#š¬āgeneral
1 messages Ā· Page 292 of 1
You can do work from home type jobs if you cannot physically work, and also im sure there will be like some sort of govt. aid or something if you want to continue your education till 12th so you should try and look for that as well.
We are So Broke Me and My Father Survive on 2/3$ A Day
Idk About Aids Even So it Will Take Long Time Before Any Earnings
Don't go with Laptop if you're Going With Video Editing
You can make 2X better Pc With Same money
Not exactly, you can continue your education while also find part time work or freelancing work in your free time.
and if you ask AI, it can give you some more ideas while considering your situation
Yeah First Let Me Strengthen My Body Then Pursue Education I Guess
But Yeah I Could Have to Do Anything in Emergency for Supporting My Family
@dry kestrel is it me or ChatGPT really upgraded their animation
Havenāt used on mobile app for awhile
Hey, does anyone know how to stop perplexity tasks?
hi
Hi
Bye
What to do on perplexity? Best features? What does pro even offer? I just got the membership for free
+1
deep dive ⤠#š¬āgeneral message
Is comet available for all pro users?
Im totally going in depressed
no
Pls suggest me best AI services with multiple models on board, like you.com, openrouter.ai, lmarena.ai, etc
Comet Browser hang on the iinstall stage, help me ?
I Just Know Perplexity
Acess to These Models Not Including Max Ones
only use gpt-5 and others ai via web browser for full power, never access via 3rd party apps because it's will probably detuned. š
Idk About Detuned But Im Sure About Token Limits
Um

Hi @everyone anyone having issue with reasoning model in perplexity they are not doing reasoning and giving wierd output
Yes thatās been happening for like a month
Numerous threads about it
I think they recently fixed Claude sonnet thinking tho
Hi, I used pro version of perplexity. what is the process to get the pro grade
but they have decreased the reasoning limit after 10-12 sec sonnet-4.0 also giving wierd output only
even with the complex task
@everyone to whom i need to tag so that they can resolve the issue asap ?
Any Max Users?
Do you mean the pro role? #š¬āgeneral message
Yes
kinda weird how they just gave pro away
Hi
.
not really jio did kinda same
hello, the pro plans limit gonna get lower and lower ...... they are just baiting for now
and maybe it's better to say something about gpt-5 to us after 10K+ messages about it instead of sending this :DDDDDDDDD
https://discord.com/channels/1047197230748151888/1047204950763122820
nothing since the gpt-5 release ! why do we have this discord if any topic feedback and... it's just gonna get ignored?
i spend 30$ not to see this type of stuff later.....
@royal bane https://discord.com/channels/1047197230748151888/1403076559207796927
+50 Upvote and not a single Mod or ... responded.... then why you guys are mod and... just to control kids in this discord??? :DDDD
well that's the primary role of a mod tbh
they aren't devs or marketing or support
Guys do check this out and leave a comment and, if you like the conten make sure to hit the clap button
Dude, everyone here is doing their own business. The main reason for that rule is to respond to people and fix issues, not just act like a king and reply every two hours with emojis! Such a shame...
link me the rule that "mods" (or anyone actually), has the non-negotiable obligation to respond or fix issues within a specific timeframe
demonstrate!
Yupp github copilot does have gpt 5 but you have to enable it from the copilot settings in github
Like, why do I have to be in this Discord when none of my messages are gonna get answered by the staff? Do I just want to chat and waste my time for no reason and get nothing back? We talked over 2K messages about GPT-5, and not a single admin or anyone responded to us. Are we just like dogs? What is this respect and customer support? They added Discord Pro as part of their setting page, so this is a main feature. We're here barking like dogs to them!
no one is forcing you to be in this Discord?
demonsrate tho
Moderators are here for server emergencies. Pinging them won't fix or add any features in Perplexity you might want, and reposting the same message over and over won't speed things up. The feedback is already there, and I'm sure the team is aware of it.
I can also say...I paid something....you should stop talking
it doesn't follow
non sequitur
have a nice day i don't see a point of talking here after 48h, and no replay even single one.
demonstrate!
have a nice one.
agreed
Havenāt I explained about why they canāt add the thinking model ?
Itās a simple economy math
As long as we or everyone want to it just simply canāt
You canāt request for a Ai sub with 20$ a month covering most of the AI models giving u nearly unlimited use per day
And still asking for the new thinking model when it just add up the price
I agree perplexity is about AI but they gotta make profits also
The cover for Chatgpt, Claude, Grok, Gemini
Iād say it had been a good trade
yo gang
bro get my point - they don't even leave a single message about the API and how does it work.....
how do you expect them to solve your problem if anything worst happenes?
Wdym? Leave a single message about the API ?
Explain to me.
hmmmm
like a hello guys we just added this and that this is how have a nice day.
they have this ecosystem and economy and they don't even update there https://discord.com/channels/1047197230748151888/1047204950763122820
the point is that they ignore user requests and suggestions that could potentially resolve bugs/lack of features
i really do hope they dont set the bar lower for pro users lmao
First of all, Perplexity mod in here isnāt really the one who got assigned to support everything, their job is to moderate discord server and guide people to contact support officially throughout emails and their contact section in your account setting
what did jio do now?
Hi iam perplexity pro user
ok thanks mate ā¤ļø
They do announce it? What do you expect about the API, explain details for me about it
How did you get that pro tag I'm also pro user btw š
go to settings in pplx pro and rejoin thru the pro discord link
All of the bugs, problem will be forwarded to the team who is assigned to do it
Ok thanks
The point of blaming perplexity staff in here is no use
leaveit
pretty cool stuff i dont pay attention to it
like i said, it's OK
.
the thing that pisses me off the most is the ridiculous lack of features if youre not willing to pay 200$ a month
a lot of the ai tools have so much potential but decide to overcharge their services
Anyone want Perplexity Max completely free on Android?
id be pissed if perplexity ended up that way
i use mac
that sounds sketchy tho
<@&1098365634498285709>
the catch?
oh MOOOOOODSS!!!
oh lord
mod....
What features do u want
I had to use
they just need to stop throttling down models on perplexity
not talking about perplexity but for example if you go to fuckingstudy.com its a website to generate fun ai videos, they limit you so much for the free version its insane
i got perplexity pro so for now im happy with it
only gemini up there works fine
Done
Itās more like perplexity RAG system may caused that
unfortunate
They donāt throttling down the models
It was most likely their system prompt that caused it
dang what did it doš
WHY IS IT SO GOOD AT CODING š
on par and in some cases better than chatgpt
*5
yo fr?
yes
indeed!
all reports and suggestions should be posted:
https://discord.com/channels/1047197230748151888/1140622008086970420
https://discord.com/channels/1047197230748151888/1111765634267742408
https://discord.com/channels/1047197230748151888/1358914210339360798
https://discord.com/channels/1047197230748151888/1393681130842226698
i thought its just some web search ahh ai
me too
but it searches only when needed
Oh
lmfao and that is?
oh reasoning
true
gemini takes the W for that
nahh bro 2.5 pro in my experience has been better than gpt at solving physics questions
huh
Who doesn't have GPT-5 working?
yeah
but anyway i use copilot rn just because it fits well with the system
i actually dont ever remember it being inaccurate in 5 months
not calling out perplexity it might just be a bug
but whenever i use the shortcut to use it in a quick access sort of way
it jumps back to the main perplexity window
which isnt bad but its just a little annoying
while copilot opens a little chatbox on screen
no matter what im using so it makes it a bit easier to use it while doing assignments
Hello š
welcome on board! 
deep dive ⤠#š¬āgeneral message
A few GPT-5 updates heading into the weekend:
- Now rolled out to 100% of Plus, Pro, Team, and Free users
- Weāve finished implementing 2x rate limits for Plus and Team for the weekend; next week weāre rolling out mini versions of GPT-5 & GPT-5 thinking to take over until your
What do you all think about gpt5, is it really that big update or just another ai gambit
GPT5 thinking mini version into perplexity please š
Like this.... they dont' even have blog page where they update us for this kind of stuf.......
does anyone else think that perplexity like downgrades their models in an effort to save money without us even knowing it
Then, bro, tell them: are they really not feeling bad when a 500+ line code is written out and you have to scroll up to copy it because they don't add a simple floating copy text or icon, just like GPT!!! OR When they allow us to upload 30 files in single prompt but we can only see 3 4 of them, no scroll or removing them all feature/button. !!!!!!
This is literally what I'm saying, and everyone is hating me for it... They should be more transparent about how they tweak the API.
i was actually hoping to be ridiculed once i sent that but its a shame that people agree with me
considering switching my plan to gpt or grok atp
@serene sand
can someone invite me to comet
i bought PPLX because of the real time search and sonnet 4 thinking....
but if they gonna always provide a weak API of the model and not telling us.what's the point of spending money? to not get satisfied?
also gemini and ... they offer you AI Labs and ProtoType - Firebase and..... which PPLX don't so they really are trying to only make profit here and not really care about the actual performance of the API they giving becuase " THEY ARE SEARCH ENGINE ",, funny.
Where's the more information button before purchase to show all of the API details, and... why should I find out about the limitations and not having the same model after chatting here? I wasted 20 dollars in the bin.
WTH
it just sounds like a silly problem (and frankly excuse) that needs addressing, not to point out no other ai platform has
hoping someone sues their ass and wins
check Tutorial ||
@serene sand have anything to say about this bro?
Like we all know, the Email and ... that you told us not gonna do anything. It's gonna go straight to their Spam: DDD or it's just a mail receiver, no dashboard.
If you are a mod, you have to provide me with a perfect answer, not just a childish reply, and send a general link !!! like search my profile you can see how much i chatted here.
i'm done
i have gemini and gpt
no worries
PPLX doesnt even care about you guys to even upload a single blog about there models they are adding :DDDD
and we're here arguing....
bye.
that tutorial is too unfortunately outdated
the context retention perplexity models often have feels im talking to gpt 3.5
Tutorial II answers everything you want to know, and I am sure the quality of what you ask would improve after you watch it
don't talk about quality man....
4k+ messages about the new model. not a single mod ... chatted here about this
quality my ASS D:DD
Have you actually used it ?
Or u just blaming everything
then export it ?
then you havent worked with AI fully at all.
go use gpt
you get what i'm saying
I have chatgpt plus so please stop with the childish
i have both team and plus plan
write 5 files more than 500+ line code in them.
then you get my point
even for grok 4 heavy and claude pro
Copy in Text Block not the entire chat.....
it's not sticky
so as you scorll
it goes away
so for you to copy it
you have to scoll all the way u
p
for like 1000+ code line file
do you feel this OK?
then OK
also see your attachement files when you add more than5 6 files in it.
no more talking i have work to do
bye'
check this too if you want
you'll get my point
we have entire extention just for this kind of messsed up thing that PPLX didn't add :DDDD
pc only atm
whats comet, complexity?
Oh š
Comet is a new web browser built for todayās internetāone that matches the way we live, work, and ask questions online. Forget the endless pile-up of tabs and fragmented workflows. Comet unifies everything into a seamless, intelligent experience, letting you move effortlessly between tasks, ideas, and actions.
With the Comet assistant, enti...
Did you tried?
Just How Many Subscriptions You Have...
You Surely Didn't Buy All of Them By Yourself?
Hi
Hey everyone, I came across a security vulnerability in Perplexity:
CVSS v3.1 Base Score: 7.5 (High)
Vector: AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N
Impact: High confidentiality impact, no integrity or availability impact.
I tried to follow responsible disclosure:
Emailed security@perplexity.ai ā They redirected me to support@perplexity.ai
Support told me to contact security@perplexity.ai again (loop)
Currently stuck in this support loop, so the vulnerability isnāt being handled.
If anyone here knows someone from their actual security team or has a better channel for reporting, please let me know.


.
Any Max Users?
hey all
anyone?
it gotta be fixed, š
Sorry Seems Too Complex for Me I Can't Even Understand
I just need a real support
@vague pagoda Help
the security team email redirects me to support, which re-redirects me to security team
That's A Very Vicious Cycle
š„²
is there a bounty program for the app
most companies have a place to report and get paid for it
How About Contacting, Aravind Srinivas
@prime siren You Have Worked for Alan Walker?
Nope, just an email:
https://www.perplexity.ai/hub/security-vdp
yea
Incredible What Was Your Work?
helped create music videos organised communities im in 1 of the vlogs too
i also founded the communuty on discord, did all sorts of things ranging from back end work to managing emails and social media accounts merch orders so much haha
i think my work at MEE6 was better though
too stressful and i suffered a bit of imposter syndrome ngl espeically when i was out partying in america with alan and vikstar
but yea i left that life behind for a quiter less stressful life tbh
Were the vulnerabilities fixed in the android app or is it still better to use the browser?
What?
Seriously?
That's Magnificent Achievement!
Now Im Sure You Must be Rich
How Did You Land Such Great Works Anyways?
What's the best model on Perplexity? Gemini 2.5 pro or the slower Grok 4?
Cause gpt 5 is crap at the moment
https://www.perplexity.ai/hub/security-vdp
Form is here
At the bottom
Tbh There is No Single Model for All Tasks Each Individual Model has Different Strength I Suggest You Figure Out Strength of Each Model and Then Use Accordingly My Go to Model is Sonnet 4 Base/Thinking
@dry kestrel does the think longer button use the weekly quota for thinking?
Yes
Touche
Touche?
TouchƩ
I think that's logical innit
FranƧais :
Salut la team ! š
Je suis super intĆ©ressĆ© par le navigateur Comet, mais je nāai pas encore accĆØs.
Si quelquāun a une invitation en trop, je serais vraiment reconnaissant dāen recevoir une ! š
Merci dāavance, vous ĆŖtes au top ! šŖš„
English :
Hey team! š
Iām really interested in the Comet browser but donāt have access yet.
If anyone has an extra invite, Iād really appreciate getting one! š
Thanks in advance, youāre awesome! šŖš„
Im So Confused
Ahhh Thank You
Is it True or Myth That GPT5 Inside Perplexity is Not Functioning Properly?
1 video and I hit a video max?
What do you guys think the response to GPT-5 will be? I know Opus 4.1 is out, but I want to know the version of Sonnetālike 4.1 Sonnet. Also, is PPLX going to provide it fully like Sonnet 4, or will it just lack the thinking capabilities of GPT-5 that we have?
Perplexity should make all the models available to the pro users. Maybe set query limit or something.
I don't think most companies operate purely reactively, since training models takes a lot of time for the most part it's just going to be business as usual and everything that's on the schedule will continue to be the plan going forward. GPT5 was reportedly an improvement on combatting hallucinations, better instruction following, improving efficiency, and slightly better all-around benchmarks - but I didn't see anything that would make competitors jump and say they have to do something imminently to respond. If there was a fundamental new capability then some might change plans to offer something similar, but for the most part it didn't shake things up that way imo.
Would be interested in hearing more details around why Anthropic blocked OpenAI's API access, and why they waited until now to make an announcement/decision about that... that was probably an orchestrated performance just in case it was good, but there was no need for that show imo.
I'm also curious to know if the Copilot GPT-5 is the same as PPLX because it's rated as 1x the capability of Sonnet 4, but it has much more understanding in my projects. So, if the rating is 1:1, PPLX should add the same model if they haven't already. Many people are saying that the PPLX GPT-5 model is weak and poor.
Since it primarily positions OpenAI in a better position for more agentic integrations, I think that'll be one focus of competitors going forward.
What UI or plugin is showing the 0x / 1x / 0.33x comparison? Usually when I've seen that on sites it's referencing relative costs to use the model instead of actual capabilities.
it's github Copilot on vscode.
Oh OK, so that just means cost/tokens to use the model, it's not necessarily related to relative performance or capabilities (except in the sense that usually more expensive models perform better). 3 o4-mini queries for the price of one gpt5 query is a good value for o4-mini imo
Good Idea
Another Optimal Solution is Pro Max Tier
What problem do you have?
Costing Less Around 50-100$
Model-gating is the new trend... Grok does it with 4, OpenAI does it with GPT 5 Pro, .... if companies are doing that then Perplexity won't have much choice to follow suit, since they have to pay much larger costs to access them.
Tbh idk i haven't used it that much but i have seen many others complain About it
Business is Business
Hmm
This is what Perplexity uses to make the search 32 times efficient and faster than others.
Is True if They Make All Models Available for Pro Users Most Won't Consider Max
Rather They Would Use Multiple Pro Accounts
Max 200$ Plan Acts as Making Pro 20$ More Appealing While Offering A Enterprise Plan
Other competitors had made all languages for free users included
Man openai subreddit is a mess
Of course, for investors lol
No people expected too much of a jump
That's Also True
Everyone is hitting a wall with llms
Gemini 3 won't get far aswell
What's important now is application of LLMs
So they're more useful than just toys or writing lol
Like Come on Man it Has Just Been 4 Years Since 2022
Everyone bit the belief that ai improvement is forever exponential
bruh this robot is so powerful
Current Industry has Hit A Plato We Would Need New Revolutionary Breakthroughs for Further Progress
yea let's study LLMs
You'd need something different from LLMs
Most Want SHI or AGI Now
Can we top this idea? Groq also offers faster inference but for search Perplexity tops!!
All These Models are Still So Powerful
yea it does hw in 1 sec
That Hope is Actually Driving The Business
Now it's time to increase practical use of LLMs and to make them more efficient so everyone can use it without low limits
We Still Have Revolutionary Breakthroughs Already
wait this server is only for pro subscription holders?
Nah
we need more meaw
#šāpro-lounge is only for pro users
Ai Skills Needs to Be Standardized as Part of Education/Course Systems
oh ohk thenks
Also Creating Diplomas
Meaw?
meaw is just uk uh
What Special it Has?
Meaw Bad Habit
oh
I was Used to Emphasize Words By Capitalising Key Words But I Ended Capitalising Most Words...
Writing and Testing extreme ethical dilemmas is fun
oh
wait tell me
can i change us market to indian market
in finance
yeah similar to Moore's law after some time the atomic limit for processors came and then its no longer possible to decrease architecture size further
No DMs promoting your products, services, or servers. You may post your own project in ā #šācool-projects only if it adds value and follows the no-spam rule.
I Think The Finance Option Automatically Adapts to Your Prompt
Solution Check With A Prompt
lets vote
or 
Thank you perplexity
paywalled vote š
I can vote but people who don't have nitro can't
An actual vote is here
didnt know this is paywalled
Why despite the pro subscription when i select GPT 5, Perplexity use 4o ? š¤”
Anyone seeing their Top Sites disappear from the home screen on Comet?
How do you that it's supposedly using 4o instead of 5?
Asking various questions with at the same time asking wich model was used for repplying to the question
The model itself doesn't know what it is
It is using GPT-5 even if it may say that it is 4o

new changlog is here (since yesterday)
https://www.perplexity.ai/changelog/what-we-shipped-august-8th
For select US equities in the Research tab of Perplexity Finance, weāre starting to roll out original research reports generated by Perplexity Lab
Well that's short
Oh alsooo finance watchlist finally supports non-US stocks
JuMp ScArE!!!!
WHAT THE ACTUAL F I WAS NOT EXPECTING THAT SINCE I HAVE SEEN THE VIDEO
can someone tell me a free AI video generator?
NGL is Hilarious Tho
The Doggo Just Made Her Cake and Eat it
What A Level of Unreal Animal Magic
I Wonder What Would Be That Dogs Reaction if You Show This to it
I would assume yes
o
Especially if it's GPT5 Thinking + Think
can i build gta 6 on perplexity?
the thinking button won't pop up if you chose gpt5 think
Oh did they fix that already? There were people on Reddit who were saying they could select all kinds of weird combinations like that.
No, Perplexity doesn't include compilers to build any leaked sources of video games.
8d?
I can't trust the GPT-5 in PPLX. What about you guys?..... It feels so cheap to me.
lmao , what do you expect
Tbh idk im just using Sonnet 4 for Now
Something Mind Blowing
Is the sonnet they are giving the real fat api model from Claude?
I think so
Love sonnet
i just let it default ,
Google is gearing up a new India-specific āHeliumā Google AI plan that comes with a big discount. Earlier, OpenAI started working on an India-specific ChatGPT Go plan as well. @distant aurora @pseudo ferry
What side skills should i focus on, i am a student and actively preparing for engineering entrance exam
I can't Understand
it give me best result
Yes they use the official API's
there is no where else to get claude lmao
AI, AI and AI
Like I like Sonnet 4 a lot, but I don't know the difference between the PPLX version and the source API from themself.
I mean the temperature limit.
Im Pretty Sure is Same as The Original
token limit and...
It's pretty much just the context that get's capped, and they might tweak things like temperature
(its pretty similar)
Essentially Main Key Skill is Prompt Engineering Then ML if You Want to Work in Ai Field
whwer can i get free api /
32k Default Context Window 4k Default Output Token Limit and 8k Input Token Limit Increased With File Uploads
yoooo
Thanks, does GPT-5 have benchmarks out now? I want to see the difference between GPT-5 and the Sonnet 4 testing.
You Can Sonar API for Free if You Have Pro
Is 32K enough? Like they gave us space, and I uploaded a 30K line code project in it, can it handle that much and still give me an answer for that?
most of them were updated on day one
Oh i see, how to get started
ppx pro or aonar pro
nike pro
thanks
After 32k = 128,000 characters The Model Couldn't See Conversation Before 32k = 128,000 characters
there is no subscription called Sonar Pro - you get $5 monthly API credits as part of the PPLX Pro subscription.
Perplexity Pro, Sonar is Model
128K is nothing :(((
okkk
Before 32k Get Summary Export as MD/PDF Re Feed for Fresh Context
You Don't Need Entire Conversation Context
Just All The Essential Important Critical Data
How can I know when I hit 32K? And if I export the MD, doesn't it consume the same token when I upload it?
I Don't Currently Have Solution for First Question as For MD That is Essentially Summary of Critical Data Far Less then 32k = 128,000 characters
I'm pretty sure that it tells you that the conversation is too long, also you will start noticing hallucinations happen way more once you start aproaching that limit.
If the LLM starts giving you nonsense, made up data, etc then start a new chat
Conversation Too Long That is True for Web
You Can Also Ask The Ai to Tell First Query of The Entire Session
From First to Last That's Current Memory
Hi
Or, from what I've seen, if it starts repeating data that it says, as if it's new... because I prompt for lists (and on free tier) I can see it repeat itself after awhile (even saying not to include items already mentioned)
You Airtel User?
Gemini 3 release date?
Nah jio
Can it be managed with a rule?
Not confirmed
Thanks, is there any rumors?
Ideally, if your conversation is organic, it will just forget the old stuff and you won't really notice when you hit the context limit
What is Average Characters of Your 1 Query?
From what I've seen, there isn't any solid evidence, maybe this or next month?
I Think By Then Previous Data Might Not Be Needed Anyways
Great, thank you for the help!
Sometimes, yeah. So for many use cases it doesn't matter. It's primarily for big projects like writing short stories/novels or coding projects or dissertations, etc. Or something that has to analyze a lot of stuff to start with.
nothing just " analyze my space : [ @myspace }
but the space has txt files with more than 2000 Line code for a project.
I only run 15 to 20k ... then new chat
Input Limit Doesn't Matter Because is 32,000 tokens (which approximately equals 128,000
I Need to Know Your Average Output Characters for Each Query While Coding
Hiii all
I can save all my work on space?
i export them as txt file
I see
If I mentioned my space current chat it can recall my chat?
No
Each Conversation has Seperate Memory
I Tried Giving Link to Other Conversations But it Didn't Work
PDF/MD/.TXT Work Best for Feeding Back All Context of Any Conversation
You can turn on "Reference search library" to reference old chats. https://www.perplexity.ai/account/personalize
Then It Can Access Data of All Conversations?
š®āšØ 2025 .. company still able to manage these things... Useless
Seems Like Im Wrong
It can then pull information from old chats.
Spaces will always retain whatever the description of the space and special instructions about it even if the context is too long - it is kinda like memory in ChatGPT - so yeah, it's dedicated to that space.
Like It Can Access All Data?
Wait Spaces Has Longer Context Window?
What do people like the least about @PerplexityComet? What can we fix to make better?
I'm not sure if it can access all your data especially if you have a lot of chats, it def works
I Assume it Will Pull Data From Old Chats According to The Keywords in The Prompt?
No memory is more like those special instructions are sent with every request, it basically always 'remembers' those specific things/special instructions (or at least resends a summation of them every few messages)
you will see it reference an old chat as a source
Well This is Still Really Useful Thanks Alot Man
Thats Reaffirmation Protocol
It Just Keeps Reaffirming Any Data You Tell it to in Every Query/Prompt
Hey guys I'm new here I was having an issue with the installation of comet browser can anybody help?
i was implementing psych1. it just converts variance of sum of random variables and a formula relationship. using summation.
variance(X)
>>> psych1
(sum(variance(X[i]),1,len(X),i)+sum(sum2(covariance(X[i],X[j]),1,len(X),i),1,len(X),j))``` from psychology's factor analysis.
very little work.
@dry kestrel Does Your Profile Hat Has COD Mobile Logo?
This is a link for the offline installer - that seems to fix the installation problem:
#ā”āask-community message
Yes I think?
You Play CODM?
No
You Just Like The Hat?
yes
Bloodstrike?
It was for free
That's Very Nice
Hi Berdo
Hi Hero
berd used that hat to hide something
nooo
I Play FPS Games on My Phone with 5 Fingers Claw
Is it legit does perplexity allow it?
How's y'all time with gpt5 been?
Sunflower seeds
Ahhh is Great CODM is Also Great
Idk Haven't Used Alot
Hi Kesku can we add this into your message here: If you are having issues with installing Comet, try using the offline installer here https://www.perplexity.ai/rest/browser/download?channel=stable&platform=win_x64&mini=0
codm my freinds play that's why i also, or else i don't play anything on phn
I Tried Using it But I Will More Comfortable With Sonnet 4
oh nonononon
i ate it
Ahhh Makes Sense for The Sake of Friends
yea for those noobs, they wanna die so i fullfill their wish
LOL!
Yeaa, they should embrace it
thanks
:(
It's directly from Perplexity, and I believe lots of other people have used that link š
Also I Just Don't Know if The Perplexity GPT5 Actually has Any Issues or Not
anyone used comet ?
Waiting for Android
Is it different from the email link or what is it? š
I Wish Phones Could Operate Windows
I miss claude
the file is distributed by the perplexity
It's just an offline version of the installer, the standard installer actually downloads within the installer, this one has all the data in the exe you download
Works just fine
Why?
Claude got the vibes
Are You Confident it Works Same as Original GPT5?
guys guys.... check this out
https://www.instagram.com/reel/DNI66jehTc3/?igsh=d2IyNzl1OWo1YzFo
Directions by Aiyana haha
#aiyana #cometbyperplexity #comet #perplexity #aigirl #avatar #aiavatar #viralreels #trendingreels #trending #explorepage #explore #fyp #explorepage #reelitfeelit #reelinstagram #instamood #instagood #reelsinstagram #ainews #googlemaps #google #automate #workflow
Well is Still Available
Ooohh okokkok thanks man š«”
Yeah? Only difference is that in Perplexity it can't decide to reason
betrayal of kimi
Guys, is the GitHub Copilot GPT-5 the same GPT-5 as the one we have in PPLX? They labeled it as GPT-5 preview!
Kimi still the goat but it's not in Perplexity...
keep pinging kesku and you will get it
Refer to my vibebench
i just copied the link from @fast meteor
(if you're running Windows this will work fine, as that link is for the Windows installer)
Should I make it my default any opinions?
what happened to grok 4 š
Will do
Wait Why?
I think it's self-explanatory
poor grok
If I was using a mac I wouldn't be using discord
I'm using discord on a mac >:(
https://x.com/AravSrinivas/status/1944214853767639489
nothing to share beyond what aravind said:
Kimi models are looking good on internal evals. So we will likely to begin post training on it pretty soon. Congrats to @Kimi_Moonshot for delivering an incredible model.
same
I'm still hoping for the best
mac users rise
Every day I go to sleep praying that Kimi is added the next day. Every night I see dreams of aravind tweeting that kimi is available.
why no ones talking about comet, its literally like jarvis
Fake Mac user
hi
Unless Kimi Get's Widely Known Like DeepSeek That is Very Hard They Even Removed R1 1776
Ya I wanna hear opinions
arent you are pc user as well
Ngl R1 was pretty outdated
Part-time
What happened to Deepseek r1
Betrayal nontheless
I Guess I Just Use The DeepSeek App for The Latest R1 0528
I think Kimi k2 beat it to the punch in general
There is 128K context on gpt5 of max plan?
It was Just Removed From Perplexity
Who am I betraying
Like across the board
Tim Cook
Guys, is the GitHub Copilot GPT-5 the same GPT-5 as the one we have in PPLX? They labeled it as GPT-5 preview!
:((( ?
Not just perplexity
What a fun command to run
There is 128K context on gpt5 of max plan?
Hey guys I downloaded comet but I really don't know what to do with it š„² š„ŗ
Oh no
I'm sure that won't do anything catastrophic
No - at least nothing official regarding this
No way 0.0
I Think Token Limits are Almost Same for Both Pro/Max
O.O
I remember tim cook giving a memento to trump during the announcement of the $500 pending š
32K limit on max
@short narwhal is Pplx Sonnet Different Then Orginal Sonnet?
Damn
The memento reminded me of glass ui
Pro has Same
I Feel Like You Need to Optimize Token Limits for Both Pro and Max
They're the same model
AI is still ass
idk
tried to find resale platforms
for tickets
like ticketswap
and it didnt offer me anything other than ticketswap which doesnt have the necessary events displayed
At Least 2030 for Amazing Performance
i just managed to do it myself via the banking app
They Perform The Same?
no need for AI
So... Can we classify the models into three tiers
literally AI is just a tool for coding/excel typa shit
AI is Not Necessary for Everything
summarizing and shi
Like how phones are
Writing
yeah but google did it better
than AI
Budget, midrange and flagship
for mac just replace it with mac_arm64
especially perplexity
Not per se, Sonnet in Perplexity is focused on search
Not the millet again
Well It Will Take Some More Years
So is Tuned?
ā¾ļø ā¾ļø ā¾ļø
Btw Any Max Users?
Why are they all vegan
so yeah pplx is ass in search if it cant replicate same performance as google
Yeah well thats not exactly the model's fault? I'd have to tell it what event Im looking for
I can't understand why you feel that way
Perplexity Pro has Been Amazing for Me
I Just Needed Prompt Engineering
because youāre bird
it is model fault? i didnt even specify what event im going to
are you stupid
š
meat for humans only
Thats your fault.....
that aint my fault
Yeah basically
lol
You can share your location for better answers
Ahhh so is not same performance i guess that's the issue with GPT5 Complains
I Guess Perplexity Models Are Just Tuned
Maybe
LLM's searching and google search is very very different
your species vegan
aight buddy
Birds Just Eat Grain is Simple
tho maybe some worms you like birdy
You need to at least tell the model which event you are going to for it to give you something useful
somehow google did it better even tho i didnt specify it
It's theur in house custom model
Because it has Access to Your Location Data...
cough cough they work very differently
PPLX has access to location as well
that doesnt take the fact AI performed poorly
LOL
I Don't Think Perplexity Stores You Locations
The whole point is that LLM's are different to google search
If you don't give enough context, how can you expect a good answer?
:/ Gives me even a section of options specifically for my area
none of these work
" i didnt even specify what event im going to"
That's What Prompt Engineering is All About Essentially
i didnt say i need to find specific event
i told both google and AI
viagogo charges you lots of fees š¢
Looks like a hotdog
"i didnt say i need to find specific event"
yeah, i didnt say i need to find specific event
lol
"dont have the event i need"
cuz u listed sites
google will infer you location based on your activity even after you turned off the location
that possibly will make me choose the event i can resale tickets on
bro is stupid š
EXACTLY
Without Prompt Engineering You Can't Utilise Maximum Potential of Ai
Lmao it's you that is complaining about AI not being able to find the event you want and also you who didn't provide enough context
No harassment, hate speech, or malicious behavior. Keep things positive and constructive.
You're just stupid. So explain to me why do I need answers from AI if it can't even suggest a website where I can resell my ticket
Yes, my answer would be āGoofyAssSale is a website where you can do that,ā but the thing is, you won't be able to resell your ticket for the event you want on those websites
but if any ai service does that, i could image how users will react to it
No harassment, hate speech, or malicious behavior. Keep things positive and constructive.
Is Not Viable That's Why is Not Practiced
You said yourself that ""i didnt say i need to find specific event" and then you say the results "dont have the event i need".
Bro is regurtitating same shit he been saying but sending exact same screenshot of prompt and answer that I got even before going to this platofmr
Google = Trusted Other Services = Doubt
So maybe add more context to your query and you will find the right result
Insulting me does nothing lol
Because I need a website where I can resell a ticket for an event I want to attend š
So say that in your query?
be specific?
I did?
Maybe Try 2/3 Tries At Least
Nah it spews same shi
"i didnt specify both google and perplexity what event im going to"
yo if the comet browser voice assistant had the ability to control over the browser then it would be legit š„
omg
It should be able to albeit bit worse than the text input one.
berd's brain fried from usage of AI and mf be defending it
Alright Guys Enough Fighting
try out openrouter, they have free models there
bro probably believes that AI's also think
Stop Washing Time and Get Back to Your Work
No harassment, hate speech, or malicious behavior. Keep things positive and constructive.
I dont understand shit there but alr ill try it out
thanks
It's you who didn't provide enough context in your query, complains about it, then attacks people who are trying to help you here.
i wonder what was the message
you aint helping shit
lol
You can filter for free models by dragging the slider here
you being pain in the ass by throwing in screenshot while not knowing these services don't provide most of the events
š
unless u actually believe AI went to those sites and put in the searching index bar the exact event i need to attend to
You never shared your exact query?
if you're paying more than $20 a month for AI services then you're doing it wrong
bro thinks imma throw in my country
20 is cheap
i made comet browser paper trade automatically lamo š
yea exactly
Openrouter is a website that offers API key with integration of basically all the models. The free ones are the ones that has the name ending in ā:freeā
Guys, is the GitHub Copilot GPT-5 the same GPT-5 as the one we have in PPLX? They labeled it as GPT-5 preview!
bro gonna wake up and see PNL -100
oh
very helpful thanks
i still dont understand shit but ill look for what you guys helped me with
I'd recomend looking through the documentation here https://openrouter.ai/docs/quickstart
i know dude i am not that dumb to trade real money š
yay i got it
Bruh. https://openrouter.ai

The one in Github Copilot seems to the reasoning variant (I'm not too sure why they added preview onto it)
and it's 10$ a month..... (ofc I know it's 300 prompt but if you know the tricks.... )
They have their own set of issues which makes me hesitant to use it.
Like, why doesn't PPLX update us about which model they gave us? I'm going back and force between sonnet 4 thinking or the GPT-5
I'd stick with sonnet 4 thinking for advanced queries
GPT-5 for quick answers
ok and how about the visual? like front-end visual stuff. is S4 thinking good?
If you have to use PPLX to code, stick with sonnet 4 for now
sure. Just cause I used GPT-5 in Copilot, I was confused about using it with PPLX
Ah
Yo I have Airtel sim card and got Perplexity Pro For FREE
So far I liked the results from GPT-5 reasoning
:DDD
Guys which ai model is Fun to use in App
You're The first person and only person
Depends on purpose
I think it did a pretty good job
Well you can add me to the list as well

Vibe with Genz Guys
People just extremely expected way too much
I have a bad feeling that PPLX will put a limit on pro users after a month or two of the Airtel hype, and... to control their cost..... only the max users will be left.
I can't see anything
I mean with grok 4 and Gemini 2.5 pro, it was already very clear that we are not going to see any major improvements to the model themselves, but we'll see more improvements to the usability
Gemini is Good at Generating Visual Data
Yll jailbreaking gpt 5
If They Put More Limitations Users Won't Stick to Airtel and Go Back to Jio
what is Jio
Jio Could Potentially Take This Opportunity By Giving Better Offer
Actually think most of the outrage is actually what I've seen so far is because they've removed the old models
Just slowly phasing them out would have been better
Jio is The Competitor of Airtel
Yep, it's not so much that "GPT-5 sucks", people don't like drastic change and thats exactly what OAI did
Jio Gives Unlimited 5G Data While Airtel has Monthly 300 GB Hidden Limit
Yeah Sam banked too much on people being fine with improvements
On the other hand, they had to at least ditch 4o if they want to have enough capacity to give all free users gpt5 mini
Because so far the ones that are truly rejoicing gpt5 are the free users
@dry kestrel Do You Think Perplexity Models are Tuned?
I know a lot of people who only know free chatgpt and they are mind blown
I know that sonar is post trained on Llama 3.1 70b. The other models I don't think so
How do they have gpt 5 pro if the pro api isnt out
So They Give Same Performance?
This is misleading - they are using GPT-5 on the high reasoning setting
Scam?
ah alr]
Yep
@KajetanMastela you have all the options
GPT-5 - minimal/low
GPT-5 (think button) - GPT-5 medium
GPT-5 pro - GPT-5 high
Well That's Assuring You Think Perplexity Will Put More Limitations on Pro Users?
okay that explains it
They are Deceiving Through White Lie
If gpt5 really saves a lot of money then I can see perplexity implementing their own router
Which of course is going to be worse

I honestly cannot answer that, I wish I could predict the future/what their internal team is doing
hi
Justified
most pro person are airtel users lol
is it still around 10 dollars per 1m tokens
Well It Will be Whatever it Wil be
Yes
Router Wdym?
kk!
Don't they already have the "best" option
Nice Poetry
i ain't poet
I Said Poetry
oh in bio
right
Come On Man
No, it will get renamed to something pretty cool
From Past Patterns Perplexity Ai has Mostly Been Improving
I Hope and Wish it Continues
Best Ultra Pro Max
Big Pplx
Pp and Lx
Perp Laxity
Guys, is the GPT-5 mini or nano better than what we have in PPLX? because they have reasoning....
Is there any page, blog, or any god damn article about the PPLX API's? Like I want to know how much input and output they set for their sonnet 4 thinking or other models....
Pretty much all the information you need is here
e
I don't need to get the API from them; I want to see what input and output they have set for GPT-5 or Sonnet 4, for example. Is it in this link?
What do you mean by input and output set for GPT-5
h
There is No API for Pro Users
you get $5 worth of API credits as a Pro user
Yeah for Sonar i Forgot That Sry
I thought your tag was for kimi antonelli š
That is Valuable
I don't know how to explain this.
It seems like there should be some difference between what the Claude Users get from the Claude subscription for the input-output tokens and what we get in PPLX. Nah?
The system prompt, temperature, top max output, Top P, is going to all be slightly different.
Input Base 8k to 32k with Files
And claude gives you a 200k context window compared to PPLX with 32k
then it's shit compare to the actual sonnet 4
(although you get like 5 messages when you do reach 200k on claude)
65K compare to 4K is baaaaaaaad.
You Can Then Just Get One Model Pro Then
I Just Use Continuation Cycles to Bypass 4k Limit
This is their API; I don't think it correlates directly with the app
Why they don't be clear about this before purchase? like they didn't give me any information about this before buying PPLX. What a Scam.....
64K Compare to 4K is nothing.
AI is here to make us struggle less, not .....
Tbh Most Companies are Not Absolutely Transparent
I still have Gemini Pro thank god.
I wonder when these guys will release something https://ssi.inc/
ok, where does it say 4K? Can I have a link for that?
Valued at 32 billion without anything to show š
what are you guys talking about
Its becasue Ilya is involved I bet
Idk About The Documents But I Got That Response Multiple Times Again and Again in Perplexity
That guy is super smart
You Need to Give A Proper Prompt
And Then Ask it to Reverify Review it Until is Certain
i'm high rn i cant xd
this is the best i can
Ai Hallucinates Often
ok
I have seen alot of Hallucinations by Now
So I Established A Verification Protocol
Perhaps You Can Get A Refund
If You are Not Enjoying Your Experience
If You Ask Me After Good Prompting I Have Got Great Results
I Think 20$ for 7 Models is Still Unparallel
For Me is Free XD
I Also Use DeepSeek 8th Model in My Arsenal
It is not 4k, I was able to make it ouput 11k+ tokens https://www.perplexity.ai/search/can-you-give-me-this-code-agai-qjjg6tqDR8mus9qxLzcAEw
Not In All Benchmarks
Both Models Win in Different Benchmarks
And Current Grok has Disadvantages
For Proper Competition We Have to Wait for Grok 4.4
Oh, is there a rumor about Grok 4.4? Or has it already been officially announced?
I Think Elon Musk has Announced it
i asked 3 models they all said 16K !
This is weird... That's why they should be more transparent.
Error
Elon Musk Hasn't Announced it
It's not out yet though, he just said they are working on it
4.20
that should be good model š
Dear Perplexity team, you let us use these fancy models with search restriction. We want research for custom models
Data Verified
No Error
Custome Models?
Yes
Wdym
Research is powered by sonar
Can you share?
Do All Research With Chosen Models
GPT 5 is generating code like this
Then Use Research with All Collected Data
There are search restriction imposed on custom models in "Pro mode"
