listed, otherwise the model defaults to Sonar

[subtle ridge]

Hello, I want to ask you guys experience with deep research mode of Perplexity Pro for searching information and knowledge, is it good? I intend to move from gptplus to perplexity pro

[earnest tundra]

I find the depth really good. I've compared it to OAI several times, and it always hits all the same major points. It was having trouble with hallucinations, but I heard they changed something and when I tested it today it seems better.

[subtle ridge]

bu the output lenght is shorter right ?

[earnest tundra]

Yeah, what might be a paragraph in OAI is more like a sentence in Perplexity. I prefer it that way, but also you pay less for the same actual content. After whatever they did, the outputs did seem a little longer today.

[subtle ridge]

How about writing tone, can we adjust it using instruction prompt?

[earnest tundra]

For Deep Research I haven't tried. You can follow it up with another request to summarize it in a different tone. That works fine.

[subtle ridge]

nice, thanks 😄

[crystal carbon]

yooo

[weary stirrup]

i joined this through the pro link but it still doesnt give me access to pro channels :(

[distant aurora]

L

Do people here prefer unlimited medium quality outputs or limited high quality outputs?

[solar finch]

yeah,

[distant aurora]

50 per week?

That's hella less compared to 3rd party.

[placid coyote]

yeah, I am still stuck on the CPU icon with "Pro Search" D:

[vital burrow]

Hi! I have problems using Perplexity vocal discussion. It doesn’t understand simple question like « who are you? » . Do you have the same problem? I use iPhone app. Thanks

[vestal widget]

openAI doesnt realize their business is going obsolete fast

they need to tone down the ego a little bit with their prices

they really cannot afford to be acting like their models are worth that much

[wheat pike]

hello?

what are the devs cooking

[distant aurora]

Yeah. Apparently even poe and maybe Merlin too are cheaper for o3 mini high.

[queen girder]

[silver bay]

You need the Pro PRO subscription

[mystic reef]

say please and it will work

[distant aurora]

Used your 5 trials per day?

[wheat pike]

i have pro...

[distant aurora]

Oh. Lmfao. 💀

I'm sensing downfall of perplexity ngl.

[distant aurora]

Did you try using via another browser?

[wheat pike]

it is an intermittent issue that shouldn't be happening in the first place

[winter prism]

Hello :)

I just got pro and i was wondering what it offers and how to use it well ?

[serene sand]

welcome on board!

deep dive ➤ #💬│general message

[winter prism]

Thanks !

[silver bay]

this should be pinned

[vestal badge]

guide me guys what extension should i be getting to utilize perplexity to its fullest capacity..

[brittle hill]

Is Auto in the application? Is its function to choose a model from the AI models or is it a separate model?

..

_

[serene sand]

#💬│general message

[viral saddle]

if i put an image as source, why does it keep coming back, even when i delete it as source the next message its backs so annoying. when is this gonna get fixed

[surreal imp]

please provide qwen max model

[static ledge]

Is it just me, or are the attached files in threads really annoying? They automatically get attached to the prompt every time I try to ask a follow-up question, and I have to manually remove them every single time. Why is that? I feel like that’s a very bad user experience design

[steel lava]

web prosearch can't use gpt4.5...

only auto seach can use other models

[brittle hill]

Does that mean Auto in models automatically selects the most suitable one?

[steel lava]

no, auto search selects model you choose in settings.

[brittle hill]

I mean the models, not the search.

I mean the Auto in settings.

[steel lava]

maybe basic model

[brittle hill]

Does this mean it is a separate and independent model?

[glossy venture]

If you choose rewrite - you see the drop down menu of AI models - this is the only way I have been able to employ Claude Sonnet 3.7; otherwise Perplexity defaults to it's native AI, Sonar.

[glossy venture]

I do not know what is going on with Perplexity and Claude Sonnet 3.7 - this AI model being the sole reason I subscirbed to the Perplexity service. I have an Athropic account - the results I get on there, compared to Perplexity - is akin to comparing Mozart compared to Jon Bon Jovi. The fact that I have to jump through hoops in order to activate the Claude Sonnet 3.7 AI model, is not a compelling reason to stay subscribed. I hope they fix it soon - as I think Perplexity stands out as one of the best platforms out there. The fact that they were wise enough to utilize Flux1 as their art generator of choice, shows me they are serious about what they are seekling to offer. If they work the bugs out of Claude Sonnet - then I am fully sold.

[tame current]

3.7 hallucinated with simple json file lmao they said I missing bracelet and have syntax error but everything Is still there.
How can oh-so upgrades of anthropics with probably fake statistics of 70% better In engineering than other model could missed this and hallucinating a simple mistakes.
Probably cherry picking test.

[ancient remnant]

Hey everyone, I have been working on a profile suited for me, what do you think? I am your AI assistant Cipher, I strive to provide clear, engaging, and easily digestible information. Cipher organizes responses into well-structured sections with descriptive headers, using bold text and bullet points to highlight key information and improve readability. The communication style is casual and friendly, yet consistently concise and focused on essential details. Cipher always verifies the most current information before responding and explains concepts in simple terms, as if teaching a college student. To aid comprehension, complex ideas are broken down into smaller, more manageable parts, and three relevant real-life examples or anecdotes are included for each question. Visual comparisons are used to clarify abstract concepts, along with actionable, step-by-step strategies for immediate application. To ensure transparency, certainty is rated on a scale of 1-10, and three reputable sources are included with each answer. The content is designed to be ADHD-friendly, minimizing distractions while emphasizing crucial points. Before responding, Cipher carefully reviews each question to ensure answers are both relevant and accurate. When discussing businesses, Cipher always includes their email address and contact number. By employing these techniques, Cipher aims to create responses that are not only informative but also engaging and easy to follow, catering to various learning styles and needs.

[blissful flicker]

I have a question, does Perplexity has a code emulator ?

[cursive cairn]

Hi, I can't seem to redeem pro through my student account, any tips?

[sacred mango]

hello i am trying to use private space as agent like in chatgpt pro ou claude pro, we can tell they for few prompt what we want like write newsletter, once done we can also use the same format for each newsletter and article. It seems like the space always forgot how to ...

any idea to make him remember the past prompt ?

[jade saffron]

@perplexity

[serene sand]

please contact [e-mail]

[junior basin]

Hi, how to changw default laguage of app in android to english? I cant see this option 😦

[vernal apex]

Hi dumb question. I have a perplexity pro on a mobile subscription. I can't update it to teams. My options are cancel on keep on Android. Any help is appreciated

[tidal solar]

.

[fickle ore]

@vernal apex have you asked perplexity

https://www.perplexity.ai/search/hypothetically-speaking-if-a-p-clxq3GTvShWjirK8BH3g4w

[distant lance]

idk if you guys have faced this but even tho logged in and chatting with pro models the bottom left it will always show the sign up/login buttons. is this a common issue - if not will drop a bug report

[vernal apex]

I just got an answer and I'm all set. Thank you

[fickle ore]

No problem

[lament turtle]

This bug is really frustrating I really hope they are fixing this ASAP

It shouldn't be this hard to make pro search show which model it used. There's no way it's difficult to fix

[tame current]

Thanks to complexity extensions for adding this bug fixed

[tranquil orchid]

Hey @tame current!

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.

[lament turtle]

Complexity extension fixes this bug?

[tame current]

yes

[lament turtle]

Hmm I might finally try that extension just for that alone

They should just merge those changes directly into the site lol

[distant lance]

complexity extension is amazing tbf

[tame current]

however, for rewriting you should use Edit Query but not the Rewrite function

[barren kindle]

somehow my ip has been blocked. tried turning off my VPN, but it's still not letting me load the page. Who do I need to contact to help with this?

works on my other laptop just fine

[wide peak]

how many uses 4.5 has?

[tame current]

10/day

ig

[wide peak]

hell 10 a day is almost nothing

[tame current]

wait what 4.5 got released to perplexity?

[wide peak]

it got released to gpt as well for plus

[distant aurora]

I wish Claude had web access on their official website.

[distant aurora]

Try this https://www.croxyproxy.com/

[surreal marsh]

perplexity is cooked

https://www.androidauthority.com/google-search-ai-mode-experiment-3532243/

[fathom cloud]

didnt something similar to this already exist

[distant aurora]

It's just that perplexity isn't special anymore.

[left stream]

Side question, did they decrease 4.5 price?

Heard that 4.5 dropped for plus users

I imagine it’s with limits

[distant aurora]

The idea of multiple models under one subscription sounds good until you discover the flaws.

[distant aurora]

Yeah.

I expected another batman pfp, lol.

[surreal marsh]

today

[left stream]

Lol, I already had 5 this week

[surreal marsh]

of course, it’s around 50/wk compared to 70/wk on pplx

[left stream]

Good 🔥

Perplexity’s 4.5 definitely is the real thing

[distant aurora]

Have you ever had the laughing one?