💬│general

[ivory bramble]

HOOWWWWWW

COMET ??

HAHAH

[valid charm]

Care to share screenshots and talk about the testing you have done? 🫡

[fierce grove]

they sent me a mail 😭

[ivory bramble]

finally a rollout

[fierce grove]

its beta

for macs only i suppose

[ivory bramble]

i think because its perplexity, the search engine is perplexity obv

[fierce grove]

but doesnt support importing data from safari

[fierce grove]

there is an option to chnage the search engine

i changed it to google

and removed the homepage from perplexity to google

didnt work

it has a lot og bugs

and idk where to report them

@zinc quartz

any idea about comet ?

[ivory bramble]

interesting

[south kindle]

They'll need to add a new channel for it I'm sure.

[valid charm]

When you go to your Gmail inbox, does it allow to do taks for you?

[ivory bramble]

when did u get the mail. today ?

[fierce grove]

was just messing around with settings. it kinda sucks and there is a lot of room for improvement ahhh

i wanted to switch from safari

[fierce grove]

elaboarte?

[valid charm]

lmao

[south kindle]

i don't have access, but there is probably a work around to convert your safari bookmarks into a format comet can use.

[valid charm]

If you navigate to the web version of email, can it do stuff like "find what Bob said about alice's project and tell it to Berth"

[fierce grove]

it can also import in html

but it will be the bookmarks

not the history

or passwords

or other data

[fierce grove]

3-4 days back. i just checked my mail today

[fierce grove]

let me check

[ivory bramble]

thank you !

[fierce grove]

it works

the gmail one

[valid charm]

Dayuuuuummm

[fierce grove]

it's working guys

[south kindle]

Can you share any screenshots? is PPLX just added as a browser assistant?

[fierce grove]

you can use pplx as standalone by adding gmail and google drive as apps on pplx

or you can use it like browser assistant like how bing was introducede

comet also has an inbuilt ad blocker

finally

[austere granite]

Wtf is comet

[ivory bramble]

perplexitys browser

[fierce grove]

browser by pplx

[austere granite]

What

[ivory bramble]

but first with a waitlist

[austere granite]

Forget i asked

I'm completely uninterested

[austere granite]

Literally just chrome with ai

Cool

[ivory bramble]

hey is there an option to share invites for comet ? i got codes for the dia ai browser from the arc company if u want. if possible we could exchange

[fierce grove]

invite??

they just sent me a mail

with a link to the dmg file

[south kindle]

for PPLX?

[fierce grove]

no

for comet

[austere granite]

I haven't eaten in like 16 hours

I want brown food

Brown drink

[south kindle]

That's what I meant. surprised that's how they handled it. seems like giving everyone the dmg file would lead to it leaking beyond users invited.

[austere granite]

What is a dmg

[fierce grove]

lol

[austere granite]

I thought iPhones used deb packages for side loading

[fierce grove]

the comet dmg is still available on some forums

i saw it there around a month ago

but didnt download cause i didnt wanted to risk it

[south kindle]

So it goes. regardless I haven't seen or heard much about it, so I'd assume it hasn't caused any real issues.

[fierce grove]

nahh

comet is anyways broken

but its useful as it has gmail integeration

which isnt available on normal web

[austere granite]

It's called opening gmail it's integrated into every browser

You just bookmark it

[fierce grove]

pplx doesnt have gmail connector on web

it cant access my mails

it is able to do that with comet

and make and schedule. meetings and browse through my mails and sort them and delete them

[austere granite]

Why would you want the ai to look at all of your emails

[fierce grove]

saves time

[austere granite]

That just sounds like a

That's definitely an idea

Let me know how letting the schizophrenic computer program set your schedule goes

[fierce grove]

i used gemini up until now

and now pplx

so far so good

[zinc quartz]

Mind sharing some threads that are gone? We’ll look into it!

[green crystal]

Gemini does it and when it works is amazing at saving time

[green crystal]

Check if deep research there still has 20 sources and low reasoning lol

[fierce grove]

whattt??

sorry, i didnt understand

[green crystal]

In comet

Try deep research there

[regal sentinel]

Heya guys just wondering is anyone else having some issues with copying answers? Perplexity doesnt embed the links and instead keeps them as bracketed numbers like [1] without actually being a link you can click on

This only happened very recently so Im curious whether this is intended

[south kindle]

I believe this is a new bug. I’ve seen others mention it.

[fierce grove]

they provide the links at the end of the text

[valid charm]

Can you to share the DMG?

[regal sentinel]

Yeah thats fine Its just more convenient if they are embedded in the text

[dry kestrel]

Can you share the file?

[fierce grove]

😭😭

[fierce grove]

nahh i disagree. i had to manually remove all the embeds when submitting assignments or stuff

[austere granite]

Oh ew

Good thing i don't use Gemini

Gemini sucks anyway

[zinc quartz]

Hey VisionArchic, as mentioned in the email, we ask that you please refrain from discussing anything re Comet

[fierce grove]

most people disagree

[fierce grove]

alright, sorry

[zinc quartz]

Any bugs or feedback you can report in app.

[fierce grove]

i must have missed that line

[fierce grove]

alr

thanks

[sullen ember]

When you copy the link, it should cite the sources and have a link at the bottom of the copied text. Is that different from what you are seeing?

[zinc quartz]

Nah all good. Thanks for trying it out

[fierce grove]

they want it to cite them with the embeds rather than citing them below

[austere granite]

Why yall get so excited for the ping button when it's this nonsense

You don't see me pinging anyone

[sullen ember]

It's plain text so embeds wouldn't work exactly, do you mean having the link be inline with the citation? Do you mind explaining the use case a bit more?

[green crystal]

I probably shouldn't continue prodding deep research but it's hilarious to see it go as low as 10 sources

[valid charm]

Can you share your prompt? I'd like to tinker with it

[green crystal]

I deleted the thread but it's
When will perplexity ai 's comet browser release

[inner isle]

I'm trying to get pro for students but it keeps saying my student email (@live.uwe.ac.uk) is invalid?

[valid charm]

You'll have to reach support. They made changes recently and are working out the kinks

[distant aurora]

That person is, iykyk

[distant aurora]

Same.

[valid charm]

+1

[gray prawn]

Here you go! If you need even older threads I can try and get them, these are just recent ones that I have noticed got cut off

https://www.perplexity.ai/search/big-hunting-expos-in-alabama-rOEZaPecQOaIDk7FSwu8Qg
https://www.perplexity.ai/search/what-is-becoming-more-popular-n8URxHRsSry5mANnyi6hkw
https://www.perplexity.ai/search/how-to-train-my-own-llm-ZxAeDOX2TfiZtrJIZzW7Vg

These are all bookmarked on the perplexity app. I had to right click and copy link because there is no other way for me to share.

Let me know if you need any other info like screenshots or something!

oh crap

my bad

You can deleted the messages if you want once you save the link in order to declutter that. I didn't expect it to do that lol

[frank island]

In the 24 months perplexity pro offer on my app it is showing that 22 referrels have been successful by my I'd but I have got 14 months extension only of pro account.

If someone can help or guide in how to get it?

[valid charm]

I love the graphic design line PPLX uses

Maybe they should reselase a wallpaper collection?

[narrow silo]

can perplexity remake images into that popular studio ghibli style?

[outer marsh]

No

[serene sand]

#💬│general message

[inner isle]

where?

[zinc quartz]

Hey there! I’d recommend reaching out to [e-mail]. anything related to billing and promos, we won’t be able to help you here.

[inner isle]

What's the average period before one recieves a response?

[zinc quartz]

Hey Trobot I’m not sure. Some have received responses in an hour and some complicated ones take longer.

[willow kettle]

Waiting on the new Sonar model based on LLAMA 4

https://ai.meta.com/blog/llama-4-multimodal-intelligence/

[zinc quartz]

No problem! We'll bring up to the team.

[gray prawn]

Appreciate it!

[shut dagger]

llama 4 too 🫠

[frank bane]

Am i missing something, or does the mac app not have an "Export" function like windows and the web?

[crisp kiln]

llama 4 just dropped

Sorry I missed that you'd be slowly rolling out release, forget what I said if you read the notification.

[zinc quartz]

We’re already cranking on this rn. I’m in the office too on Saturdays.

[crisp kiln]

are we getting llama 4 or sonar 4

[tame current]

Yo

[distant aurora]

@zinc quartz are the rate limits for deepsearch same as pro search?

[green crystal]

Y'all should prioritize fully releasing deep research high

[plucky talon]

Is there a way to add perplexity AI as a bot to a discord server? If so, does it provide up to date info

tried chatgpt and all data is from 2023

[left jasper]

It's not that good. Imo it's not better than their previous deep research - it just takes longer.
Googles deep research is far superior and often even their gemini 2.5 is better (and clearly faster).

[green crystal]

Before it was removed it did quite well

[dry kestrel]

gemini DR performs much better than pplx deep research

[green crystal]

And they have to bring it back anyway
I mean it's a part of the sub value...

[serene sand]

#1358062760474443786 message

[left jasper]

That was the old deep research. And yes, it did pretty well. Taking a few minutes and giving a good answer.

The new model, waiting 10-20 minutes for a subpar answer (token output limit?!??) is just terrible

[green crystal]

10-20 minutes? For me it's 30 seconds to a minute max lol

[left jasper]

That's not their new model then.

[green crystal]

Yeah you probably got picked for testing their partial rollout

[crisp kiln]

It says deep research high for me currently

and it has never taken that long

it takes like 3 minutes

https://www.perplexity.ai/search/manchego-BJoy7Q4NRuiDjGD6kv6l2A

Gotr this from it

probably took less than 3 minutes

[valid charm]

Are you using Complexity?

[left jasper]

That's not really a prompt though?

[crisp kiln]

Yeah, does it add the option for deep research high?

[crisp kiln]

It should have thought longer and HARDER about manchego

[left jasper]

Make the prompt ask for that then?

Try something more complex, like:
I found this text as a status update on discord. What could it be?
:3@[m@<*K$FWbd7E,]6-+D\m'FECu7Ci![#+DDs-@<6O5DCt

[valid charm]

Yes, it seems like it does

[crisp kiln]

these are poor excuses 🥱

[left jasper]

It'll think for many, many minutes

[left jasper]

Why would LLM spend 30+ minutes on a less thought through question? Is it even a question? It's almost like a statement...

It's like saying "Banana" and wanting a company to spend 30 min of GPU power to try and understand what I actually mean which is actually stupid

[crisp kiln]

Terrible excuses 😴

[left jasper]

Stupid "question"

[crisp kiln]

He's getting defensive over these poor excuses 🛌

[left jasper]

What was your expectation? What kind of answer were you looking for, just asking Manchego?

[crisp kiln]

The space itself has a very long instruction 😪

[left jasper]

You do know that this is not shared in your link...

[rich onyx]

I can't see a Generate image or any other button in a sidebar in a thread, like the docs and other articles say. (see https://www.perplexity.ai/help-center/en/articles/10354781-generating-images-with-perplexity). The focus mode is also not there, I can select sources or something ,but that seems a subset of focus modes as the docs describe (https://www.perplexity.ai/help-center/en/articles/10354753-what-is-focus). I have a Pro subscription.

Am I missing something?

[crisp kiln]

It doesn't show the space 🤷‍♂️

[left jasper]

Exactly. To everyone opening the link, they only see "Manchego". Which is not a question.

[crisp kiln]

Right but the 20 instructions I gave the space are

So they need to stop with the excuses 💤

[left jasper]

Or you need to share the system (space) prompt + your one word...

[crisp kiln]

Oops I put the instructions in the description field and the description in the instruction field

done that 3x now

uis fault stop the excuses

👼

[left jasper]

😂

[crisp kiln]

😓

It took less than a minute this time, but I think it's because complexity makes it say deep research high when it is just the regular old one

[left jasper]

You still asked it a single word...

[crisp kiln]

Space prompt:
Research assistant for cheese monger in training. You will find all cheeses that exist, and explain everything about them from origin, history, making of the cheese, and all other information a cheese scientist, cheese historian, cheese educator & cheese monger should know about the cheese. You will also find what the cheese is made of, what it is suitable for and what it tastes like and the flavor description should be detailed, talking about what you taste at first and what it develops into in your mouth as time passes.
Always respond in an enthusiastic tone, and be as educational as possible while remaining memorable & understandable for the layman.

[crisp kiln]

the prompt is in the space

[left jasper]

Try and copy the space promt into a new chat with your "one word 'question'"

[crisp kiln]

I give the prompt to the space so that I can just tell it a name and have it evaluate everything based on that one time I gave it a list of prompts

[wanton elk]

Hello

How are you guys?

[civic cairn]

Why does it say I have 3 uses left for the Gpt 4.5? Wasnt it 600 before?

[vapid elm]

It was always 3 and it is now 0 since they removed the model - afaik it simply redirects it elsewhere if you're using Complexity

[civic cairn]

Oh Why they removed the model. Is it because of the cost? Is Sonnet 3.7 way better than 4.5?

[serene sand]

1. #👑│pro-lounge message

2. #💬│general message

[zinc quartz]

Yep, I think it's like some 500 or something?

[rich onyx]

Thanks, found a thread indicating the UI was kinda ruined, since you can only generate images if youget images results, and that's not consistent at all.

[serene sand]

patience, it will get better 🙏

[valid charm]

https://tenor.com/view/hello-hi-hy-hey-gif-8520159980767013609

[valid charm]

This used to work but got patched a few days ago

[valid charm]

Try it this way

[left stream]

Llama 4 release, heard 10 million tokens, true?

[valid charm]

Yes, it is true

A beast but let's see if it is any good recalling all that memory

Early versions of Gemini suffered from bad recall making the large context window useless

[west remnant]

Is there a reason model quality seems to vary so much? Specifically Claude 3.7. Is there a hidden limit that switches to a lower quality model after a certain time?

[green crystal]

well we do pay for it soooo

[distant aurora]

It got nerfed slightly I think.

@left stream heard of this? https://www.rxddit.com/r/singularity/comments/1iec2p9/deepclaude_combines_claude_sonnet_35_with/?rdt=35660

[distant aurora]

What was the prompt?

[distant aurora]

I don't think it even utilises half of it in most cases.

Applies to most models.

[mint rampart]

i found someone was selling genspark tutorials with loaded accounts, are those valid?

[distant aurora]

Even on chatgpt, outputs stay close to less than 1000 tokens though the limit is 4000 for outputs.

No matter how I prompt it.

I'm talking about gpt 4o specifically here.

It stops mid response around 1500 tokens I think?

@left stream these rate limits look very generous on the surface https://www.giz.ai/ai-rate-limits/

[left stream]

Who knows, all I know is that the two they released are nothing in comparison to llama 4 behemoth

I thought the two new models were amazing until I heard it’s thanks to distillation from Llama 4 Behemoth.

288 billion active parameter with 16 experts

[left stream]

I’ll check them out, haven’t heard of gizai before

[distant aurora]

Copilot is lame af

[left stream]

Of course

Their o1 is garbage imo

It’s good

But not as good as o1 from other platforms

Well its arguable, I noticed o1 from copilot did somethings better than o3 mini medium reasoning but not better than o3 mini high

[distant aurora]

But I think it should be better considering that it's the only model they offer I think?

That 20 dollars is probably in the name of platform integration.

[distant aurora]

@left stream is grok that bad? They were saying it's the best model during launch. https://bigcode-bench.github.io/?utm_source=chatgpt.com

[left stream]

Yes

They should’ve assessed o1 pro to the benchmark

O1 pro is a monster

[distant aurora]

The price killed it's hype I believe.

[left stream]

True

Just imagine o4

They already have o4 mini

O3 pro, o3, and o4 mini are releasing soon. Let’s see how it’s gonna go

Might be this month for o3 and o4 mini

O3 pro much later probably next month

[distant aurora]

I'm assuming O3 series will be replacing o1 and o4 mini will replace O3 mini.

I wonder what the open model would be.

Might just make some old model open.

[shut dagger]

If I must say, don't keep much hopes for open model

For one, they probably won't try to release model better than 4o

[distant aurora]

It's more of a pr act I believe.

[shut dagger]

From the above chats, I assume you want to buy a subscription, why not perplexity?

[distant aurora]

Claude and O3 mini glitch.

And perplexity is still on the menu, I have not outright rejected it.

The deepsearch and pro rate limits are value for money.

[left stream]

Been using llama 4 maverick and liking it so far

[distant aurora]

Is meta ai using llama 4?

Yes it is apparently.

@left stream https://ibb.co/Swzc8pm8

[left stream]

Yes, you can also try it via openrouter they have scout and maverick

It’ll be awesome once behemoth is available

That is gonna be a monster

Both are keeping up and can handle very long pieces of text

[left stream]

That last response lol

[warm perch]

is there a list of schools that perplexity considers for their 1 year membership

[cobalt vale]

that's grok V1... I don't think the latest model is on there

[red atlas]

where do i get the pro disocrd invite from !?

[distant aurora]

Ikr

Bro lost context fast.

[worthy harness]

When are we getting Gemini 2.5?

[distant aurora]

They're working on it.

[distant aurora]

Claude?

[distant aurora]

https://artificialanalysis.ai/leaderboards/models?utm_source=chatgpt.com

Select human eval or scicode in intelligence index.

Grok beta is grok 3.

And pretty low.

Probably got ignored by people I think. Or they didn't enable the reasoning switch.

[frigid oracle]

I actually wish they had scout that can take 10M context length.

From what I understand, behemoth might not be good for conversation

[late rampart]

When will the check sources option be available for everyone?

[random ocean]

Any timeline on when can we expect Deep Research high and Google Gemini 2.5 Pro integrated into Perplexity?

[serene sand]

https://youtu.be/j6QENSMY2qI?si=E5F2jtsSWNqG2ESE

☕

[foggy stone]

Hi chat

[wheat pike]

same prompt, deep research visits less sources than sonnet 3.7 reasoning

interesting

[crystal cosmos]

• Chicken
• Watermelon
• Dodge Charger

[green crystal]

My deep research gets on average 20 sources max...

[wheat pike]

it all depends on the prompt

i was planning for a trip, so i gave mine a list of locations and clearly specified the format of the output i wanted

[green crystal]

It's supposed to look for information surrounding my prompt aswell

Next to just directly looking for what's asked

[wheat pike]

what was the prompt?

i find that it usually just goes deep enough to answer your question

[green crystal]

Asking what analysts expect gold price to be

Deep research high did very well with it

Looking up info from multiple perspectives

I mean even grok did it

[wheat pike]

i see

yeah these kind of open ended exploratory questions seem to be what deep research is good for

but if you have an idea of what kind of output you want, sonnet reasoning could give you better results

[green crystal]

If you try to include all those extra searches in your prompt to force it then it'll just stop searching after a short time and start reasoning
Leaving out allot of information

[green crystal]

Yeah sonnet reasoning actually has better search atm lol

I know they mentioned that it'll be rolled out to everyone soon but geez I'm slowly losing patience

[wheat pike]

what will be rolled out

[green crystal]

Deep research high

[valid charm]

Good morning PPLX village

[mental bone]

Bro, with Llama 4's context window and cost, there's no reason not to increase Perplexity's context limit past 32k.

[warm sapphire]

what app is this

[valid charm]

This is what we know aboyt increasing the context window

[green crystal]

i hope this doesnt include deep research

eh the reason why its so shallow

[valid charm]

🤣

[summer crescent]

I still find this phrasing amusing, I literally use another search with unlimited use along with perplexity because perplexity does have a limit..
the very unlimited limit isn't unrealistic to hit

not quite sure why it's advertised as unlimited, during research it can easily be eaten up especially if it doesn't find what you're looking for in the first "many" tries

[green crystal]

its capped at 500 per day i think

[summer crescent]

If I remember correctly the base models is 500, but the pro search is 300

so you can followup ( not using the pro model ) with the base models for 500, but new searches ( especially pro ) was 300 last I checked unless they upped it 🙂

but it's fine, I use another model that's very comparable that does give unlimited searches with a custom llama model with CoT.

I just find the advertising a bit disingenious

pay a subscription to pizza milano and get unlimited pizza, up to 2 slizes a day

[distant aurora]

Grok deepsearch is underrated.

[green crystal]

its very transparent with how it searches and analyzes things

[distant aurora]

Gimme prompt

[green crystal]

what is expected to happen with gold price tomorrow

[distant aurora]

Does look slightly nerfed compared to 50+ in usual experiences.

[green crystal]

the amount of search steps is also much lower

like 2 to 4

[distant aurora]

R1 search gave me more sources.

Deepsearch https://www.perplexity.ai/search/what-is-expected-to-happen-wit-BgctazcsTZ2xkmlXu2WMEA

[shut dagger]

it's nerfed? my 99% searches gives max 20 sources only

[distant aurora]

R1 https://www.perplexity.ai/search/what-is-expected-to-happen-wit-Oep3mmrGQlGQlT_t1qpHYw

[distant aurora]

I mean, I have seen it go over 100.

Depends on the topic maybe? 🤔

Okay wait.

Could also depend on the way you prompt.

[shut dagger]

for me, only deep research exceeds 100 sources

[distant aurora]

I haven't experimented enough since the recent update.

[distant aurora]

I was talking about Deep research.

Pro search used to be 30-50ish I think.

Anyways, more sources doesn't necessarily mean more info I believe.

Unless it's something scientific or analytical with research papers of trials.

[green crystal]

its mostly dependent on how many search steps are done

but its supposed to do more since the whole idea is that it can research the prompt from multiple perspectives by itself

[distant aurora]

@green crystal Provide a comprehensive analysis of the factors that could influence the price of gold tomorrow. Include technical indicators, recent market trends, macroeconomic data, geopolitical developments, investor sentiment, currency movements (especially USD), central bank activity, and any scheduled economic announcements. Also, explain both bullish and bearish scenarios, and provide expert predictions or consensus if available.

Try this

Though the entire point of deepsearch is not having to refine a prompt. But I want to see if it makes difference in number of sources.

[green crystal]

your prompt is literally gemini deep research plan lol

[distant aurora]

Just put it on perplexity and give me the result. I'm free user so I ran out of 3 prompts per day.

[green crystal]

36 sources

[valid charm]

Did you use High?

[green crystal]

but it doesnt look like it

[green crystal]

nope dont have the option

complexity doesnt show it anymore aswell

[distant aurora]

Nerfed af

[green crystal]

im trying grok now

[valid charm]

To make sure I get you the most useful and up-to-date insight, could you confirm if you'd like this analysis to focus on:

Spot gold prices, futures, or both?

A particular exchange (e.g. COMEX, LBMA)?

Any specific timeframe for technical indicators (e.g., 1-day, 1-hour)?

Should I include gold mining stocks or ETFs in the broader sentiment, or just the metal itself?

Once I have that, I’ll dive into the full analysis.

I'll use it on ChatGPT and PPLX High

[green crystal]

its very thorough when going through sources

102 sources and counting

[distant aurora]

@green crystal

Give me a detailed report of relation between healthy fat consumption and brain health. Focus on dementia and Alzheimer's

[green crystal]

you just tried it?

[distant aurora]

No. That's old.

Try the same prompt to see if you get as many sources.

If you don't then it's obviously nerfed.

[green crystal]

alright

[distant aurora]

Don't forget to enable social and academic.

[green crystal]

oof time to redo it then lol

[shut dagger]

thinking steps of Deep research reminds me of Sonnet 3.7

[green crystal]

but yeah its been confirmed that its nerfed due to their infra not being able to handle it
it should be good again when they're switched over to golang

[green crystal]

the issue is i hope it doesnt take forever because its usually not a smooth ride from one language to another

[green crystal]

50 sources and only a single search step

[shut dagger]

more sources or less sources doesn't mean better answers

https://www.perplexity.ai/search/provide-a-comprehensive-analys-pLNBDCSGT7eWRCCOHbWddQ

it says 55 sources, but only used 14 sources to form it's final answer

[green crystal]

search step reasoning quality and amount does

grok just goes looking around by itself

so even if it were to use less sources

[valid charm]

Dude at least change the text a bit each time you post it

[green crystal]

it looks at the prompt from different angles by itself

[left stream]

https://tenor.com/view/sukuna-scale-of-the-dragon-recoil-twin-meteors-world-cutting-slash-gif-1831960037484152553

[left stream]

True, might be the reason why it’s still testing.

Hopefully while it’s being tested it’s flaws get mitigated as much as possible

[left stream]

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

[valid charm]

Misspellled Grok?

[left stream]

Yes

[left stream]

I mean no

Claude 4 maverick

Llama 4*

[distant aurora]

What app is that?

[left stream]

Pal, let’s you use any ai as long as it has an api

[distant aurora]

Likely. Though groq is a website that offers distilled R1.

[left stream]

If it doesn’t you can tell the maker and they’ll add it

They have openrouter

[valid charm]

Huh I see

[distant aurora]

How to get api?

[left stream]

For openrouter?

[distant aurora]

https://groq.com/

[distant aurora]

I thought api was for models.

[valid charm]

Interesting name choice

[distant aurora]

I have never gotten or used api key.

[left stream]

https://openrouter.ai/chat

[left stream]

Yes to access them they’re the key, but not all the keys work for every door. There might be a door made by openai but you have an open router key

[distant aurora]

Complicated.

[distant aurora]

Does a key have a usage/time based validity?

[left stream]

No

Thanks

[muted drift]

Grab it!

[left stream]

Already did I got lucky I have two edu emails my college gave me 😭

[muted drift]

Don't forget to share my referral link with your peers

[polar lily]

why does perplexity always default to pro when im using deep search?? it anoying when doing long sesions and it destroys the whole session because i forgot to hit deep research evry time 😒

[green crystal]

You can delete answers if you accidentally generate using pro

No need to make a new thread

[shadow raptor]

Do you think Perplexity subscription is worth it?

[polar lily]

but still it's anoying. i dont need it to chose a engine for me. i want to be abel to folow the rabithole on the same engine

[polar lily]

i never tried it without. but thats because i have the subscription because my bank gives it to me for free 🙂

[serene sand]

#👉│feedback-general message

#📢│announcements message @polar lily

[cobalt vale]

new deep research doing agent things

[tropic crest]

how did u get it?

i saw that once and it never showed up again

i still dont like it

its nowhere as close to chatGPT

[tame current]

😦

[cobalt vale]

yes, I believe they are limiting access

feels like 1 prompt and we get rate limited

[tropic crest]

i sitll wasnt happy because i feel like all that perplexity is doing is chucking 100k tokens into R1 and asking it to write 10k tokens

and the content is just very surface level and feels like it just broke down your query into 4 parts and stitched it together

and doesnt feel coherent

it feels like u asked a highschooler to write a 10k word essay

idk what chatGPT does but it just feels like a much more coherent report. It goes into proper depth, asks follow ups and the entire report is very cohesive

[cobalt vale]

ChatGPT is just smarter

[tropic crest]

agreed

[cobalt vale]

it uses o3-preview

[tropic crest]

yeah like it feels like chatGPT reads everything, internalizes it and then writes a report. Almost as if you let an expert learn about a topic for an hour and then they wrote it themselves

[cobalt vale]

that means it's way too expensive though

[tropic crest]

whereas perplexity feels like u just asked a high schooler to copy and paste random stuff till it hit word limit

agreed

rate limits suck

[polar tendon]

Fr

[cobalt vale]

these are all called deep research, but really there's 2 tiers

[tropic crest]

agreed

[cobalt vale]

one is chatGPT with their truly deep research, and then there's everyone else's enhanced search

actually 3 tiers maybe

old perplexity deep research and grok deepsearch, possibly the smaller competitors such as chatLLM as well

[tropic crest]

even the new deep research wasnt upto par tbh

[cobalt vale]

and then new perplexity deep research, grok deepersearch, and google deep research

[tropic crest]

and i think it was mainly because theyre still bottlenecked by R1

[cobalt vale]

not really ChatGPT level but a still more comprehensive than basic search

[tropic crest]

fair yeah

[cobalt vale]

well R1 is their only choice

[tropic crest]

I mean they can use Claude too right

I guess soon we'll get llama 4

[cobalt vale]

only if they release llama 4 reasoning

[tropic crest]

Hopefully if what they say is true about it's perf they might run post training on it and get good results

They will soon enough

It'll take months before pplx can launch it

[cobalt vale]

and llama 4 at the moment is not very good

[polar tendon]

Pplx doesn't feel worth it for its quality anymore

[cobalt vale]

simply too expensive

[tropic crest]

Yeah valid

The only reason I'm using pplx is bc the search + reasoning combo is v good

[cobalt vale]

could be worth it if we get more than 1 deeper research prompt

[tropic crest]

I just used deep research from chatgpt to generate a textbook chapter about this field I'm trying to learn and was blown away by the quality

[cobalt vale]

also feels like perplexity is the last third party to get revolutionary new models such as Gemini 2.5

[polar tendon]

What's the chat gpt paid limits for deep research? 20$ plan

[cobalt vale]

10 per month🫠

[tropic crest]

yeah it sucks at 10 per month but man the quality is just insane

[polar tendon]

Which site you guys think is kinda better value for its quality?

[tropic crest]

well at the moment perplexity ngl

I just hate the perplexity sucks at coding/communicating with base models

[polar tendon]

I feel perplexity is very bad at anything long and continuous

[tropic crest]

exactly

luckily for me i cycle between claude and chatGPT free limits for anything long and continuous

[cobalt vale]

completely depends on what you are doing

[tropic crest]

grok is good enough for the long and continuous

coding too its good enough for me

im anyways paying for twitter so it kind of works out for me

[polar tendon]

Its very annoying. Doesn't remember context and all. Overall I'm annoyed by perplexity

[cobalt vale]

don't sleep on AI studio or even Gemini if AI studio rate limits

[polar tendon]

Oh Gemini is pretty good. I have been using it recently

[cobalt vale]

2.5 Pro crushes benchmarks everywhere

[tame current]

No way 😭

[tame current]

This the worst thing about Perplexity. Can’t remember anything for more 2 or 3 prompts

[distant aurora]

@sturdy skiff

[shut dagger]

Hey @vague pagoda, events announcements are done at #📰│server-news

I thought we got announcement for deep research high or gemini 2.5 pro 🫠

[vague pagoda]

Thanks for flagging, but this was intentionally posted in #📢│announcements as the events channel will be modified with the incoming changes - apologies for the inconvenience!

[vague pagoda]

For anything like this we'd ping @Perplexity Announcements

[cobalt vale]

personally I was hoping for more Comet news

[lone hound]

Check:
#1357104230770540574 message

[zinc quartz]

Hey guys - Gemini 2.5 Pro API isn't available for commercial use yet, only preview modes. Stay tuned! Once we're allowed, we'll be adding.

[shut dagger]

quick question, what's context size of R1 ?

[zinc quartz]

128K I believe

[shut dagger]

in PPLX?

[left jasper]

What is Claude 3.7 reasoning output tokens limited to? Been running in to stopped output (single prompt, first output) a few times recently which doesn't happen for R1.

[shut dagger]

"I believe"....ambiguous, doesn't confirm anything

[shut dagger]

longest output I got is 5.7K tokens

[zinc quartz]

Breh I added "I believe" bc responding "128K" seemed too curt

"128K confirmed"

[green crystal]

Deep research high would be quite useful for me now

[shut dagger]

That's more like it

[zinc quartz]

Definitely. We're continuing to roll it out to more and more users so stay tuned for that.

[bold imp]

currently? finally? any other info we need on that?

[zinc quartz]

No firm launch date announced yet. We're still getting feedback and fixing bugs. Taking a bit more time to work out the kinks vs. shipping it too quickly.

[cloud hollow]

Has anyone seen/used the new Deep Research (High?)? By the way, I'm getting really good results (I can say clean, to the point) with the existing one, and I'm looking forward to using the new one.

[bold imp]

I think these disclaimers are important to prevent people from looking for it and then being disappointed not seeing it in their perplexity

[green crystal]

Inb4 it gets released right after my sub runs out

[atomic whale]

heck yes

[valid charm]

Yes, you get good models integrated into a great search infrastructure. Stuff like 4o performs better for searches in PPLX. Also you can switch from competing company's models in a single conversation seamlessly.

[sly dome]

Hey guys. It's my first time asking or answering a question. Anyways I was reading a little bit about AI and I was wondering if you guys need of premium put together your project or an AI that's completely free to help me gather any declassified or classified legally about my time in service that can gather in depth analysis about toxic chemicals, toxins, hazard material, biological radiation, airborne hazards? I know about the EPA, DOD, and filling out a form to request records . But there's another AI I can get from GitHub, online somewhere legally 😉 if course, plus it would have to be able to find and be able to comprehend the material written on the documents and make a case with EPA, atsdr, legitimate documents stating what was at ft Irwin CA before 1993- 1996. And what each chemical is identified and then be able to break those down and find out how those little separate chemicals could have caused most of issues that are going on right now? If you don't want to help me than that's cool, but I figured I'd better ask just in case. Thanks for your your time and patience.

[tranquil orchid]

Hey @sly dome!

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.

[sly dome]

I have perplexity pro. Also I need it be able to make a case for me and link it all together.

⭐

[distant aurora]

Good one

@left stream Microsoft's stupidity needs to be studied.

[sly dome]

Is there any others out there that someone has put together myself that could 📄 me

[sly dome]

⭐

⭐⭐⭐⭐⭐

[left stream]

LOL

You can possibly bypass using image generation on chatgpt

It did some things that I didn’t expect on chatgpt

Also

This model dropped a hard line

[distant aurora]

Peace was never an option

In a nutshell.

[distant aurora]

The problem with chatgpt is that it says copyright issue everytime I ask it to make iron man or Darth Vader

This wasn't a problem earlier.

I don't know why they made it this way now.

[raw harness]

Hey

[left stream]

You can bypass it but you would have to paraphrase it

I had it do that so I can have it make superman smoking kryptonite like it was marijuana

[frank apex]

Hey guys my API suddenly dissapeard.. When I try to add my a new payment method it gets authorized on my banking app but nothing happens afterwards..

[misty quiver]

Not sure why in the same thread when I started with Reasoning R1, the follow ups start with AUTO model.

Perplexity gave me this answer

The lack of options to stay in the same model is just a frustrating experience

[tame current]

When is Comet coming for Windows?

[lone trellis]

Only Linux?

[tame current]

Is it?

[cobalt vale]

no chance, they will support windows for sure

[tame current]

Yeah thought so

Lol

[valid charm]

It's currently available in macos durint the testing phase

And yes, it is definetly coming to Windows

[cobalt vale]

you got access?

[valid charm]

https://tenor.com/view/yes-man-no-man-no-gif-7347715096894588969

There were to users who came and commented

But turns out they are not allowed to share during the testing phase

[cobalt vale]

commented where?

[valid charm]

Here in this chat

[cobalt vale]

ah

so us normies have a chance

[valid charm]

Start reading here

[fierce grove]

what?

[wise maple]

hello

[valid charm]

Nothing lol I'm just letting someone else know the little infor we have on Comet

[cobalt vale]

comet user got summoned

[left stream]

@distant aurora

[valid charm]

lol sorry

[fierce grove]

ahh hahaha

heyy lol

[fierce grove]

you could have turned off the mention 😭

[fierce grove]

but lmao it's alr

[cobalt vale]

were you early on the waitlist?

[valid charm]

Sorry I'm discord noob. Will do next time

[fierce grove]

ahh i just got access to it

it's fun and cool

although man6 improvements can be made

[cobalt vale]

yes, but is it because you signed up early?

[fierce grove]

nope

i just got the invite

[valid charm]

I think he means the waiting list

[fierce grove]

yeah i understood that

and no i didn't sign up

[cobalt vale]

h o w

[rocky wagon]

Can you verify me so I can add it to my server please?

[fierce grove]

idk 😥

maybe cause i am a paid user? and once tried searching for comet

i didn't sign up for comet
i was interested but i didn't find the sign up page

so no i didn't

[cobalt vale]

maybe they accidentally deleted the waitlist and just rolled out to random accounts instead

[valid charm]

Whaaaaaaaat

[fierce grove]

no idts

[valid charm]

https://tenor.com/view/joey-joey-tribbian-funny-stare-wink-gif-20597720

You must have a friend on the inside

[cobalt vale]

like half the people in this chat probably

[fierce grove]

i wish 😰

[valid charm]

Maybe you use it a lot?

[fierce grove]

no

[valid charm]

Hmmmm

[fierce grove]

i just use pplx to critic my plays

and get data and insights on my research paper

hardly an hour a day

[valid charm]

You should play the lottery. Seems like you exude luck

[fierce grove]

😅

how about you bet on me

😆😆

[valid charm]

lolololol

[fierce grove]

why these emojis so big

[cobalt vale]

emojis are bigger when sent without text

[fierce grove]

i used a whitespace i suppose

[cobalt vale]

Kwak pls explain how non waitlisted accounts get comet earlier

[valid charm]

That is a interesting question

[tame current]

Are there any mods here?

Where is the reading out loud feature on the perplexity ios app?

It went missing last week and it still hasn’t come back

I use it to help with my adhd

I haven’t updated perplexity on my macbook and it still has the feature

im not going to update it anymore if they removed it

[valid charm]

Yes there are mods here

[valid charm]

You are the second person I see making this report

[ocean wharf]

I spent my whole life trying to understand how social media worked. It's very stupid. If your video doesn't meet certain criteria, it will be trashed. If you upload it again with the missing criteria, it will be promoted. It's ridiculous to think that this is the problem. Some people are arrogant enough to think that people didn't like the video, so it didn't spread. The AI discovery system and the bullies should not let you down about your efficacy.

[shrewd crest]

Please how can I get my referral link to refer my friends

[tame current]

Thank you, I’m glad it’s not gone, but scrolling all the way to the bottom of the answer, tapping on the headphone icon, and then scrolling all the way back up just so you can follow along is such a bad design decision

It was better when it was at the top right

[hollow quiver]

Does Lockdown mode need to be completely disabled for app to work on iPhone?

[untold delta]

I've question, Why Perplexity is not using Llama models even though it is Open source

[zinc quartz]

If you're a college student, you can get your code here: https://pplx.ai/students

If your email is not working, please DM me the college email and we'll see if we can add it.

[wheat pike]

idea: have a 'is my school email supported?' checker and add schools that get asked about a lot

[scarlet bramble]

Is ASU supported?

[distant aurora]

What am I looking at here?

[distant aurora]

Lmfao 💀

Didn't know 4o mini was so squeezed.

8b can literally run locally on my laptop.

[gleaming cloak]

They do use Llama models. Their in house Sonar is trained on Llama.

[distant aurora]

What do you think they use on the non pro search?

[wheat pike]

@zinc quartz @vague pagoda doesn't this break no advertising rule

[distant aurora]

I wasn't sure whether or not to bring it up.

[zinc quartz]

Yes we will handle thanks for bringing it up

[distant aurora]

Real. They can tell gender from eye balls.

[polar tendon]

Anyone knows Gemini advanced usage limits for 2.5?

I can't find it anywhere

[untold delta]

I was talking about latest model llama 4 as one of the options for pro search

[untold delta]

They might be using it for internal use-case but I was expecting it to be one of the options for pro search

Just to see results

[cobalt vale]

Honestly I don't consider it more useful than other dumb 8B models

[cobalt vale]

It's probably high enough that you don't need to worry about it. Most consider it unlimited

[shrewd crest]

I just sent you a dm

[narrow path]

Hello, how do I obtain the pro role

[distant aurora]

They do an okayish job at extracting information from documents locally.

[distant aurora]

@halcyon coral can you help her?

[narrow path]

I was told to dm ok.alex I believe, but I'm not sure where to do that

[distant aurora]

@signal hamlet ?

@narrow path hope this helps

[vague pagoda]

We've sorted it out for you, you've got the role now

[distant aurora]

Thanks kes

[vague pagoda]

[distant aurora]

Btw, in response to the nohello thing, I usually wait for a response to make sure the other person is available/willing to listen.

It's like knocking a door if you get what I mean.

[distant aurora]

https://www.perplexity.ai/search/gemini-2-5-advanced-rate-limit-x.Yb80E_RDi7UgixBfyOag#0

[cobalt vale]

I don't trust them

small models hallucinate

[cobalt vale]

that response does not include gemini advanced, only API

[distant aurora]

Fair

[distant aurora]

But yeah, I don't think there's any info on that there.

Wait.

So O3 mini hallucinated.

But grok didn't?

W grok.

[cobalt vale]

deepsearch?

[distant aurora]

Nope

Grok 2

Hilarious

https://grok.com/share/bGVnYWN5_46668bf5-a670-4894-8651-84cff597a947

[cobalt vale]

stop using grok 2 pls 🫠

[distant aurora]

Grok 3 has rate limits.

Like 10 or 15 per 2hr for regular prompt.

And 5 for thinking.

1 for deepsearch.

Yes they nerfed it.

[cobalt vale]

grok 2 should be worse than both 4o and 2.0 Flash Thinking

[distant aurora]

Well, see the results for yourself.

https://chatgpt.com/share/67f3967f-f14c-8013-aa7a-d519a780af2f

Okay nvm, only o3 mini hallucinated.

Non reasoning search didn't.

[tame current]

Good morning chat

[primal cave]

It’s 8 at night for my country

[boreal dew]

time is a lie that the government wants you to believe in

[primal cave]

So all this time am I being lied to

😭😭😭

[boreal dew]

i am sorry that you got to know the truth like that😔

[serene sand]

completely agree with you

[raw swan]

I need help from support

I don't see my Perplexity Pro subscription on my account. I had 12 months free because I have T-Mobile in Poland and after one month later, which is today my account has "free" tier instead of PRO. Anyone had same issue?

[serene sand]

please contact [e-mail]

[primal cave]

Wake up to reality…

[boreal dew]

harsh wake up call is better than no wake up call
remember to wear your aluminium hat today, you never know

[polar tendon]

It's unlimited

[distant aurora]

Yeah. I realised that eventually.

[primal cave]

Yeah thanks

[primal cave]

I might get shot you’ll never know

[boreal dew]

exactly. the aluminium hat will protect you from exactly that and ofc the chem trails by the air planes

[primal cave]

And the harmful suns uv rays that can explode my brain if I don’t use the hat

[boreal dew]

and now imagine that combined with 5G! outrageous

[distant aurora]

If you are making a bunker, make sure there's enough food, movies and books. Leave an extra seat for me too.

[boreal dew]

the seat has been reserved for you and prepared in case the world will end when AI takes over the planet
idk who would even use this dangerous stuff!!

[polar tendon]

Perplexity ai after becoming sentient : forgets 😂

[boreal dew]

i think if pplx became sentient, it would force us to use citations for EVERYTHING 😂
"what is your name"
"kim"
"HOW DO YOU KNOW; WHERE IS THE SOURCE; SHOW ME SOURCES!!!"

[rustic steppe]

ahem

[marble ice]

Quick Question: Is there some sorte of magic trick I have to do to get the „Finance Digest“ Summary?

[rustic steppe]

Why cant i generate ghibli images bro
like what do i have pro for?

[serene sand]

#💬│general message

[vague pagoda]

This is a feature of GPT-4o image gen, it isn't in the OpenAI API yet so it can't be added to perplexity

When it's available, we'll have it

[valid charm]

Good morning PPLX village

[tame current]

anyone got access of comet ?

[valid charm]

There are two users who have it and are on this server

[shrewd crest]

I'm trying to learn how to use perplexity for efficiently especially for promoting and other features

I've read the official docs but didn't get much information on how to prompt as a user

Is there other materials that are specific to perplexity

Official or unofficial

[valid charm]

Hmm.. seems like the conversations that take place here is the closest you can get

[cursive jacinth]

no latest news on DR sir?

[tame current]

daym

who

[valid charm]

Start reading here

[valid charm]

Some people reported DR High via (Complexity) is getting better

[serene sand]

deep dive ➤ #💬│general message

[distant aurora]

Thanks

[drowsy fox]

Claude 3.7 thinking is way better than Gemini 2.5 pro from my testing

[dense crest]

Waiting for the DR upgrade

[warm wolf]

Just got the student free ChatGPT pro and I have to say, perplexity you been slacking. The advancements they’ve had are insane

[valid charm]

How did you get this??

[warm wolf]

They’re offering 2 months free for students

Just search it up

[tame current]

google's early adoption tactics at college campuses now being used against them with PPLX/OpenAI

[left stream]

Do any do military discounts?

[silent pier]

no

[dense crest]

are you taking about the deep research?

[rain thunder]

I'm sure everyone knows about this

[shrewd crest]

I can't get my referral link

But I've registered and gotten the pro account

Can you help

[serene sand]

please contact [e-mail]

[warm wolf]

Deep research, basic interaction with AI which it gives suggestions on what it can do and image generation is a big one . Perplex image gen is seriously lacking

[meager flower]

Здравствуйте! Здесь есть рускоговорящие?

[valid charm]

Yes, there is a lot to be done. They have been stepping up recently. Let's hope it stays that way

https://tenor.com/view/fingers-crossed-gif-10027140023077878069

[green crystal]

I wonder when I'll see that I have the new dr

[narrow silo]

Did sentient AI mention me,
Or what is happening? 😲

[valid charm]

What do you mean?

[ruby heart]

I found out how to get unlimited manus for free

[valid charm]

How???

[ruby heart]

Mobile app

Hold on

I’ll send a tutorial vid

[tame current]

my math ai

has steps printing

(((1+(x^2))^-1)*(e^arctan(x)))
>>> Sdx
THOUGHT PROCESS = 

integrating (((1+(x^2))^-1)*(e^arctan(x))) wrt x
rewriting as (((1+(x^2))^-1)*(e^arctan(x)))
  substituting arctan(x)=y
  integrating (e^y) wrt y
  rewriting as (e^y)
  the solution is ((e^arctan(x))*(1^-1))
the solution is (e^arctan(x))

(e^arctan(x))
>>> ```

thought process being printed into the output console

(((1+(e^(2*x)))^-1)*(e^(2*x)))
>>> Sdx
THOUGHT PROCESS = 

integrating (((1+(e^(2*x)))^-1)*(e^(2*x))) wrt x
rewriting as (((1+(e^(2*x)))^-1)*(e^(2*x)))
  substituting (2*x)=y
  integrating ((2^-1)*((1+(e^y))^-1)*(e^y)) wrt y
  rewriting as ((2^-1)*((1+(e^y))^-1)*(e^y))
    taking the constant outside of the integral
    integrating (((1+(e^y))^-1)*(e^y)) wrt y
    rewriting as (((1+(e^y))^-1)*(e^y))
      substituting (e^y)=z
      integrating ((1+z)^-1) wrt z
      rewriting as ((1+z)^-1)
      the solution is (log(abs((1+(e^y))))*(1^-1))
    the solution is (log(abs((1+(e^y))))*(1^-1))
  the solution is ((2^-1)*(log(abs((1+(e^(2*x)))))*(1^-1)))
the solution is ((2^-1)*log(abs((1+(e^(2*x))))))

((2^-1)*log(abs((1+(e^(2*x))))))
>>> ```

integration questions.

such an invention i did.

[zinc quartz]

For those that wanted to escalate something to Aravind: https://x.com/AravSrinivas/status/1909284104530698595

Aravind Srinivas (@AravSrinivas) on X

Morning. What’s the number one issue on Perplexity right now that’s bothering you and needs to be fixed by us? Comment below.

X · Apr 7, 2025, 4:36 PM

[distant aurora]

Lovely. Hope it's in announcements.

@left stream chatgpt needs to fix this fast 😭

[shrewd crest]

I just sent a mail

[distant aurora]

They'll get to you. Through the crowd. Eventually. Hopefully.

[serene sand]

https://x.com/askperplexity/status/1909294939323924839?s=61

Ask Perplexity (@AskPerplexity) on X

Guess the final score of the game and win a free year of Perplexity Pro 🏀

Comment your guess below and tag a friend. If they get it right, you'll get Pro too. No entries after Tipoff!

X · Apr 7, 2025, 5:19 PM

[valid charm]

Seems like the production release is looming

no longer experimental but preview

[left stream]

@distant aurora

[left stream]

LOL

[left stream]

You checked if there was any updates available for the app?

[livid aurora]

Will there be a News API for Perplexity?

[valid charm]

What model is it?

[distant aurora]

Yep

[distant aurora]

Multiple personality disorder final boss.

Hemlock should have been Walter white.

Willow should be Quagmire.

[distant aurora]

4.5

[valid charm]

Care to share the jailbreak? 😆

[autumn raven]

give me the jailbreak please

[zinc solstice]

hello

[zinc solstice]

what app is this?

[naive sage]

Yeah, LLAMA 4 Maverick or Behemoth would be great for Perplexity to add

Maverick seems to edge out GPT 4o on benchmarks

And has a very large context window

[indigo valve]

They finally added Google Astra to Pixel phones. Pretty dope if you guys have a Pixel with advanced

[timid laurel]

Good evening; how do I prevent the AI ​​from always showing me in its sources an image that I sent?

[ivory bramble]

for my testing its really ass

[regal coyote]

How to change the model of pro on mobile?

How to change the speed of tts in the mobile app?

[signal hamlet]

Hey @regal coyote! If you are referring to the Android app, that will be changed soon - there's a model selector in the app settings. That model you chose in the settings will be used if you start a thread with Pro.

[signal hamlet]

Currently, there's no option to adjust the tts speed, but I'll forward this to the mobile team once again.

[regal coyote]

Thank you Alex, you're awesome

[velvet stone]

@signal hamlet I've uploaded a document for perplexity to quote from in Pro mode but it keeps making up quotes and ignoring me

How do I mitigate this

Its like it doesn't understand basic instructions

[signal hamlet]

Could you please share the thread URL in the #1140622008086970420. We'll check

[jolly cypress]

yo guys, i accidentaly have excluded all my topics (searchs)...do u guys know how can i get them back?

[timid laurel]

.

[signal hamlet]

You can remove the image from the 'Attached Files' section in follow-ups; it should work on the desktop version. As for the apps, this is currently on the backlog for future versions.

[jolly cypress]

?

it was today, like, 45 min ago

[zinc quartz]

Hey there! If you deleted all of your topics, there is unfortuantely no way to restore them.

[jolly cypress]

ooh, it would be nice to have like a trash can to find these deleted chats...

[dense crest]

https://www.perplexity.ai/page/google-s-ai-analyzes-complex-i-t9y3oZs5T9643xF6kYCP5Q

Google releasing ai mode w image

[simple salmon]

I want to live in a world where Perplexity and Notebook LM can easily be connected.

AUDIO AUDIO, imagine in Spaces being able to curate podcasts?! And expanding on partnership with ElevenLabs, have custom voices be able to read out answers (a narcissist like myself might want to hear myself answer my questions)

[distant aurora]

@left stream what is live bench smoking?

Putting Claude 3.7 thinking below R1 32b is diabolical.

[autumn raven]

its below, but the points that it gots is more than R1 32b

[distant aurora]

Look at coding scores.

52 vs 45

[autumn raven]

yeah, it doesn't make any sense

[deep sleet]

Garish vanity metric humping

...

[heady gyro]

Pro user here.

I am trying to upload 8 .pdf files to my thread and after 5 minutes of loading it either uploads one or two that instantly disappear with an error pop up saying "file upload failed". Tried individually and same thing. This also is currently happening on new threads. It's only been going on for the best 20 minutes but I've tried reloading the site/thread, shutting down and reopening my browser and restarting my laptop. About to try restarting wifi but I'm not hopeful.

I can do it on the Android app but on the website and windows desktop app I am having trouble

Anyone else experiencing this at the moment?

[vapid onyx]

Definitely not intended behavior. How large are the pdf files?

[heady gyro]

114kb to 9,502kb

[vapid onyx]

That's really strange. And this is on Windows devices where you're having the issue?

[heady gyro]

And the website

[gray prawn]

Man if I get one more ad about "Harvard and MIT are advocating for Perplexity" I am gunna lose it

Clearly all their new fundraising has gone between comet development and pushing that single ad campaign lmao

Plus it is not even true, it is a half truth, student organizations from those 2 schools have advocated for it, but they carefully choose their words to make it seem like professors are coming out and saying "use ai search engines!"

[gray prawn]

interesting!

[zinc quartz]

@sour sable you're welcome to ask general questions here

[left stream]

@distant aurora entitled people thinking they deserve everything in the world for nothing is crazy

I ended up finding a way to make chatgpt’s image generation do things it wasn’t supposed to and my post on reddit was soaring but then they were calling me a liar and saying it wasn’t chatgpt’s image generation until I gave them proof and I told them I ain’t giving nobody anything for insulting me by thinking I’m a liar 😭 @distant aurora

[left stream]

Damn, I understand o3 mini high since it’s op for coding and a monster at it, but claude 3.7 losing at coding that badly? Damn, r1 32B must be made with vibranium and all other fancy future tech 😭 also ngl, ever since gemini 2.5 pro came out, I’ve been using it with other ai models. I recommend for long context

[left stream]

First gemini model that finally processed my pdf file for once

[random ocean]

Any updates on Deep Research high?

[next wedge]

How do we generate images on the Android app?

[brisk patio]

Hi there,I am the one who is from JP and using PPLX for web research w/ the prompt that i made.From few minutes ago my PPLX returns the error message like "Sorry we cant do web research without comet browser. Use comet browser to avoid this kind of prob." (Not 100% accuracy cuz the original message was Japanese). As I remember this is very 1st time to face this kind of problem. Any help?;(

[distant aurora]

Have you tried notebook lm?

[distant aurora]

Send link.

I mean link of reddit post.

[distant aurora]

Bro the subscribe button on grok doesn't even work for me.

[tame current]

anyone heard about genspark ai ?

[kind badge]

Hi

[shut dagger]

on mobile or pc?

[crimson otter]

Is it possible to receive local or general news from my country? The only news I see is: 🦅🦅🦅🦅💸💸

[distant aurora]

Both mobile and web.

[boreal dew]

i could be wrong, but maybe try going to settings and then personalize
this shows up for me there
i mean it is just weather and sports but it is what it is😂

[green crystal]

im still also waiting

[distant aurora]

@left stream https://www.rxddit.com/r/DeepSeek/s/zFUYlP8NeV

[green crystal]

the counterpoint in this here is that grok got somewhere by just copy pasting allot from deepseek

[prisma lake]

Hi @everyone

I am using perplexity instead of Google search the recent upgrade pro and deep search when to use where to use I am bit confused it provides unrelated answers too while searching can you pls guide me

[nova wolf]

Is it just me but I'm not getting any notifications from perplexity?

From discover and it's enabled

[distant aurora]

Directly through source or training through responses?

[spring sigil]

I just love the design of Perplexity. The logo, icons, graphics etc. Amazing. I wish there was a wallpaper channel where we could access all the pictures that Perplexity uses.

[distant aurora]

Tomorrow?

@left stream https://www.rxddit.com/r/DeepSeek/s/7GODR60ylE

[green crystal]

Someone ask him when deep research high rolls out

[sick arch]

its already there, use complexity extension

[green crystal]

It just reverts to the current deep research

[sick arch]

then ask it on reddit

[kind badge]

#AMA heard nothing in/ for india after your last india visit with pm modi ? Is there something coming

[sick arch]

what do u want?

[kind badge]

On a broader perspective, about plex in schools/ colleges & for Indian companies, as AMA

On personal level, the free plex for students , but almost no college here gives .edu or any emails, so how to fix this

[sick arch]

most of the college gives .edu emails, its ur college fault on this.

[kind badge]

Most, and what about rest, any solution #ama

[sick arch]

u should ask this on reddit, aravind will answer

[distant aurora]

Anyone who's actually gonna talk to aravind please pass on the

How many r are there in terrrrific

Issue when using O3 mini.

I'm not sure if he's gonna respond on twitter today if I reply to his "what should we fix first" post now.

[slim star]

Hi.

[copper rune]

does anyone here use chatgpt projects? how do you think it compares to perplexity spaces?

[valid charm]

Good morning PPLX village

https://tenor.com/view/kuzco-emperor-emperors-new-groove-grove-groove-gif-5555210683836092267

[placid coyote]

I only read about claude's projects and chatgpt's custom gpts. it's been some time since I used (and paid) chatgpt, but custom gpts were more powerful compared to pplx spaces. instructions in custom gpts were real instructions, could result into running code (python sandbox) or do a web search. on perplexity sadly it is more about final report language/formatting, because space instructions cannot trigger "Programming" step (python sandbox) nor affect web search (eg whitelist or blacklist domains). also RAG in perplexity spaces is rather aggressive (removes most special characters) and returns rather small chunks, though quite a lot of them. I think instructions are also more limited by length on perplexity. depending on your use case, it may or may not be enough. I documented some things me or others discovered https://monnef.gitlab.io/by-ai/2025/pplx-tech-props

[proud basin]

good morning

[zinc quartz]

Mind sharing a screenshot?

[unkempt sedge]

I use Perplexity for fitness. When I try to upload a physique picture, I’m met with a “attachment upload failed moderation” error. What to do?

[distant aurora]

Probably cuz "skin"

💀

[unkempt sedge]

Jesus … I even asked before I purchased Pro if it could analyze physique and progress pics

[distant aurora]

Yeah, you might wanna try it on grok.

💀

Or chatgpt idk.

I have never tried body pic on chatgpt but it accepted a wart pic on my toe.

[unkempt sedge]

ChatGPT did fine, I just recently moved from ChatGPT to Perplexity because of the rambling ChatGPT did sometimes

Not really useful lmao

[distant aurora]

Tf happened to your chatgpt?

Lmao.

Give me the prompt.

I have never seen such hallucination on chatgpt before.

Actually, prompt kinda lacks structure too. Is your main language polish?

[frozen willow]

https://tenor.com/view/bugs-life-a-bugs-life-bruh-not-smiling-ants-not-smiling-gif-8815283262918161887

evening floks

[distant aurora]

Evening flok

[frozen willow]

[worthy bridge]

what do i do when my tmobile magenta moments yearly subscription just dissapeared randomly after a month

[frozen willow]

I'd recommend reaching out to T-Mobile support ASAP , sometimes the Magenta Moments perks can glitch or get removed if there's a billing hiccup or plan change. They should be able to reinstate it or at least explain what happened.

[worthy bridge]

like there was a new code on magenta moments (i didnt ask support it was just there) but i cant use it since i had used the previous code

[frozen willow]

so once you’ve used it, you’re locked out maybe. Classic T-Mobile move, i think i not sure sorry!

[worthy bridge]

ill just ask support i guess

thanks for help

[tranquil orchid]

Hey @worthy bridge!

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.

[frozen willow]

you're welcome miss!

[left stream]

Yes and it would fail. Even with 2 million context memory back then. Somehow it would never process my pdf. I guess it was too much. Ai studio, gemini, api, etc. doesn’t matter the pdf would fail until genini 2.5 pro released and it actually did the job

And damn it’s a good writer ngl

[valid charm]

How can I read the server rules?

[distant aurora]

#📙│rules

Lmao

I can't read it either.

[valid charm]

There were a lot of people exploiting the TMobile promo for selling 1 yr codes. They cracked down on it and banned a lot of accounts. This means you should reach support to get it looked at your case

[worthy bridge]

perplexity support or tmobile support?

[valid charm]

PPLX support

[valid charm]

[e-mail]

[vague pagoda]

good catch! should be fixed now - let me know if not!

[distant aurora]

Thanks

[distant aurora]

I tried documents on Gemini 2 I think. It sucked. I haven't used Gemini 2.5 pro much yet.

[left stream]

Use it, I had gpt 4.5 create a story against gemini 2.5 pro

I had 1 story only made by gemini 2.5 pro

3 stories by gpt 4.5

All random and no author name given to claude 3.7 extended on the native claude platform

Judged them and

Gemini 2.5 pro’s single story beat the other 3 stories by gpt 4.5

I compared them 2 at the time

I am a fanboy for OpenAI but this time I gotta give credit to gemini 2.5 pro

Oh yeah forgot the jailbreak for you @valid charm

Sending rn

[valid charm]

Yep, same thing happens to me when doing research and text translation. Too sad to se the OGs slip but nice to see competition driving the field forwards

[left stream]

Don’t leak tho

[left stream]

Yes, I feel bad for meta ngl

When one releases something

3 more companies release new models lol

[valid charm]

Yes lolol

[left stream]

O4 mini and o3 will be the trump card, let’s see how openai does with those two

I wonder how op o4 is

[distant aurora]

Compare grok

[tame current]

take a look at sympy and mathematica if you get time. because i have figured the secret behind those types of software. hype them also. along with grok and gemini.

this is a whole new subject. its not about being useful... even though it is useful.

it changes mathematics completely.

i think geometry and chemistry also... i am trying physics.

[distant aurora]

How do I access mathematica for free?

[tame current]

use sympy and maxima cas... mathematica is paid even though more powerful.

[distant aurora]

Who has a high end phone here?

[tame current]

you want to run sympy on phone ?? you can run it on termux may be. by accessing pip and python through command line.

soon my library also i will release. my library already i have.. but.. i will do some more improvements. then i will send it to everyone.

mahtematics equations are trees. ((1+2)+3). the brackets are the nodes. f_add d_1 f_add d_2 d_3 this is most central thing.

you need to be good at search algorithms and at manipualting trees in order to create your own library.

integration of sin(x^2)*x wrt to x

this equation is a tree too.

f_mul
 f_sin
  f_pow
   v_0
   d_2
 v_0
``` we need to manipulate this... take ``x^2`` as  a substitution.

[distant aurora]

Are you working on sympy?

[tame current]

i don't understand sympy's code. i developed my own library..

[distant aurora]

Oh.

Library of math formulae?

[tame current]

using this library you can differentiate any equation. for example. ANY.

>>> ((-2^-1)*cos((x^2)))
((-2^-1)*cos((x^2)))
>>> d/dx
(sin((x^2))*x)
>>> d/dx
((2*cos((x^2))*(x^2))+sin((x^2)))
>>> d/dx
((-4*(x^3)*sin((x^2)))+(6*cos((x^2))*x))
>>> d/dx
((-24*(x^2)*sin((x^2)))+(-8*cos((x^2))*(x^4))+(6*cos((x^2))))
>>> ``` differentiate a multiple times. its very fast.

or expand an equaiton like this

((x+y)^4)
>>> expand
((4*(x^3)*y)+(4*(y^3)*x)+(6*(x^2)*(y^2))+(x^4)+(y^4))
>>> ```

[left stream]

Me

[regal coyote]

new fresh chat, simple question and Perplexity gave me this reply: I'm sorry, but I can only assist with questions related to Perplexity or Comet. Let me know if you have any questions about these tools! 😄

[zinc quartz]

What was the question?

[regal coyote]

what is https://www.upvoty.com/ ?

model: best

🙂

selected 4o

[zinc quartz]

Mind DM'ing me your threads? I'll have team look into it

[regal coyote]

yeah

[idle zodiac]

Fyi... Manus Invites launched today ...

[distant aurora]

https://cznull.github.io/vsbm

This will make your phone lag.

[crystal cosmos]

🏆 Rank 1: Gemini-2.5-Pro-Exp-03-25 (Google)
🏆 Rank 2: Llama-4-Maverick-03-26-Experimental (Meta)
🏆 Rank 2: ChatGPT-4o-latest (2025-03-26) (OpenAI)
🏆 Rank 2: Grok-3-Preview-02-24 (xAI)
🏆 Rank 3: GPT-4.5-Preview (OpenAI)
🏆 Rank 6: Gemini-2.0-Flash-Thinking-Exp-01-21 (Google)
🏆 Rank 6: Gemini-2.0-Pro-Exp-02-05 (Google)
🏆 Rank 6: DeepSeek-V3-0324 (DeepSeek)
🏆 Rank 8: DeepSeek-R1 (DeepSeek)
🏆 Rank 9: Gemini-2.0-Flash-001 (Google)
🏆 Rank 9: o1-2024-12-17 (OpenAI)