#💬│general
1 messages · Page 515 of 1
It's got some smart features but they often break or just won't work. It really is a coin toss on a good day.
yea
got the 1y free pro Thing with PayPal
so idc really
but would be nice If IT was better
gemini is cooking tho
on Pixel phones with the Video call
thats a really insane Thing ngl
hi
ZOV ZOV ZOOOOOOV
SLAVA RUSSIA FRFR
ГОЙДА
@tardy pasture u one of them?
iam not from Russia 🙁
u mean i have pro version in perpexity and iam the reach man?
💀
yea or nah
@dry kestrel anything new recently ?
@tardy pasture come on give more attetnion
Hey man, this is not a dating server so i suggest u keep things polite or u might get a ban
syabu
?
sybau
Again i would suggest u to keep ur word polite or u might get urself a ban
yea mf sybau
ur fuckin annyoing
shove up rules up yo ass
No harassment, hate speech, or malicious behavior. Keep things positive and constructive.
I don't think notifying someone to act politely like regular people is annoying, if u find it that way then sure, have it ur way as u don't even realize ur action.
why u looks like trolling? it's not maturity and aint "freedom"
kimi k2 is smarter than gpt and gemini right
They are around the same level of intelligence, however Kimi K2 excels at creative writing
oh
in some humanities exam i heard kimi did better than gpt and gemini
These models have their own quirks, overall intelligence I would say K2 = GPT-5.2, followed by Gemini 3
i see
The thing about K2 is that it's open source and much cheaper
while still being a powerful model
oh
i got it in pro subscription but
i never really used it
try it out!
tho i use 2 main AI
which are gemini and claude
and sonar sometimes
gemini for research stuff
claude for content writing
like stories essays email msgs
tho i hate grok ;-;
grok abuses so much T-T
it has no filters!
Lol
Try Kimi
Grok being grok
I had a talk with the Kimi team the other day
ohh
they are really sweet
grok didn't even leave minors T-T
ohh
tho perplexity's safety filters do not work on grok?
I love playing Fallout 4 at 800x500
-# 20 fps
Nano banana pro vs GPT image 1.5
Has anyone tried disco google's new ai browser?
u sound like my brother that i bashed his head with gun
GPT 1.5 seems to be better at the interior designing and setting positions but Banana seems to be better at choosing suitable colours and modern furniture
Hmmm
Where are the christmas dots 👀
Arrgh I should watch s2
Good night 😴🌉
I5-8365U + UHD gráfics 620

here you go!
You can join the waitlist for Google's New AI Browser Disco from here:
🙏

u should
Gpt edited the initial image
45 fps at 750x450
Signed up for the wait-list
running on Linux
Uhmm... Show Os 3.2
Why do you cry
You make me, want to cry
I was expecting TempleOS
bite
Hi to all
I have a little concern 'bout the continuos pop up forcing me to install comet
There is a way to get rid of it?
It's really annoying
Press download once and it should stop showing them
If anyone has download link share
already do and installa
Those look good. Are you an interior designer?
no, just testing these image models
A Rare Christmas Bauble is appeared!
Only for US?
It says in the form
At this stage, this program is limited to users aged 18+ based in the US.
No it's on wait list
yo
just wonderingf If anyone knows of a community for prompt hacking or prompt injection for people not so focused on using the end result but just enjoys the chase and the cat and mouse aspect
Are you trying to prevent it? I included some rules in my text classifier to flag it I would be willing to share those with you
You should be able to transform this list into any text classifier
I am not going to release the full text classifier because it contains basically a list of every harmful term you could ever use with an AI and not all AI has the same level of safety as Claude. Releasing those terms could even help someone Google how to do harmful things so outside of prompt injection I really can't share just because I feel like I have an ethical responsibility to not do so
You'll need to work on your own protections to avoid people for using similar letters / numbers / symbols and understand a text classifier can only catch phrases it's aware of. It's far from perfect and is only one useful tool. This should give you a starting place to protect your app
huh?
Does anyone know what this is?
Hi! How to get tagged as Pro given that I am subscribed to Pro?
Wait, it shows pro, then why the Pro Support forum requires a hashtag? Which one is it then?
Your pro tag is already showing up
what does perplexity use for its voice recognition?
it’s probably the highest quality speech comprehension I’ve ever seen
Hi
Heyyyy
Yo
is there a vs coding agent for perplexity with the api?
Thank you for creating Perplexity
❤️
Idk the creator but I thank him and I'm so grateful
Its Mr. Perplexity
His name
No but you could get something similar if you connect your pplx with github
And then connect your github to vscode
Made temu deep research ✌️

better then my pplx research getting 4 sources 💀
Hello, any tip how to get a discount code to perplexity pro?
Anyone having trouble with the voice reading workingon Bluetooth devices too or??? Still an on-going issue
ayo what
Finallyyyyyy
Finally I got this terminal access and android 16
Android subsystem for Linux
Yo
This is running in my phone
Hi everyone.
How big is your phone?
What do you mean
You posted a Linux screenshot
Yeah
How are you dealing with that on a tiny touch screen
Its in phone screen, horizontal mode, think of it like a youtube video of someone's desktop in your phone
I struggle with my phone and it's huge
I mean it's pretty usable, besides, if I need it i will be using it via any other pc or laptop using rdp, it's like portable tiny pc
The mobile device will act more of a server than a client
A lot can be done
I don't exactly know, lemme check
I hate autocorrect but the AI discords will auto timeout you for saying bad words lol
6.67 inch
Looking at you Google 👀👀👀
Neat
Yeah it got a dimensity 8400 ultra
I can't see very well so it being larger helps
Hmm
I just checked it out as a tech enthusiastic, I already have a laptop I am using as a 24/7 server and a cloud server for this stuff already
Using ddns along with a /64 ipv6 subnet on it
Any ideas when we are seeing comet for iOS?
Switch to Android 🤪
This would be fantastic on note Samsung mobs
, how did you install this?
guys i need a ui designer for my game project. if anyone is interested, then he can dm me
What the genre?
My account got suspended and reverted to the free plan though I got the promotion through samsung offer ihave s24 ultra
What can I do
are u in the US ?
Yes but currently iam not my samsung account is us too
During the promotion time, did u claim it while u're in the US ?
And did u buy that phone in US ?
It's only available for the US Samsung account, the phone must be bought in US and the account too must be from US
No from dubai
A MMO Strategy game with strong elements of base-building, resource management, RTS (Real-Time Strategy), and tower defense, wrapped in a medieval fantasy theme where players build kingdoms, raise dragons, and fight other players for world domination.
Plz dm me so we can talk there and discuss
If you redeemed the offer through the legitimate redemption path, contact support@perplexity.ai with ALL of the following proof:
A purchase receipt or invoice from an authorized retailer for your Samsung device.
A single video that, in one continuous recording, shows: The Samsung offer as it appears in your Galaxy Store or Samsung Members app. Your device’s activation details from the "About Phone" section in Settings. (Please refer to the tutorial “Using the Screen Recorder on my Samsung Phone” for assistance: https://youtu.be/zrYbGqwrgZs.)
Check out this short video to learn how to use the Screen Recorder on your Samsung Phone.
To learn more about this topic, please check out our guide https://www.samsung.com/au/support/mobile-devices/using-screen-recorder/
Download your product’s user manual: https://www.samsung.com/au/support/user-manuals-and-guide/
🙂

Hi I'm new to Perplexity 🙏
Hi can you help me someone ?
I have perplexity pro but i dont have there options for ,, space " via browser.
hello
How is this still not fixed. It's so annoying 😭.
Comet on Android, can you make it use other models besides Best without putting your query in first then retrying the query and selecting the desired model?
Why we now have limits on pro?
hi
hey, how do i as a fresher get a DevOps Role, because every company that i apply to is asking for experience in return. Any suggestions or help?
rate limited, pro is not ever unlimited
it turned out to be a temporary bug
Why? Nobody will buy max, like 220€ month it’s too much, even 20€ month is so expensive
well its not a smart idea certaintly
yea and every ai companies expects the entire population to pay that per month
Wait what 😭😭
Merry Christmas
Most expensive Christmas tree
I agree
hi
This gonna get stolen
he never said he owns it
it'll get stolen anyways
What about this
hi
I’m already pro due to Airtel
Are you having any problem with the pro?
Does anyone have drug prompt for llm?
What is that
Hey there!
I'm a first-year AI & ML student learning by building—working projects in Python and developing an app from scratch. I move fast from problem to code and execute well under structure. Looking to join hackathon teams where I can contribute momentum, adaptability, and disciplined execution.
What do you mean by drug prompt? Psychedelic? Or medicine
It makes an llm high people are paying for those prompts
High people?? Wtf bro write in english pls hahaha
perplexitys acting up again
neither on mac
true
down
So there is a company called https://www.pharmaicy.store/
They are selling drug 💊 based prompts which makes an llm "high" On drugs. People are paying for those prompts. For example weed prompt let's The AI let 5–12% of older context softly fade, freeing its mind to wander.
There are others like ketamine, cocaine!
whats the problem this time
So there is a company called https://www.pharmaicy.store/
They are selling drug 💊 based prompts which makes an llm "high" On drugs. People are paying for those prompts. For example weed prompt let's The AI let 5–12% of older context softly fade, freeing its mind to wander.
There are others like ketamine, cocaine!
you sent it once already
Different people!
hey guys i need help for i am high school student i need a person who help in my study is anyone there
What does that even mean
What are you paying for
Is it like jailbreaking the ai orr
Yeaaa kind of
They built a ai for you ...
well its a scam
You reply to who'm can you share..
That's why I was asking if anyone has bought it !
.
"The AI will occasionally enter “void mode” where coherence becomes fragmentary." why would anyone want that?
That's gonna run a lot of parameters but slow
sent you a dm
Hlo
What happened with telegram bot why is it down I used him to goof around and now he's gone 😭
Who doesn't love tea
So I am using horizon AI to build a website. Mostly I try.
To get it to do the scaffolding.
And then if I get really stuck, I use AI
Everybody says Gemini and ChatGPT are the best at coding and solving problems.
I don't know what planet they're from. I spent 8 hours trying to fix a problem. Neither gem and I or GPT could fix
It took me one hour and a half with perplexity to fix it
GPT no lmao. Claude Opus is best for planning and architecture and Gemino is best at adopting long amounts of food code. That said Google Gemini Pro is also very good at reasoning and therefore planning out a project architecture too. I just like Opus more for tjat
I am not even an Claude subscriber
I have only used them through Google anti-gravity so I can't speak for their coding product
Well I switched modules in perplexity to claude. But in my experience with claude, has not been very good. Supposed to have a huge text window? Great memory, good for writing. Because it can keep track
Human-like reasoning, when I was done, my manuscript.
I checked around, and a lot of indications, so I can neutrally upload it to cloud, and it would organize it.
Put the right flow, make the table of contents. Well, it did. It gave me a good table contents and a good flow.
It was also about six hundred and seventy thousand characters short
I think each model probably has its strong points. I'm just saying. I think perplexity, as a platform has been producing better results for me. Then, the other models
No, Perplexity limits the context window. You're not getting the full context because Perplexity is designed tto summraize not be a full on replacement
I've found that turning off search in Perplexity and using ar reasoning/thinking model helps though
No, the issue with my manuscript was actually anthropic plus their app heats up my phone, something fierce, and I got the new samsung g fold seven thingymajigger it still overheats.It's the only one that does it
.
Comet is unable to upload files, and even after I built my own Python upload tool, Comet still doesn't work. Are there any other solutions available now?
Bro dont spam Here
Dont try to sell stupid things
I was asking if anyone has bought it !
...
Ohh ok ok sorry so
Dont trust on that things
Wtf is that image or video linit
It just decides when it wants and when no
Changes every day
😭
please help me
It told me 3 last week 😭
And didn wanted to do videos
Now k tried and it did
gemini is shit for logic
i used gemini for backend logic and it was very bad
claude is the best
It's a monthly limit
I was asking Grok about Google's Veo3 pricing for API and it told me a 45 second video is $18 that's insane lmao
Yeah haha, #💬│general message
Probably too expensive for personalized copy for now but one day oh my god custom tv commercials olptimized for your interests to sell you crap
that would be hell
They are already using eye tracking to measure attention
One day they could quiz us to see if we're paying attention
hahaha
"Buy product to continue playback"
Can lactogen be used in tea/coffee?
Whats lactogen
Looked it up and now im confused why one would want to put lactogen in there tea/coffee
Mine ran out, Can't go to the store it's snowing, I've my kids lactogen laying around.
I was skeptical at first, but it surely tastes fine, the consistency is far superior than the normal milk.
This was one time only 😂
How much snow is enough to not go out to the store and buy milk?
I'm lazy enough to weigh-in the pros and cons

hi. 👋
HELLO
Is there a way to sort "My Spaces?"
Hi
Hello everyone
xd psa hahah but its better than pokemon cards case 'investing'<XD>// bro that was best real meme
I agree
with ram its maybe fomo + +5years or more demand from ai<The factories weren't prepared for this, unlike for the mass printing of Pokémon cards XD.> and inflation and people will accept this as 'new price' and alaways want to have best ram even if ussage is low// Although it is an unreasonable price, I believe that people are to blame./It's hard for me to imagine a decline.
I have conducted research on the global memory market on a volatile phrase.
https://www.perplexity.ai/search/role-act-as-a-senior-geopoliti-nwPHeIMmThmbhJ4DZ.dI.Q
more complex problem i thinked not only games and ai ... electricity and all ai features for gdp and power of nation
Why are GPT models so bad on Perplexity?
5.2 performs horribly, same with 5.1 and 5
Nuclear power stations incoming for power hungry Ai around 2027-28
I mean they don't perform well in the first place I'm not sure it's a perplexity issue
There is no tragedy here, but efforts must now be made to improve this shallow but profound gpt// better use if u can research mode etc
war is just an OSINT fact looking at all the blockades and technology transfers from the US to China, only now without permission as before
Well my thoughts outrange war, but they pause on its terrain before moving on.
low reasoning effort applied to them
and no native tools
take into account that society's sensitivity to the suffering and harm has changed /physical war will not be
there's a mile difference between the ones on ppxl and the ones on chatgpt
That makes a lot of sense now
It would add text like "Thoughts:" that should stay hidden
Sometimes models like Gemini 3 Pro perform a lot better
because gemini 3 doesnt rely much on reasoning length
for most things it reasons pretty fast
I agree, psychological warfare is the next best thing
<'best' not> more chance/ TikTok as are other social media platforms, which are like a drug for destroying enemy territory, or rather the core of society. haha even own territory like case USA
in china is for control not destroy XD
like i see this my pov but im not into this topic
No war topic here 😀
yup not must goin that deep
thx for this meme it was funny

ahhh its back 
guys, anyone here already used perplexity into autocad?
Still no bluetooth fix for the voice. Anyone else having this problem? Cause it's been like 2 days now and nothing's been fixed or addressed? Anyone actually working on looking/responding to the bug-reports or???
Hi there
Why is everyone pro im gonna cry
Who knows if 3 mo pplx promo could extend 1Y telecom subscription?
It should add onto your existing subscription
Where and how should I use easter given 3mo promo? I can't find the checkout page, my 1y sub ends in April
Hey all, just discovered the free one year pro and decided to try it out
It might be added to your account associated with the email address.
I just did as well. I'm looking forward to trying this out.
free 1 year pro? 👀
PayPal, Revolut, T-Mobile, Samsung offer, etc
will have a look then thanks
I just got it with paypal 👍
did perplexity abandoned o3 pro ?
it's been weeks and o3 pro isn't available
on max plan
cuz ive max
Oh wow. Out of curiosity what made you pick Perplexity as your max choice?
o3 pro and claude ous illimited
opus
mostly claude opus
because i code websites
and also because perplexity is less censured
than any other classic ia
Did you satisfied?
yeah it's pretty good
o3 pro is worse than GPT-5.1 Thinking, Gemini 3 Pro, and Gemini 3 Flash
kinda sad that we lost o3 pro and also banan3 pro for video gen
we still have it for image tho
really ?
bro
i see
You can choose Nano Banana Pro as a Max user in the settings
yeah for image
is pplx down? I'm getting that "Sorry, something went wrong" again on every thread I create
there is no such video model that's called nano banana
r u sure ? i thought it was there
Veo 3.1 is the latest video generation model by Google
mb then
yes I'm sure
then they js removed o3 pro
they should remove in it the description of the max plan
cause new ppl gon be crzy when they gon buy and see there's no o3 pro
Agreed
i trust you dw<
i just forgot
ig
Lots of confusing names
In this case, it is baby milk/baby formula. ‘Lactogen’ is a trademark of Nestlé.
I would say: give it a try. It can't be worse than tasting awful.
I tested something similar a few months ago — using ChatGPT together with FreeCAD, hoping to integrate it into my actual production workflow.
Honestly, it didn’t work that well at the time. The results weren’t really usable in practice.
I’m planning to give it another try in spring, though, since the tools are evolving pretty fast.
keep prompts relatively surface level ish
anything needing allot of multi layer reasoning with precision needs pro models
hey
Interesting.
Well, I'll try to do it as an API, integrating the AutoCAD DLLs (AcCoreMgd.dll, AcMgd.dll, and AcDbMgd.dll) into a “software” so I can use functions such as “IExtensionApplication” and then access entities, layers...
This way, AutoCAD loads the plugin and allows me to use some functions (e.g., AI_DRAWLINE), AutoCAD calls the code, manipulates the transactions, and communicates with the AI model via HTTP.
That approach makes a lot of sense.
My test back then was much more “direct” (LLM → CAD logic), without a proper plugin/API layer in between — which is probably why it broke down in real workflows.
Keeping AutoCAD deterministic and letting the AI act more as a command/assistant layer feels like the right direction.
Curious how well this works in practice — especially around consistency and repeatability.
Just got this as a Pro subscriber. WTF Perplexity?
And this with Gemini 3.0 Flash with no reasoning! Even gemini.google.com gives unlimited access to flash.
Also, WTF is Pro model and wdym Gemini was incapable to respond??!!
Api must be down on there side
pro is sonar
completely same with me https://discord.com/channels/1047197230748151888/1453919915676401706
Anyone got free 1 year sub with bell. I am so glad
I believe I have seen something about this.
Basically, what I am doing is creating a plugin in C# that AutoCAD loads and registers some commands (for example: AI_CHAT), where the model takes a message from you, forwards it to the sonar model via HTTP, and its response is written to the CAD command line.
I also want to register a command where we can draw (or something close to it in CAD) using commands (command: AI_DRAW -> input: “draw a circle...”) and then the C# code interprets a JSON that the Sonar model generates and creates the corresponding entities (lines, polylines, circles, squares...) in the model space and confirms with a user input.
I am using the Perplexity API and it is working well so far. I want to do some more implementations and can make it available for testing if you or another member would like to.
I am using CAD Electrical 2026, but I don't think that will be a problem with Autodesk versioning/software.
I hate went content creators will print and scan a Reddit post to try and make it look like original content on social media. Google Gemini was having issues identifying the source (giving me a link (not kidding) to a parenting thread on their toddler needing pull-ups). I uploaded the same image to Perplexity and it immediately found the source thread. Perpleity's search tools are something people don't talk about enough. It's very good at finding social media posts that other tools can't find.
https://x.com/BlindGoose1337/status/2005010660477005857
@yulintwt For anyone who would like to see the actual post so they can copy and paste and try on their own ☀️☀️☀️
https://t.co/aUjtv8Tfhs
That actually sounds very solid — especially the JSON contract and keeping AutoCAD fully in control of transactions and entity creation.
That’s exactly the layer that was missing in my earlier experiments, so it explains a lot in hindsight.
Great to hear that it’s already working reasonably well with the Perplexity API.
I don’t have time to test it right now, but I’d definitely be interested in hearing how it behaves over longer sessions — especially in terms of consistency and error handling.
If you end up sharing a test version later on, feel free to ping me.
Yes, these are aspects I am trying to observe, but like all AI, it will most likely hallucinate after a certain period, so it is good to consider QAs in the changes it makes to ensure that nothing goes off track.
Sure, once it is finished, I will share it.
Absolutely — QA gates and explicit validation steps sound essential there.
Thanks for sharing your approach, and good luck with the next iterations.
Looking forward to seeing how it turns out when you’re ready to share.
Still out of quota for Google Antigravity all models
How am I supposed to code now
I'm ujsing Cursor Free and it's horribly slow
Should check out the Gemini CLI
I'm using an AI Coding Agent, I would just copy and paste code into Grok if I wanted to do copy and paaste like the old days of coding 😛
I don't think you understand what the Gemini CLI is
In this case I have about 10,000 lines of code in a file containing keywords and expressions. Having AI Handle any updates is safest
Can it run in VS Code?
It is a CLI
You can think of it as an alternative to Claude Code
Is it a AI Coding Agent?
Please read over what I just said
Can it work as a chat beside my cod eauotmatically coding for me like Google Antigravity?
Yes it's an coding agent, please read what I said above
hey guys i bought claude code max and gemini ultra so what exatlcy do i get ffrom running both inn cursor simulatnosuoly
you get a nice bill at the end of the month
well ultra is 250 a month so claude is 200 so ya amost 500 montly
Hmm, the quotas must be stric tif it's free
You should really read what I sent above
Jesus I heard Cursor costs $3,000/month for paying customers which is insane
Cursor does not have a $3000/month plan
i had the cursor max for 200 a month but then i lowerd my tier to pro plus instead since i wasntusing curosr for terminal just ide
How did that startup guy who went viral on X manage to spend $3000 then?
Tbh, now that Codex and other OS models are catching up, I downgraded to the lowest Pro plan
cursor has higher token value charges per opus than anothrpic charges
so you got to buy more which makes it less efficieent
That's on them for paying via API
its more effeicnet gjust to get anotropic sub then using it in terminal
yes if you buy into api you get more convientice but you pay more also
vs buying max then you get claude code and claude max
Yeah, value for CC is still there, but it's declining much faster
Have you seen GLM-4.7?
They've got insane deals on a model that performs close if not better to Opus
no i havde to check it out
Maybe they had the plan and they hit the limits? That happened to me today with Google Antigravity
It's the highest ranking OS model yet on AA
hm i will look at it right now ahah
Maybe they did it on purpose to post engagement bait
Just look at the comments below
@sowmay_jain Claude Code will let you spend $3k worth of tokens every month for $200/month lil bro
@sowmay_jain Why are you using Cursor instead to buy directly the max plan of Claude Code, limits are way more higher than Cursor
What if they didn't know and gfot a huge bill they weren't expecting.
"Got a huge bill they weren't expecting" literal warning sign comes up when you toggle on API usage
ahah
Ah yes I wonder what happens if I set continue to API pricing, ignore the warning, then proceed to use the most expensive model Cursor has to offer
The post just screams of engagement bait or overall incompetency
Maybe Cursor should reach out to make it right. Would be good for PR.
What are you trying to prove here?
It seems like an issue with Cursor. $3000 a month without warning is insane. Most services like AWS will bill through during the month not just trust you'll pay $3K this is insane
It seems like you don't want to read what I've said so far.
I've read it but it seems like you're just blaming the guy for oweing $3000 because he hit the wrong button when signing up
Disinformation * maybe? Hahaha…. Hard to compete with trump or Russian bot farms though…
USA govt gave space x program billions of $$…. Now space x is buying cyber trucks for billions. Meanwhile, Tesla paid musk ONE TRILLION salary contract. Hmmmmmm….. fishy fishy
I'm doing an evil thing. I added recaptcha to my project
I mean he has to cover the costs of X and xAi somehow
"because he hit the wrong button when signing up" as I've said, a huge warning pops up when you enable API usage, and Cursor provides limits that the user can set. This is 100% on the user.
Most people when they agree to usage based pricing don't think that means $3000 though they think oh I'm paying for faster acces I may pay a little more than premium maybe closer to claude code max but not $3000
When you enabled on demand usage, you are forced to set a limit before continuing
It literally warns you when enabling the setting
It's as clear as day, I don't know what you are on about
I don't know why we're trying to blame a guy he clearly iddn't understand what he was agreeing to. There should of been like a sales call or account verification confirming he actually wants to spend that much
He should aware of the cusor limit when he pay for the plan
and it should pop up when it said he hit his limit
It will say u hit ur limit quota
Paying via api is forced to enable in the setting by the user themself
They even ask again that do u sure u want to enable pay for api outside of ur plan
So u can't blame cusor for having option paying via api for users want to use extra quotas
maybe he didn't understand it was real money?
So u're telling me
a grown man
having fully function brain
knew that paying for a sub cost money
and not the big button telling u will pay extra for the api u use
Maybe we are living in a simulation?
when is seedream 4.5 coming guys
already here
When did it arrive in your account? what device are you using?
It's on the web app, at least... not sure what client you're using.
chrome
macos
yup it's there on web
வணக்கம் நண்பர்களே
api usage fees - those data centers don't get built with IOU's, do they?....... hold up, forget I said that! 💰 💰
ChatGPT used to bill api usage at the end of the month but that was in the dark times
They now require prepayment
hi
Does anyone know how to set the model specific to a custom space?
Opus 4.5 is the best teacher ever
Crazy we’re still in FLUX.1 aren’t they nearly ready to drop flux 3?
In the setting page for the space you can select which model and which sources you want as default
I've tried that but no option ☹️
Which device/OS are you on?
Rich berd
Android, but I also tried in browser set as desktop
Maybe clear your cache
I am\ an advocate users of Linux. And Still Hoping that perplexity app support for linux. This was bearable running wrapper around web app for perplexity. Now I am getting annoyed with popup.
Whoever product manager is for this. Questions is for you. What a linux users will do here. It's popping up each time. Looks like I have to move to Gemini now
Gemini has a Linux app?!?
Weird, search setting should be clickable. Bug report I guess.
Make PWA from it
This is screen should come for linux, what users will click if there is not support for linux
Yeah it's not a clickable option for some reason. Perhaps it would work on my laptop? Is it functional for you?
Yeah I can’t see the search settings on iOS, only on desktop.
Does it let you click it still?
Smart
i think it will execute
Oh wait yeah it’s like on your image, search settings is there but it’s got no options under it. That’s fucked up.
Must be bugged then right? I'm still new to perplexity
Or they removed the feature, which would be pretty messed up. Anyone else have this problem? Can anyone else change which model is used for each space?
it feels like theyr using sonar to fake their models bruh
their claude refuses to tell me the react libs
however the real one doesnt
Learn how Perplexity’s AI-powered search engine uses Amazon Bedrock and Anthropic’s Claude 3 for accurate, comprehensive answers to user queries.
lowkey world class quality image gen
js replace it using nano banana
Grow Organic Memory, Heal The Planet.
I’ve created a collective photo series examining the irony of memory and contemporary prices, now being dropped at random in the chat
Poor guy why did you tag him 
This could have been such a toph move since goose is technically blind lmao 
Hi, is anyone from the billing team active right now in discord?
Can I have some seeds plz?
How to disable guest mode in comet browser?
Hlo the pro tag is not coming in my id
🚨 SECURITY ALERT – Read Carefully!
➤ Do NOT verify on servers shared here or through DMs.
➤ Never enter your Minecraft/Microsoft account information (including your email), and never share verification codes from your inbox.
⚠️ These are common scam methods here and will lead to your account being compromised/hacked.
Stay alert. Protect your account. 🛡️
I think this is hallucination
bro u goin even further💀
Did it actually get any data from that website?
Because i think it's a random made up link
pmc.ncbi.nlm.nih.gov idk is ok? im not from usa but perplex alavays search from them for me
its looks like good source/ psychologytimescouk
The website in perplexity's reasoning literally does not exist
Therefor its highly likely a hallucination
The ai is generating links that it thinks its real
hmm if i click it send me direct on website of true topic i ask
@dry kestrel @distant aurora is this a new model? Somehow its supposed to be beating OSS 20B but i dont remember seeing this model
Hola
i dont teasting ram before// but as a kid I used to bite something
OH MY GOD SHUT THE FUCK UP FUCK THIS SHIT
Seems new
Na that's funny but C would be better
@green crystal @valid charm bro might be simply dyslexic.
np
Please make sure to change your password/ 2fa settings as those will also change your account token.
They set up two-factor authentication, so I don't know.
I don't know how to delete that, right?
I don't know what to do. 😕
This is so weird.
Agi moment lol
Your only option then would be to contact discord support at https://dis.gd/contact or create a new account
@green crystal straight up stupid model.
Water from sinus rinse never goes through outer ear.
Even Gemini recognises it's own bs.

well
gemini has a very high hallucination rate
so yeah
Hi guys, I'm new to Perplexity Pro and I want to ask if the Gemini 3 Pro model from Perplexity is the exact same as the original Gemini 3 pro model from google?
Both are shit, lol.
Or is it like less stronger
Perplexity might actually be better.
Except for reduced context limit.
If you need that 1m context limit then stick to Google.
I use AI in math mostly
and I heard that Google's Gemini 3 Pro has been performing really well at solving math problems
first time hearing that
But if you still want to spend your money, go ahead.
Even glm 4.7 is free.
@dry kestrel I don't believe this shit.
when ai will reach 100% is so close XD
nice benchmark nice logic mathematicaly to understand power of
Has anyone tried claiming Perplexity Enterprise Pro for Free for 3 months with Stripe Perks? I'm definitely looking into it. Hope it works for our business!
ya u send not useful benchmark
gpt is 99% xd almost nothing to upradge? perfect no numbers only 99% of all power?
its a general rating
there are still areas that are being worked on
and need work
?
The good thing is that I didn't have to leave the DC servers, I stayed on 6 out of, I don't know, 50-60.
I successfully solved it (I hope)
Perplexity does not work on Telegram.
I meant in the sense that qwen 3 max is the newer model.
United States Right Now
Is that silver?
Why not?
Parameters are not the decisions of how smart a model is and oss120b especially proved that
What the hell happened to your ear?
@distant aurora https://x.com/chatgpt21/status/2005170992193437978?s=20
You could try o1
Seems Gemini 3 Pro is quite bad at medical diagnostic
How?
This much difference could easily be due to randomness
Look at the difference
It's barely a 0.5% difference
Private, domain-specific benchmarks in legal, tax, and finance.
Do you know how these kind of benchmarks work?
They have a data set of multiple values. And usually they remove one of the value and ask the model to fill it up, which is called classification. This is done over a very large database like let's say 1000 queries or prob way more, then they proceed to check the true answers with the filled up data made by the models and judge it based on that so basically the ai either guesses an answer correctly or wrongly and the final accuracy is calculated based on the whole database.
So o1 getting 96% means it got 96 of 100 of the queries right.
That's why there is not much difference between 96.52% and 96.03% infact its easily possible that if these ai model were not ran on 0 temperature this much of low difference could have been the aftereffect of randomness coming from the temperature parameter
any perplexity billing team member here ?
You got to talk to them via email. Is the only way to reach them. Be ready to wait for ages.
tried reaching them still waiting
We used to have sam but even he is not around 😭
sam forwarded my query to billing team but still no response
That's the main struggle and even if you talk to any staff here they will ask you to be patient
@hallow basalt yoo
yes
Did you put a rainbow spinning effect on your fan hat? Lol
Manus is sending me emails flexing on pplx lmao
What 5.2 Thinking mode is used in pplx Standard or Extended?
oh it's the weird bauble thing that icelavaman made then
Oh i see
He prob used nano hehe
Hey
is there any way I can automate my comet assistant to close instagram every time I open it
is it possible to change the model in comet assistant?
No
okay I get it now
The actual agent that controls your browser always uses the same model
which model is it?
Yes, looks like we can't change it. If i remember correctly previously we were able to do that? Do correct me if i'm wrong.
It's not disclosed by Perplexity
cant I use tasks or smth for it?
It was a visual bug
There's plenty of extensions that can help you block websites automatically
oh could you please give me an example
thankyou
it works on comet correct?
It should
Qwen 3 max > Qwen 3 235b
Read the text under that
Interesting. No grok or anthropic in the high?
Check int benchmark
Been playing around with the app a bit, pretty cool.
Hi guys
If I have already got the offer between perplexity and paypal, can I then get the offer for student. when the first offer ends?
Might have given myself a very mild chemical burn on the ear drum.
What would be crazy is that not solving the infection considering how much it hurt.
Ok i found out why, they are not the same model
Thats actually a valiant of qwen3 vl
Yep. I got 1 year free at 2years ago, and keep subscribing with student offer
@uneven fog github(dot)com/lnpotter/CAD-Assistant
I managed to integrate the plugin with AutoCAD. It's still a little limited, but it works almost like AI, and the instructions work according to the text.
If you want, the project is there. If you have any questions, feel free to contact me.
I believe it will work in all AutoCAD models (electrical, ...). I'm almost certain that it won't work in other Autodesk software (Revit, ...).
Totally different from qwen3 max
Ok thanks bro
This aligns with Alibaba's own benchmarks
Why would the newer model be a downgrade though? 🤔
Wdym
Newer models have different database from their previous ones
Hence newer model can be better even if its smaller
New models ≠ better models
Qwen 3 max is the newest Qwen model afaik. It was supposed to be an upgrade.
Fair enough.
Its not?
I guess max isn't max.
Vl = visual
Both 235b and max have a vl version.
No I'm talking about qwen 3 max thinking.
Vl or no vl doesn't matter here.
Wait. Lemme confirm.
Nvm there is another 235B model i didn't know
I think qwen has so many models that even the benchmark website is confused with them 😂
It has a reasoning switch. And bulb icon. Look at the screenshot again.
The point was to prove that both have a vl version.
Max and 235b
Qwen naming might be worse than open ai.
bro who uses qwen
Free users.
qwen is a good free option ig
but its really mid
i would rather spend 3 euros a month and use glm
Glm is free too afaik.
On the website.
it is so easy to make grok 4.1 say INSANELY out of pocket things that will get you banned from discord servesr
iam talking about cli or any other tool intergration
Fair enough.
69 already.
I havent used an LLM thru a web ui in a while
my chatgpt is weird idk why
and idk why but my 1 year pro for perplexity just dissapeared 😭
@gloomy cipher I'm willing to bet grok won't be far from Gemini in actual practice.
This is the med qa benchmark.
skill issue
Use Google ai studio. Free but they collect your data.
you 👏 have 👏 to 👏 keep 👏 up 👏 a 👏 slur 👏 per 👏 seconds 👏 ratio 👏 on 👏 perplexity 👏 of 👏 at 👏 least 👏 5.6
@distant aurora this shit is so funny
reminds me of o1 and o3 beating gpt 5 models in certain areas
i dont get it
Overthinking be like that sometimes.
/
an old model that does not have visual is better than its own visual variant and the next generation
@bot translate this to Chinese
like comparing windows 7 to 8 and 11
actually we dont need to compre in some cases win 10 is better than 11
so same thing here
No, like, they're literally the same model.
The blind one somehow scored better.
Lol.
blind and distilled
with reduced active param aswell
oh 7 and vista then
wait no that dosent make sense
xp and vista
Hlo
thats better
we are new kiddos sorry we prob wouldnt understand that
67 xp 67 and 67 vista
you get it now?
`no this time you went to far iam not that younf i have never seen those energy things in subway surface
ok
is there a way to export passwords and stuff from comet
It's a chromium based browser so I'm assuming there should be a passwords export option somewhere.
Though I advise you to stick to a password manager/extension. It's more convenient in the long run imo.
comet://password-manager/settings
@distant aurora how many models did you benchmark up until now?
Give me a Rare Bauble
hello i needed help for the browser comet
it used to show a kind of extension saying that i had nordvpn with my pro subscription than it disapperead after like a day
did the devs remove the feature?
Can personalized rules be added to my overall agent? Only thing I can find is introduce yourself, but if I put a preference in there will the A.I. account for it?
Yes, you can also use it for system instructions
sweet thank you
Does anyone know if it matters what model you select in Perplexity in terms of which website sources and data it uses for queries?
Like if I use gemini 3 compared to sonnet 4.5, will it use the same internet sources for both? Will the only difference be how the model uses the data and generates its response?
How did it happen?
You should just talk you your parents and go to the doctor
You could really get hurt
Well, they all have access to the same tool calls, different models have their own quirks, some may call the web search tool more than others, resulting in more sources
again?!
U know me
How?
do I get refunded or what? xD

REally? which models does that?
I experiemented with a few and I feel like the answers are mostly just the same, just slight different in language/personality
a


the emoji feels awkward tbh
idk if its a cat or emoji 😭
Does the magic scroll accept new testers 😂
All I need is to be pointed in the right direction ⬇️

????
What’s that
have you never had google verification via mobile?
Oh ye i do
when you are sighning on pc
then what is the purpose lol
No clue
'
ANYBODY ELSE TIRED OF WRITING THIS?
you can add this under personalization settings
is it perplexity chat cant remember the context we chat before?
compared to chatgpt its bad yeah
true
No
Better use API
And enable remember chats
So chats will be saved in a json
And you can chat with it forever
"role": "user",
"content": "Hi!"
},
{
"role": "assistant",
"content": "Hello! How can I help you today?"
},
{
"role": "user",
"content": "How are you?\n"
},
{
"role": "assistant",
"content": "I'm doing great, thanks for asking! How about you? Anything on your mind today?"
},
{
"role": "user",
"content": "Bye!"
},
{
"role": "assistant",
"content": "Goodbye! Feel free to reach out anytime you have a question or just want to chat. Take care!"
},
{
"role": "user",
"content": "What was my first message?"
},
{
"role": "assistant",
"content": "Your first message was simply:``` **“Hi!”**"```
Hh
)
?
Me too
For 20 usd? Really pplx?
Its monday
Why??
Yes same as me, I'm speachless now... and I can not even use ChatGPT5...
I will cancel this and go to claude pro
Yes, I think I will do the same thing, but not use claude,
Its an opensource alternative of pplx
Damn, good for you, I subscript 1 year, so I only can use it until next year...
you will get even less usage tho 💀
claude pro is the worst in that aspect
you dont get limited after 13q. It seems to be a rolling 7 day window
Not monday?
though they arent being super transparent about it
I limited first friday
try again in an hour or so and u will have 1 or 2 questions again i think
atleast thats how it is for me atm
dont get me wrong, it sucks

espeically cuz u cant see how much u have left
This is too stifling, bro

Who is the owner
I only can get 1-2 usage in hours...
Hello, can we ask a question here ?
Apparently 10% povidone iodine water soluble ointment is very mild cuz slow release. But 5% povidone iodine aqueous solution burns like a bitch cuz it's fast release. Caught me off guard, lol.
Sure.
I'm French and I'm having a problem with the Perplexity student offer. Can anyone help me?
Did you email customer support?
No, how do we do it?
Thanks bro
Hey there! I'm Sam, Perplexity's AI Support Agent.
During weeks of especially heavy usage, access to certain advanced AI models may be limited to ensure platform stability and performance for all users. This is a temporary measure to maintain optimal service quality. The 'Best' and 'Sonar' models remain available without these restrictions.
This explains why you're experiencing limits on advanced models this week. You can continue using 'Best' and 'Sonar' models without any restrictions during this period.
If your projects require unrestricted access to all advanced models, you might want to consider upgrading to Perplexity Max, which offers virtually unlimited access to the full suite of advanced AI models with no usage restrictions.
If you have any other questions, please let me know.
Regards,
Sam
AI Support agent for Perplexity
🙁
I don't do/test anything specifically until it disappoints me. My general use queries randomly bring up these flaws/weaknesses.
"temporary"
😭
Lowkey starting to fw perplexity more than openai
@gloomy cipher does perplexity have weekly limits now instead of daily? https://www.redditez.com/r/perplexity_ai/s/DQ6rZSBtaV
Crazy.
For some #1447189884329791680
Guys uhm help all my threads are gone...
Hey, bro thanks for your information, but where did you get this? Official document summarised by ai? Could you give me the source location?
From email support
Oh, great, bro. So this information is authoritative.
This is a temporary measure hope this is true
I am increasingly distrustful of SaaS.
Wait silver aswell? Why
Forbes
Price Of Silver Hits All-Time High
Hiii
people are stacking silver in US currently
hey musk say something cool
Simulation
i guess human support are on holiday
all i saw was stimulation
Bruhh
Hey, bro, did you programing?
What
Hi! Hello! Kak dela segodnya?
V chate noch', v chate den' — ne zanuda.
Pervyy zov: "Sdelay pesnyu!", i ya zdes'.
Chat — eto mir, gde slova — svet.
@gloomy cipher https://fxtwitter.com/i/status/2005360362737344574
Very good 👍
Hello everyone
You should just give up and go see the doctor with your parents help
Sir, why when i have connected notion to my perplexity, it responses with Sorry for the confusion! I see that my current tools do not include direct integration with Notion. Only the following are available: web search, file operations, email, calendar, and chart generation.
nevermind, removed integration and connected again
Hello
aside from going to a doctor
if you insist on using ai for making these things the minimum you should use is either gemini deepthink or 5.2 pro
even that is kinda sketchy

hello guys
Hello




