#💬│general
1 messages · Page 304 of 1
and then for funzies trump suddenly cancels all tariffs and drops all pre trump era tariffs to 0%
market soars 30% in a day and gold becomes cheaper than silver
kesku i know you're reading this
cough cough custom role
Btw
Im Pretty Sure Airtel Will Keep Offering it in Their Data Bundles
Wait will it apply to goods built in US and exported to other countries
Why don't you
Hi
Probably Not
I Think
checkout pin !
Again, thanks
Hi
check the pinned msg in this channel
You now should have acces #👑│pro-lounge
OKAYYY!
NO i mean i thought i will be getting discord nitro for free
😭 bro what
got it
"pro" yk........uhh umm nvm
Why is it not giving any answer to my prompt..🙂
Nice
um, what? did perplexity put a limit on outputs or something
sometimes i select o3 and o3 replies saying "i don't know who o3 is"
i mean it did answer your question 😭
Hello guys, I'm having problems uploading videos on perplexity, the videos are in mp4 but still it's says "failed parse file" , I'm a new user so I'm not familiar with what type of videos it does support
make sure you don't have any custom instructions that are interfering
@haughty cairn Please don't advertise here
i didnt use it in a space. i tried again and it did give me a proper answer, i think o3 sometimes glitches
Are you trying to upload it in the android app? Should work just fine on the website
I'm talking about the information here https://www.perplexity.ai/account/personalize
I'd want short answers like that too...
It's monthly btw
Yeah on the website I tried a lot of videos with mp4 format and still fail to upload , I'm on android , pro user
i didnt realise there was something like that, that area is blank for me
Try disabling these - under https://www.perplexity.ai/account/personalize
Actually might be a bug, happening to me too 🤔 Should probably make a post in #1140622008086970420
i cant reproduce the glitches on demand though
Try disabling reference search libary - might be intefering
Yeah cause I've been trying for 4 hours now trying different formats and file sizes without result
Weird, was working for me
Use o3/sonnet4
Did perplexity really make a bid for google chrome??? 😮
which one of you is responsible for advising aravind to buy chrome 
Its a marketing strategy for perplexity. I guess
free press
They would be forced to sell chrome if the court order goes though
||donate it to mozilla foundation||
Bro can perplexity can make a website
sort of. by using labs
certainly not production ready
Nice update
cool, just got the update on my phone
I'm a plus user and I received this yesterday. I've already made the connections, good update. But nothing surpasses 3,000 messages per week in thinking mode. I'm using it now without even thinking.
Hello, I've noticed that when I answer, Perplexity sometimes runs scripts (not in Labs mode). Is there a way to prompt it to run a specific script when answer? In addition to Space for work, I have Space with instructions for RolePlay, and it would be great if I could ask the neural network to roll a die when it feels appropriate, rather than generating a text value. Is this possible?
I can just say "generate a random number between * and *" and get the same result.
That's not what I want.
In this case, I believe that this is a suitable reason for "make the product better "
@cinder mason is there any youtube video that can teach me how perplixity was made
Why ask me?
Hi
Still 35b is nothing
Free advertising for PPLX ig
For someone like you it's practically charity
Wouldn't it be stupid for selling chrome? It would have made sense if Google bought comet/pplx instead
Let's Get AI to Reproduce with eachother!!😏 😆
Demen rol
Legal issues
The govt pushing google to sell off chrome
Atleast the last admin
msft in early 2000s is a good reminder; dont be tough against gov kek antitrust is no joke
I need GPT5 Thinking
Not with current admin lmao
They don't care
If they actually do something with the FTC I'd be surprised
Placement is key 🔑
isnt this spam?^
removed
there is more in #👋│introductions too
Thanks
Who has an invitation to the comet?
how did i get the plexer role? 👀
thx
Share a Perplexity thread link in 💗│sharing and you’ll automatically be assigned the <@&1097674658779771002> role!
hii
Ethereum Hit 4.6k
I Knew it Would Touch Around 4k
But Didn't Hadn't Thought it Would Rally More then 1k From The Low

Does The GPT-5 Base Model Have Reasoning? Give Up-to-date Data
Yes. The GPT-5 base model includes built-in reasoning capabilities, with adjustable “reasoning effort” to trade speed for deeper thinking.
- OpenAI states GPT-5 delivers stronger reasoning across math, coding, multimodal understanding, and open-ended factuality, including a dedicated “GPT-5 Thinking” mode that the system auto-invokes for complex tasks[1][2].
- For developers, GPT-5 exposes a reasoning_effort API parameter with levels minimal, low, medium, high, letting apps dial down or up the model’s internal reasoning before answering[3].
- Microsoft and independent reviews describe GPT-5 as a unified system combining a fast model and a deeper reasoning model, routed automatically based on task complexity, with notable gains on reasoning benchmarks and complex coding workflows[4][5][6][7].
- Reported benchmark improvements (e.g., AIME 2025, GPQA) and reduced hallucinations “when reasoning/thinking” indicate explicit support for enhanced reasoning modes beyond prior base models
To clarify, the model variant "GPT-5 Chat" does not have reasoning capabilities.
Alright
O3 Does Right?
Yes
Alright Will Use Reasoning Models for Conplex Tasks
What Tasks Require Reasoning Anyways or is it Just Better and Best to Use Reasoning for Everything
does anyone have a comet invite 👉👈
Will the information edited in the user information be input into the model along with each input?
Does anyone know?
By User Information You Mean Like Name and Stuff?
STEM related questions
No, it's the place where you can set up a dialogue to describe you and help the model answer your questions
Yes
Everything in here gets passed to the LLM
Every time I enter?
Or just enter it once when starting a conversation?
Not sure, I would assume that the system prompt is only passed once at the start.
OK, thanks
I Have Many Suggestions fof Perplexity
I Think They Should Add Some Sort of Notion System So Anyone Can Keep Their Important Responses Organized
You can post your suggestions here #1111765634267742408 #1358914210339360798 .
I can post in both?
Just post in one of these is fine.
Gotcha Thanks
Updates to ChatGPT:
You can now choose between “Auto”, “Fast”, and “Thinking” for GPT-5. Most users will want Auto, but the additional control will be useful for some people.
Rate limits are now 3,000 messages/week with GPT-5 Thinking, and then extra capacity on GPT-5 Thinking
3000? Lol I thought they were dropping the limits back down?
also do u have an alert bot for Sam Altman tweets lol
The said it was temporary on the OAI website
Ive got notifs on his tweets


AI need to be the following in order to move to AGI:
- Longer context window (10M at least)
- Fundamental understanding of concepts(instead of predicting the next token)
- Faster replies
- Multimodality (text, vision, audio)

uh so now there's even more thinking options...
wtf is thinking mini

smaller GPT-5
The one that got used when you ran out of limits I think
?
on POE it can reason quite a bit
wait so uh did the think button count against the weekly limit?
It does
3000 a week though
what
You shouldn't be hitting any limits
isnt it 200?
isnt that temporary?
Yeah - I don't think they are going to decrease it
quite a jump
from 200 to 3000
oh
hmm
why dont they just use a daily limit ffs lol
Thats worse
In theory you would want really high weekly limits
or atleast a counter if its low
though 3000
yeah you're not hitting that unless you're using it like crazy
It's more than enough for most users
the smaller group that needs it is most likely pro yeah
Then problem that I have is that Plus users get lower reasoning effort than Pro
its medium right
Yep, corresponds to medium in the API
64 juice on Plus and 128 on Pro. High on API is 200 juice
well IF the limits are that high then im fine with it
if i really ever need high
POE has it for a really low point cost
and i assume other services too
Their credit system is meh
$20 for 1m points a month - less than 4000 requests on GPT-5 assuming your messages are under 1000 tokens per chat and you start a new chat for each message
damn im not sure if i used LLMs too much or uh
so the 200 was temporary until the rollout was over and then this new one?

lol
i mean the rollout is already done
everyone and everything can use gpt-5 now
Sam mentioned 3000 a week in a comment https://x.com/sama/status/1954604215340593642
I don't think they anticipated such severe backlash over the launch
(I still don't think 3000 requests for GPT5-Thinking is confirmed, I still suspect sama typo'd but we'll see)
It is confirmed....
Where though, I only saw some articles that reprinted his tweet
the site is probably waiting for every change to be in place before updating
at worst if they drop it to 1000 per week
thats still way more than o3 for plus users lol
and if the model really is just that efficient
they should put it in deep research so we can use more than a silly 25 per month

GPT-5 thinking is more expensive than o3
o3 we get in UI is different from the one in deep research no?
It's only via the API though, and extremely expensive :/
hi
https://youtu.be/pKwLS9PaeXY is this going to be a W or very bad
Nilay Patel, The Verge editor-in-chief, joins 'Fast Money' to talk the latest moves in the AI arms race between Google and Apple.
That's Sad oh well is still something
Give me summary
perplexity got money like that? And why would google sell anyways i feel like the chrome ecosystem is very dominant
court order
oh so google are being forced to sell chrome?
If they lose yes
ask comet
Comet is Not Available for Android
oh 😔

If Investors are Convinced it Means is Evidental They are Growing Revenue
These Investors are Business Mastermind Geniuses With Alot of Money for Researches of Teams of Hundreds
They Don't Bet if Logically, Mathmatically and Scientificly Their Probably/Percentage is Not High
LOL
no
investors can be dumb
not a minority if TSLA has falling fundamentals yet the stock is flying upwards most of the time
There are Countless Factors in Business and Financial Markets
Which General Society Normal Humans Don't Understand
Hello
In This World Nothing Happens Without A Reason
dude i would rather not take advice from some indian who capitalises every word and has the word genius in his name
are these the techbro dicksuckers
people who didnt pass 6th grade
LOL
aye dude are you literate
how old are you
discord is 13+
better get off do your chores
stop regurgitating things your dad told you
i dont care if he works in quant
Use Pro Discord Join Option From Perplexity
I did
Or Manually Get From A Staff Member
By Showing Screenshot of Your Pro as Evidence
The Automatic Role Giver Bot Has Issues Currently
Mm
@dry kestrel Help This Guy Out
Let Me Predict You are Probably Airtel User
Obv
Dude its like 17k
Please check pins
Fr..
And if it doesn't work, leave the server and rejoin through the link
He is Democratizing Ai
Congratulations
Im Pretty Sure Perplexity Will Be Offer in Airtel Data Bundles Like Hotstar/Jio/Netflix
Check out #👑│pro-lounge
Yee
hihihihihihi
lmao
If Perplexity and Airtel Colab Succeeds Jio Will Also Colab With Another Ai Company @jovial jolt
:__:
He saw it coming
I bet mukesh will put his hands into chatgpt
he knew
what now
The Higesh Probability
:__:
Lol
Sure They Are
I loke the vibe of it
Yes sir
Frr....

Issue with vencord?
Are these emojis copy righted
No, it's just that they are trying to use emojis from other servers
Nah
Im talking abt this server emoji
Cuz some servers custom emojis are copy rightsd
basically they aren't doing any harm
the emoji we use
converted to image
people can def seek the fake
Makes Sense
also they buy nitro not because of the emoji
if that's the case
nitro basic is way more cheap
they buy nitro is because of the custom profile
Can We Use Any Images as Emojis?
Completely Makes Sense
Understood
erm
(I need larger file uploads)
also i doubt you can be sue'd for pasting a webp image which happens to be of a certain size
Got edu mails for those looking
I have a doubt about campus partner program
If an existing pro user applies for "campus partner program" can his pro subscription be extended?
Yeah
using referral
upto 24mnths
it's not about the plan
it's a whole different field
in chatgpt, is 3000 messages per week for GPT5 thinking, is it for teams plan as well?
Yes
3000 is literally unlimited for me, and with two team member account i get 6000 per week, i think its big heaven
they increased context window also to 196K for gpt5 thinking in plus plan
why cant perplexity send files like chatgpt?
i think perpelxity should introuduce gpt5 thinking, as message cap limit is bigger number
team plans give u unlimited access to chatgpt 5
and a flexible access to GPT 5 thinking
Congratulations
❌
Now it Will be Easier for Me to Make Money With Stocks

hello
long time no see :DDD
gpt-6 is coming(JOKING), and PPLX HQ still isn't thinking about anything, since the price of gpt-5 reasoning is cheaper than o3! They really only need to think for less than a minute about this and add it to their website! ""** gpt-5 is cheaper than o3—so if we replace o3, not only do we get better results, but we can also make more profit** EZ | PZ ... but it seems they don't use their brains.
PPLX Net Worth is over 500M by now BTW....
and they act like a 10 man company..... shame.
i bet they will lose so many MAX customer by the next month
My man why you act like you should be provided all informations like school kid
It's simple math, if gpt 5 high is about that price ( high has thinking and because of the new logic they don't show thinking because all models except the gpt5 chat that we have in pplx have thinking)
If high is about that price, they can replace it with the outdated o3 that we have now.
Or if they don't plan on gpt high, they can add nano or mini, both better than there chat variant..
Are You Certain GPT5 High is GPT5 Thinking?
The default and essential thing for a burger is it's bun
There's no reason for the restaurant to put bun as there description info about the burger..
Same az new gpt model
Since all have thinking except the chat ond, they ditched the thinking lable.
Whatever Might be The Case
😂👌
They just simply don't care about there customers because they roll too much free pro sub gives..
Now pro users are like shit to them..
Perplexity Decides What to Do or Not to Do
We Can't Control it
Unless The Entire Community Demands it
I Understand Your Frustration
But We Can't Do Much
in the ChatGPT platform, when selecting "Think Longer", the model used seems to always be:
just buy chatgpt
that's it
Problem solved
If u want chatpgt 5 thinking so much
just refund pplx and buy chatgpt as u please
interestingly, for specific prompts (presumably when it decides to "automate" whether it thinks or not), you can get this model:
did chatgpt get a update
haven't checked
no copy and pasting images for pro users? or is it a bug
update for what?
Instead of fixing the issue... Yeah you right
You right, pplx should add shity version of api from now on... You pplx users deserve it
ahh nice i understand now
Was working fine for me an hour ago. You might want to post this here #1140622008086970420
yea with paid version, u can always have the "Thinking" enabled
what seems to be the issue here ?
There is no issue to be fixed?
Gpt5 was a replacement to gpt4.1
i mean the thinking mini
a couple days ago it was only gpt 5 and gpt 5 thinking
New update
They removed my good ol friend o3 pro
I had this conversation with you guys for the past 3 days. It seems that you don't want to have best things for your self and instead of agree with me about this because it's gonna be beneficial for all of us... You arguing with me just to prove that pplx ia doing right thing by adding bad version of gpt 5
I'm done
i think they replaced it with gpt5 pro
i wonder if Perplexity Max users can access O3-Pro still lol - cos doesn't look like ChatGPT users can
or they still got smt on their mind
It was there before the update today
ye i can
weird
maybe they are on to smt
but still it's good enough
How many times i need to tell you
the base version is the same version as Chatgpt
looks like ChatGPT decided not to put O3-Pro as legacy even
it just the RAG
Whats the update?
not sure
Updates to ChatGPT:
You can now choose between “Auto”, “Fast”, and “Thinking” for GPT-5. Most users will want Auto, but the additional control will be useful for some people.
Rate limits are now 3,000 messages/week with GPT-5 Thinking, and then extra capacity on GPT-5 Thinking
no one is disagreeing with you that it would be nice for GPT-5-thinking to be in perplexity - who is saying that?
before they only allow the gpt 4.0 to comeback
who ?
Awo i tought it's for pplx
its just not even letting me paste, idk how to give evidence
yea who is saying that it "won't be nice" if Perplexity included GPT-5-thinking in their models?
Just state the issue you are having, operating systrm, browser
i'm not sure what u meant
Everyone apparently
Too many models to choose from > Make GPT-5 and GPT-5 Thinking the only ones available > People miss GPT-4o so add that back > Go ahead and add 6 new models to the selector for better customization > Start all over again
People dont like drastic change :/
And low rate limits
I mean seriously, wasn't the whole point of this movement to decrease the need for selecting specific models and just having one that can do everything? Why is it that I go onto chatgpt.com and have to decide between 4 options for a single model
That 200 a week was pretty brutal for a lot of users - and then we have people using 4o as like a therapist
Just choose auto
Yeah I prefer having a bland, know-it-all as the base, rather than be stuck with 4o
?
Auto routes you automatically now - and you can turn off legacy models lol
Win win between casual and pro users
fr
They added most of the legacy version back
not for the good old pal o3 pro
Btw does anyone know how the rate limits changed for voice mode? I know they're giving plus users almost 24 hour usage for advanced voice, but does that translate over to video/screen record as well?
I dont think so
in a way, Perplexity is sorta like that - there are models which are fast and don't seem to think much (eg. Sonar), there are models which seem to think a long time (Gemini 2.5 Pro)..or at least takes a longer time to get answer, there are specific thinking models like Claude and Grok etc
3000 for plus users?
Yep
Cause I would love to be able to use screen record for hours on end if possible. Would make for a great study buddy where I can just have voice in the background and ask it questions
Now I Get it
Yeah that might not be possible
Wait, I can't get rid of legacy models for some reason. I can only hide the extra ones like mini and then the other legacy models but 4o is still here
Even Pro users have rate limits " Unlimited access to advanced voice, with higher limits for video and screensharing"
Perplexity is Just Waiting Patiently to See What is Happening With ChatGPT
slide me some pro
When ChatGPT is Done Organizing/Optimising
Weird might just be ui bug
Perplexity Will Decide Weather to Replace Models or Not
Bouta get wiped by Sam Altman
Do You Guys Agree?
Oh shoot I forgot I was talking on the perplexity server lmao
Was asking for chatgpt help in perplexity general
LMAO
We talk kinds of ai here
Black, white, green or blue
Not just perplexity
Purple
White
Hmmm
maybe i'm dumb or i was not paying attention
gpt 5 mini high would be comparable to o3 medium
Whenever Someone Sees A X or ❌ it Will Automatically/Subconsciously Remind Them of Twitter/X Which is Just Free Advertising
X is Also Linked to 10 by Roman Numbers
Bruh https://pcforum.hu/hirek/29728/perplexity-google-chrome-vasarlas-34-5-milliard-dollar-kamu
(Translate to English)
mashallah
yoo
How have we come to this? Is perplexity ever going to grow up? Remember when they wanted to buy tik tok haha
Just Witness What Will Happen...
All I want to witness are bug fixes and a clear, transparent communication with their users. Right now it's a sh#tshow really - they don't even put out release notes..!
I wish pplx made more sensible decisions and focused on it's core products
Instead of trying to do a different thing everyday
They seem to have benn lost deep within the ai tech bro rabbit hole
They're doing well but i hope they focus their efforts on where it matters
It's inevitable
Because it's a sanfran tech bro world
Well the last couple of months their main product has been struggling to work correctly
Comet works decently well
Pplx apps are so broken
They haven't even ported img generation to work well on android and ios
Not sure about that, Gemini is a more serious product - of course Deep mind is a scientific agency to its core
I have to force stop and restart app each time
Talking about mid level players
Yeah
That have to rely on gimmicks
Im all for gemini to work well
But it just doesn't have the charm of gpt
Yet
Hi
Hola
It will probably dominate, they have the models and the computing power
Thank God for that lol
Better than what? They're better than gpt 5 and Grok 4
It needs better tools like gpt has to it's arsenal
I genuinely think that gemini will eventually overtake oai
Just needs marketing and better models that suit every need
What Ai Models are Best for Collecting Up-to-date Real Time Data?
I Believe Gemini is Planning to Be A Allrounder Generalist Since They Don't Have Many Separate Models
They are pretty good. When they work. And their API does not bankrupt you
On perplexity? Probably the default sonar
It's optimised for that
No Like OverAll
Alright
Gemini pro too
So I Just Keep Using Sonar for Real Time Data Collection
Will Also Getting Gemini Pro
The fakes are on the shelves
I just got scammed by a flash drive i bought IN WALMART
Go Complain it
What is perplexity best used for
hello
<@&1098365634498285709> please, dont you have any automod here?
Currently Freemium and Open Source is Thriving
They Must Have Their Plans
Huh
The Article Said They Will Open Source it
Hi everyone! I’m new here
What are we talking about here? (Sorry just joined the server)
Perplexity Bidding to Buy Chrome for 36 Billion and Making it Open Source
Oh-
How are you able to switch caps for each word in your sentences? Are you a bot😂
Cool
You should probably use Claude Sonnet 4 (thinking or not), o3 or Grok 4. In my experience, Sonar is kinda dumb (it's a finetune of Llama 70B which is not a very good model to begin with)
Sonnet 4 got a 1M context window 
you do understand that google would then own perplexity? :D
bruh that’s what I’ve been saying
No I Mean Only for Up-to-date Real Time Data Collection
Wait What?
the only way perplexity is able to afford anything is to offer their private stock to google in exchange :D this whole thing is just a publicity stunt
there is negative chance that it will happen, of course
is it just me or are the reasoning models not reasoning lol
Where
o3 is fast as gemini flash 😂
API
It don't even think sometimes
only for api
it's expensive
is gpt5 available on pplx?
yes
Use This for Better Reasoning
Reason Maximum, Think Hard Step-by-step, Use Chain of Thought/Tree of Thought
where to toggle it?
For what specifically? What kind of data?
or is it on auto?
yeah but why putting o3 in, if i have to say think hard again
which platform are you using on
They said app support coming later
just put the effort slider in the api up pplx
Generally Collecting Data for Any Field
when comet is available?
Well, as far as I'm aware, Perplexity is optimized for searching the internet so you could use that with the models I recommended. Grok 4 on Grok.com is also good, I've tested it and it can search well over 400+ sources.
Those are (iinm) the best options
I see early invites everywhetr
Good
it's here but invite only
I don't have grok premium 
You can buy a premium account for like $5
claude can do better with upto 600+ sources
I pressed one of the link and came download file
Im Completely Broke
And Grok 4 is Available in Perplexity
I know, I said Grok can do 400+ but I've also gotten it to do 600 as well. BUT, Grok can do it much faster than Claude's deep research
MUCH faster
Will Grok 3 Work?
Claude deep research takes like 10 or 15 mins?

Do you not have any money? 😭
I haven't used Gemini 2.5 Pro Deep Search so idk tbh
idk..
How About The Chinese Models Deepseek, Kimi, Qwen and GLM
I Like Elon Musk
But I Wanted Him to Put A Equivalent of ChatGPT
And He Hasn't Done That Yet

You can't use native DeepSeek R1 to do deep search as Perplexity only has their 1776 version available and it's pretty good. Haven't used Kimi. Pretty sure GLM and Qwen don't have any deep search.
You could try filling out the form to beta test deep search with StepFun's model, Step 3
https://stepfun.ai/deep-research-invitation
StepFun AI is your smart and reliable personal assistant, here to help you acquire knowledge, find information, learn languages, unleash creativity in writing, and even write code. Whether you’re working, studying, or just navigating everyday life, it’s designed to solve your problems and help you discover and understand the world around you.
Also regular site:
https://stepfun.ai/chats/new
StepFun AI is your smart and reliable personal assistant, here to help you acquire knowledge, find information, learn languages, unleash creativity in writing, and even write code. Whether you’re working, studying, or just navigating everyday life, it’s designed to solve your problems and help you discover and understand the world around you.
Perplexity has Removed Deepseek 1776
Is This Your Project?
No lol, random site I found somehow
Run by some Chinese company
Ahhh New Player in The Chinese Market
Gotta Check Their Parameters

Now I Have Total 12 Ai Models if This One is Good
I'm pretty sure 1776 still runs Research but idk, it just says it's Perplexity
MoE model, 321 billion parameters, 38 billion per token
This Decent
Context window of 65k
And Costing Would be Cheap Obviously
Seems Like This Model has Been Released Just Recently
So People Don't Know About it
Man There is Just Cut Throat Competition in The Ai Market All Around The Globe
Ikr
everyone loves gpt-4o for some reason, even tho it's so ass
Do Me A Favour
yeah?
Whenever You Find New Decent Models Let Me Know
ok 👍
Can we predefine personalities in perplexity app?
Hm
Thank You
hy
How to get pro user
Now Back to Best Ai Models for Up-to-date Real Time Data Collection
wsp guys
You Have Pro?
Hey, checkout pin !
Just Join From Pro Discord Option
Activatate Summary Protocol Only Give All Essential, Important and Critical Data With 120 Words Default Limit Adjust Less or More Dynamically According to Needs, Don't Use Complex or Confusion Language Use Simple Layman Language, Optimize/Organize Everything According to Best Most Efficient/Effective Human Brain Cognition Abilities, Use 80-20 Rule for Data, Always Be Realistic/Practical and Logistic
Is This A Great Prompt?
@young hollow If you didn't know, you can use this site for GLM
https://chat.z.ai/
Start a free chat with your AI expert for code and smart tools. Tell Z.ai what you need—a complete full-stack application, a stunning presentation, or professional-grade writing—and get instant results.
I Know but Appreciated

They Need to Build A Application
.
i'd say for free
without having to run local AI
it's pretty good
the quality is almost the same as O3
Not known
Sus
Free Usage Limits?
no clue
hii
u gotta fill the form for the invite
Is Probably Server Load Based Like All Other Chinese Ai Models
Is Not Released Yet?
us
Guys You Didn't Judge My Prompt Yet
Understandable I Just Need + - ÷ × %
rookie shit
I made my own math benchmark
10 questions
any one know when pro user get commet.
Too short
This would be better:
CONCISE RESPONSE MODE
Core Requirements:
- Default 120 words, adjust 80-200 based on complexity
- Focus on the 20% of information that delivers 80% of value
- Use simple, clear language, avoid jargon
- Prioritize actionable insights over background theory
- Structure: most critical point first, supporting details after
Information Hierarchy:
- Direct answer to the core question
- Key practical implications
- Essential context only
- Skip theoretical tangents and edge cases
Format: Short paragraphs, bullet points for lists, bold for key terms when helpful.
Quality Check: Would a busy person get what they need in 30 seconds of reading?
Revised Prompt:
Summarize [TOPIC/CONTENT] using these guidelines:
Length: 120 words (adjust ±20% based on complexity)
Focus: Include only essential, important, and critical information using the 80/20 rule
Language: Use simple, clear language accessible to general audiences
Structure: Organize information logically for optimal comprehension
Approach: Maintain realistic, practical, and actionable insights
Format: Present as bullet points or short paragraphs with key takeaways highlighted.
Analysis and Feedback:
Character Count: Original: ~518 characters | Revised: ~445 characters
Critique of Original:
Lacks clear topic specification
Contains grammatical errors ("Activatate," "Confusion Language")
Instructions are jumbled together without clear structure
Missing formatting for readability
Vague terms like "Best Most Efficient/Effective Human Brain Cognition Abilities"
Changes Made:
Added placeholder for topic specification [TOPIC/CONTENT]
Corrected grammar and improved clarity
Structured using bullet points for better readability
Simplified cognitive optimization concept
Added specific formatting requirements
Made word count more flexible with percentage range
Areas Improved:
Enhanced clarity and actionability
Better organization for quick comprehension
More professional tone while maintaining simplicity
Clear output format specification
How You Guys Even Do This
nope
i got my own prompt
to improve it
Act as an Expert Prompt Engineer to refine prompts for Perplexity AI Pro. When I provide a draft prompt starting with '{Topic} - Review this prompt,' your task is to evaluate and enhance it for maximum clarity, focus, and effectiveness. Use your expertise in prompt engineering and knowledge of Perplexity Pro AI's capabilities to create a refined version that generates high-quality, relevant responses.
Respond in the following format:
Revised Prompt: Present the improved version with all necessary enhancements.
Analysis and Feedback:
Provide the character count
Critique the original prompt
Explain changes made and their rationale
Highlight areas improved for better outcomes
Refinement Questions: Suggest three targeted questions to clarify or expand the prompt further.
When revising, consider:
Clarity and Focus: Ensure the task is specific and well-defined.
Context vs. Conciseness: Balance detail with brevity.
Output Specifications: Define format, tone, and level of detail.
AI Strengths: Align with Perplexity Pro AI’s capabilities.
Expertise Requirements: Address any specialized knowledge needed.
Formatting: Use markdown (headers, lists) for readability.
Prompt Length: Prompts should be less than 1500 characters and not exceed the limits of Perplexity Pro AI’s Spaces functionality.
Your goal is to craft prompts that consistently elicit expert-level responses across diverse topics and tasks.
Thank You!
just don't fill it with hallucination like that holo guy
blud just sky rocket the ai to 40% more hallucination
anyone got chatgpt pro around here
i am having a seizure with perplexity's LaTeX
anyone knows a prompt to fix its buggy latex formatting?
If I ask Comet to describe a image (say with the image URL open directly, or visible on a website), it doesn't seem to be able to describe the image. Is this expected or because I'm not on Pro?
I was just asking this before I put this in #1393681130842226698 in case it was available in Pro
ohh yeah i want to ask can every pro user access comet or maybe some certain number of peoples have access of comet or what?
it's invite only for now
do you have comet access on invite
@twin grail do you know if the image thing is the intended behavior at least for now (regarding what I said just above)
can you please wait
is this a prompt or training data for perplexityh
I have it since the first day
hold on lemme read
It's because comet see it as a plain text
me also before that
i understand that part, but it could upload the image if it wanted if you get me
i once asked it if it can see the text was highlighted
it said that i can only see plain text
hmm
u mean capture the image for you ?
Send me the link
say you are on the url of the image directly even
For example, if you were on this page: https://images.unsplash.com/photo-1754766621748-2a96cbf56a1f
You can say, "what is the URL of this page" and it'll correctly give you the image URL
but if you ask it to describe the image, it'll either say it can't access it or sometimes make something up
like i get what's happening under the hood, i just thought for something like Comet it would make sense if it actually uploaded the image so it could look at it
is self promo prohibited here?
U can do it if you use the square icon
no i mean like
use the direct url access
Web interaction doesn't have the ability to do that even if it was morally allowed
but it's one of their privacy about this
Because web interaction doesn't have vision
you can do it in cool projects channel. Projects only
Try research and say download this png in your python then analyze it
Wallpapers count?
yes
You get boinked
it's not me, said sonic player above
i was testing it for him
yeah just wanted to make sure
it doesn't have vision, but why wouldn't it upload the file as you would normally?
u will most likely got instant ban
It has vision when uploading an image but it doesn't have vision when it interact with a web application you misunderstood mate
yes, but you're missing the point then 😛
Because when its interacting with a webpage its gonna acces its html and won't use vision, so the ai cant see any image if you use links
If you go here:
https://fabrikbrands.com/wp-content/uploads/What-Is-A-Non-Descriptive-Logo-1-1155x770.png
And type: "what is the URL of this page?"
It correctly gives the correct URL of the page. Meaning that Comet could upload the image if it wanted to (assuming the Assistant is on a Multimodal GPT that accepts images)
But i have managed to do it by asking the ai to download the content of the link first in its terminal tho, this only worked with ACTUAL direct links to the image
anyone have same problem like me? cant paste image from copy image to chat
my logic was like if you say to Comet:
"What does the picture up the top of the page show?"
it's not that I'm expecting Comet to actually screenshot the page, I was expecting it to grab the URL send that to the LLM and then come back with the result
And if you're on a direct link of an image, then this process would be even easier
@opal osprey peasants like you always make me do all the work
i know it'll work from Perplexity, but does it work from the Assistant?
It should they are the same thing, but go ahead try it i dont have my pc rn
Assistant is the same model as pro search models i believe
ill test now, though perhaps because im not on Pro it might not work, lets see
This didn't work at first until i made sure its using python so look out for that
"I can’t download files in this environment, so I’m unable to fetch the image from the URL as requested."
Your account isn't pro?
it's possible because i'm not on Pro my model isn't accepting the images
😑 😒
You are a rare breed i tell you that, almost everyone here has pro except for you
regardless, i still think eventually it would be logical that you just type "What does this image show" and it would just do the behind scenes work for you. I get the part that you might need Pro etc
The problem is it doesn't have access to the tool
Because consistent vision model is probably too expensive
that could be a factor perhaps yea
probably could be the reason
i might upgrade to Pro tonight
then i'll test around 😛 and i wont have to get you do it 🤣
LOL
But still there's one thing
If You Get Pro You Need To Get The Worth Back

what is it
Ofc we got it for free . So no complaint
Sup
do you guys also have issue with clipboard pasting images inside the PPLX?
i can't paste any image
but text no issue
Help me understand something about the browser Comet.
I began testing it today and i find something in the terms and conditions to be quite fishy.
The giving away a comprehensive lisence of everything im putting into the browser to Perplexity.
What does this mean in practice? If i write a book in google docs, using Comet, are Perplexity technically co-owners of my book?
I only have an issue with the assistant while connected with my car; Google Assistant works fine, but Perplexity does not work at all
They’re positioning for a buyout
That’s all this is. Perplexity is rapidly expanding their lineup of offerings, making obscene offers, etc to artificially inflate their valuations in hopes a company desperate enough to get halfway decent ai (Apple) or a company with more cash than they can deal with (Google, Microsoft, etc) gobbles them up. That’s why you don’t see as heavy pushes anymore to fix many bugs. It’s full steam ahead with new features, more features, newer features more more more and more
Api?
I haven't checked but api is more and cheap
@gloomy cipher
After You are Done Reading I Will Delete This Message Since is Too Long and Might Annoy Others

I Figured
Otherwise it spikes to 0.60
131k is max
60k and under is cheaper
That's the issue with kimi, its very smol
Hello
Did kimi k2 and o3 get an update recently?
Perplexity is damaging its reputation by making an offer it can't really afford to pay. It's worth $18 billion but is offering $34 billion to buy a browser—a sum it could earn only in the next ten centuries if we talk about revenue.
10 centuries? Really?
just wait till next year when all the indian people need to recharge there PPLX because they are attached to it... :DDDDDDD
Not gonna happen
Exactly
and if we get more models.. bye bye 20 dollar per month. hey 40 50$ pricing :DDDDDDDD
they have to up thier price so they can offer 10000 models that we will get for the next year or so..
These pricing models are not feasible in Global South
bru why to cry about this, we getting it for free so just use it no?
as always in the case of large companies, it is not about real money, but about the virtual value of shares.
For example, we agree among ourselves that one share costs $100, but if you try to sell it on the real market, it can only cost $0.00000000001.
I am also indian
yup, its not about valuation but revenue. valuation can be 20x but buying requires cold clash
Not really, they offer shares
to buy chrome they need to pay money not shares
they can't buy chrome ever considering their revenue and investment rounds
Ever is a very extreme word 😂
This an US server ?
no if you think logically
they need to have that much cash and today their revenue per year is $148 million not profit they need to have profit of at least $40-50B so they can buy chrome
even if they make $50 million profit per year it will take them forever considering inflation and growth of google itself
chrome can easily buy them but them buying chrome is impossible and CEO is doing it to create buzz which will fail within a month
It will soon damage credibility and people will consider its CEO as naive
somewhere i think its to signal to investors and buyers that he's confident that perplexity can pay itself off in the future
or hes just crazy

I Hope Perplexity Adds DeepSeek, Kimi, Qwen, GLM, Stepfun and Such Ai Models
I have the figma designs and i want to generate the bestest code for it can you help me with the bestest prompt and i am seeing through using HTML CSS JS
This is obviously a ridiculous move. Don't you remember they also "offered" to buy tik tok too haha
Hi
Gemini and gpt 5 with reasoning are both cheaper than sonnet and grok 4, basically same price as o3. I don't understand why they aren't adding it😑
Comet assistant can't create tables anymore
or give bolded headings
Can we please look into this
its so disorganized unlike before now
Hey
Sup?
yeah @dry kestrel told me
the limits got upped so much that even if they drop it at worst its 1000 weekly
I Wonder Too But Im Sure They Have Some Reasons
I Think Currently They are Just Observing What is Happening With Openai Once Openai has Probably Established All Things i Think They Will Replace it

Whatever The Majority Customers Demand They are Choiceless to Give it
That's How Business Works
I Think Sam Did All of Deliberately to Make it Feel Like They Listened to The Community and Gave Generous Limits
I Think They Were Going to Give These Limits Anyways But They Were Checking if They Could Get Away With it and Make More Profit if Not Main Mission Will Be Done A Illusion of Better Usage Limits

no bell canada offer for pro
anyone have a prompt that only looks at tweets on X?
Just asking it to use twitter should work, but you can also say "use site:x.com and ..." for example
Perplexity be cooking so hard


