#💬│general
1 messages · Page 272 of 1
ok. so what u do for prompts?
I... just make them myself
I just explain to perplexity what is it that i need for that space
that's all
np, have a good day
So nice of you ! Do you still happen to have one left by any chance please ? Thanks for the sharing anyway !
Sharing it, Check your DM
Thank you so much !
Hii
But what does it doo ?
Remember to invite more peoples
Do not gate keep it
Comet is a new web browser built for today’s internet—one 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...
Helps me cheat in stuffs
Pretty good
specially Quizes
That’s good but uhm what is agentic ai
Thanks another thing will perplexity assistant feature be on iOS
I mean we have Siri but does perplexity have more advantages
That is Crazy stuff
But mind blowing
It does have it rn
But the phone assistant from preplexity app
Atleast the one on Android is not that great
So it would work better if it was with iOS you say? I mean imagine apple and perplexity together
Perplexity Assistant is already in the app, and from there it can do some stuff. Sadly you can’t replace Siri with it due to limitations of iOS
That will be a major upgrade for Siri
Oh I see
Huh
U misunderstood it seems
Yeah sorry
But do you think if perplexity and Siri come together
It will be a phenomenal upgrade
No idea
Never used apple products
Yes ! Just did that ! Will be sharing back to the community as soon as I have other available !

Yeah Siri ain’t that great maybe on upgraded iPhones is of use but like perplexity could be dang nice with it
Can anybody guide me how to use perplexity api locally on my app ?
Need to pay more to use APi
5$ per api ig
The free credits of 5$ ? I can use that to make call?
You can
Getting 401
I see
What model are you calling? Don't think your code is right
It is right have tried with almost all models
When i am trying to make a basic curl to just fetch some information still it is not working
curl https://api.perplexity.ai/chat/completions
-H "Authorization: Bearer YOUR_API_KEY"
-H "Content-Type: application/json"
-d '{
"model": "sonar-pro",
"messages": [
{
"role": "user",
"content": "What are the major AI developments and announcements from today across the tech industry?"
}
]
}' | jq
Is it?
Wait i will check
All models like sonar work
What's an api
in simple terms it's hosted somewhere and used somewhere
another way API (Application Programming Interface) is a way for two software systems to talk to each other. It's like a menu that tells you what you can ask a program to do. It's usually hosted on a server and used by other apps or systems to get data or perform tasks.
all good bud?
I had to get my second molar removed 💔
dang, must be a pain
@tame current https://www.instagramez.com/p/DMtzZ0ZIL1r

lf web dev interns
Chatgpt can show weather widgets?
Probably idk. This was just in chat though.
Damn. That's too low
Hello friends, If anybody here has invite link for Comet, please share it with me, thank you so much🙏
It's rainy reason.
Try your luck on twitter.
I will try there, but I think somebody has the invite link here as well, thanks for advise
@languid star https://www.instagramez.com/reel/DMuht4jSy8A
🤪
How to get api?
PPLX api? if you're pro subscriber, you get monthly allocation of $5 USD.
https://www.perplexity.ai/account/details you can get you api key from your settings
@dry kestrel now you can use that deepseek api with raycast!!
Does it expire after a month?
its $5 every month even when you don't use it. when you go over then you have to pay more
and you can auto reload
i will experiment with my app when an agent needs to get more information rather than use from training
useful for agentic coding when solving for bugs. the agent can tool call (search in this case) so it doesn't go in a loop in an attempting to fix a bug
perhaps PPLX has an MCP server that I can just put my API key instead
bruh... i already have google workspace with 2tb storage... this is all i get when i upgrade my plan... dufuq?
im logged into google... they already know a lot about me... why are they not serving dynamic recommendations based on my user state? 
no thanks,... i get way more done with PPLX
I am pro user
Got free subscription from my router provider
Network
airtel ?
Yeah
Lol
They will never give
🥺 why ?
Ambani is very poor
😯 ooo isee
From heart
he only spent in wedding
@languid star https://www.instagramez.com/reel/DMmWo2COm7M
💬 2.1K 🔁 0 💜 101.9K 👀 470.8K
Ehhhh-Man
Pc: @patandar_z
#memes #humor #viral #funny #satire #edits
If anybody here has invite link for Comet, please share it with me, thank you so much
maybe he still .............
I like cursor ai agent
#☄️│comet-general pls go there chances are high that u will get one
yeah they are very good
twitter?
but only in frontend
what ?
Hmm
hmm
where?
#☄️│comet-general here
⠀⠀ ⢠⣄⡀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⢠⣿⣿⣿⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⣠⣾⣿⣿⣿⡆⠀⠀⠀⠀
⠀⠀⠀⠀⢠⣞⠧⠡⠄⠨⢉⠿⣆⠀⠀⠀
⠀⠀⠀⣀⣾⣏⣯⠷⡀⣠⣖⣺⣿⣄⡀⠀
⣠⣤⣿⡟⢿⡆⠁⣀⠃⢡⠉⢣⡿⠿⣿⡤
⠛⠻⠅⠁⠘⣽⠀⡍⠒⢊⠀⣳⠃⠀⠉⠁
⠀⠀⠀⠀⠀⣿⡄⠙⠉⠋⣸⠕⠀⠀⠀⠀
⠀⠀⢀⠀⢸⠐⢻⡤⢤⠶⡛⠀⠀⠀⠀⠀
⠀⠀⠀⠘⠡⠈⠀⠄⡋⠔⣫⠀⠂
aha, got it, thanks bro
ur wlcm
smol potate
⣿⣿⣿⣿⣿⣿⣿⣟⣿⣿⠟⠛⠙⠋⠙⠿⠿⠿⠍⠙⠋⠛⠙⠋⢉⣻⡹
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠐⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⡯⢝⡮
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣈⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⡊⢙⣪
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢯⢼⣻
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡗⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⢾⣿⣿
⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠏⠀⠤⠀⣀⣀⠀⠀⠀⠀⠀⠀⠀⠀⢸⣿⣯
⣿⣿⣿⣿⣿⣿⣿⣿⡿⠁⠠⢈⠆⣝⢶⡻⣞⣟⠶⡶⣄⠠⢀⠀⢸⢹⣿
⣿⣿⣿⣿⣿⣿⣿⡿⠁⠀⡁⠎⠞⢼⣳⣿⣷⣾⣯⣷⣬⡓⠄⠂⢸⢸⣿
⣿⣿⣿⣿⣿⣿⠿⠅⠀⠀⡡⠤⠰⠄⠉⠹⣿⡿⠟⠉⣀⡉⠀⡁⠀⠙⢿
⣿⣿⣿⣿⠏⠀⠀⠀⠀⠀⠵⣀⠰⠤⠂⢈⡾⠁⣠⣄⡀⠁⠆⠀⠀⠀⠀
⡿⠀⠘⠛⠀⠀⠀⡔⡂⠀⣙⢶⣿⣾⡟⢠⣿⡀⢸⢾⣵⢢⠃⠀⣀⡀⠀
⠁⠀⠀⠀⠀⠀⠈⢾⣆⠀⠣⢞⣿⣷⠀⠻⣿⠟⢌⣿⣏⢣⠁⣜⣿⡲⠀
⣦⡀⣀⠀⠀⣸⠷⠀⢥⠄⡙⠈⢟⠑⣭⢓⣴⣂⡊⢻⡙⠢⢀⢟⠛⣧⢾
⣿⣷⣿⣴⣿⣿⣿⣦⣞⡃⠸⡙⢼⡀⢀⣈⣁⠉⢀⠞⢁⠍⣈⡍⠀⡿⣿
⣿⣿⣿⣿⣿⣿⣿⣿⣏⠇⡀⠀⢹⣿⣦⣙⣈⣤⡾⠃⠊⢰⣿⠇⢂⣷⣿
⣿⣿⣿⣿⣿⣿⣿⡏⠲⠑⣆⠆⠈⢛⣿⣿⣿⠿⠁⢠⢀⣿⣿⡄⡈⣿⣿
⣿⣿⣿⣿⣿⣿⣿⠧⢀⡻⣬⣏⠶⣀⡄⣄⠁⣠⠜⠂⣻⣿⣿⠃⠐⣿⣿
⣿⣿⣿⣿⣿⣿⣇⠰⣼⣹⣷⣿⣿⣱⢟⡮⣜⢧⡻⡀⠤⣷⣿⣌⠀⣿⣻
⣿⣿⣿⣿⣿⣿⣿⡮⠓⡝⠞⠻⠻⡙⠮⢓⠹⢚⡑⢁⢊⣿⣿⣿⣿⣿⣿
Lmfao
does pplx ceo aravind ever lurk/chat in #💬│general ?
Nah
Never.

The official was smashing so his son replied.
Le phortnite ar pubg
No way embed failed.
............
how tf pplx turned into meme server
Better than a racist server which was/still is time to time
yeah we meme...
...on ai and internet culture
Mat kar Lala Mat Kar
Is this ai generated? Cuz the electricity poll is wiggling
💬 14 🔁 0 💜 502 👀 17.4K
MRo ;
KcuF you Very Much #Milliejackson 😝 ...
The middle Finger is universally recognized
#Confidence #Naughty #middlefinger
#fuck #love #memes #like #meme #follow #sad #happy #instagram #lol #dankmemes #funny #art #love #attitude#guts #music # #anime #life #edgymemes #likes #funnymemes #instagood #me #mrosmalls
eyo chat
Is perplexity good ? , if so in what ,m
deep dive ➤ #💬│general message
Do yall have pro perplexity
Nah , give me summary
Indeed
Not even close
So u mean perplexity is better?
Agreed
damn, didnt see that somewhow. ty.
U must have a Airtel sim lmao 🤣
Good 👍
Better than jio atleast
i have dth frm there i got
I notice that when I use perplexity as a search engine on a non-comet browser, there is a slightly different layout for source results (left is safari, right is comet). is there any way to get the comet layout on other browsers?
Not possible
rip
I really want to go back to safari but it's so hard to do now that I got used to perplexity search
it's just so perfect
and it's way worse on other browsers
I guess props to the perplexity team for making a really good ecosystem
ms edge sidebar thing still pulls me back to edge lmao
bruh edge browser free for all windows users
lmao
I'm asking about the AI features
Hi Guys, I would love to use Comet, so please if anyone have a code for me i would be thankful.
yeah copilot free even deep think is free
alr
unlmited
i remember using deep think in copilot for a large project and it didnt even hit the rate limit more than 100 requests i made
go to comet browser website click on join waitlist
they will give u invitation eventually
but it takes time
no if they accept good if not just wait till end ofaugust maybe at that time they will make it public
Tsunami waves arrive on Japan's coast after earthquake in Russia, triggering Tsunami alerts for US West Coast
Tsunami waves arrived on the eastern coast of Japan following a powerful earthquake off the eastern coast of Russia. The 8.8 magnitude earthquake triggered fears of a tsunami across much of the Pacific Ocean basin, including the shores of Hawaii, southern Alaska, the entire U.S. West Coast and British Columbia.
People in japan, stay safe

US, Russia and Japan!
Stay safe guys!
Billy, it's Strongest earthquake since 2011 in pacific
how

Yooooo
Airtel wifi subscription
i wanna be a bug sniper too 🙏🏻
ahhh... airtel
Just post a bug in #1140622008086970420
Yo ?
How much u earn , by hunting??
finally
Tie up
How come that be a issue
Even thow I didn't , it was automatically upgraded
Check the mail
a generous 0$

nah, idk
but think abt it this way, its fun. & it helps improve the browser w eall love

Contact support@perplexity.ai
What is this about i even saw a video on YouTube saying "he is back"
Can u send me invite pls
Really u do it for public satisfaction
I tried many times to upload a picture with text or even have the plain text as an input and asked to create an image and the output is always unreadable. Do you know what happened?
What is the best way to have text as an input to create an image (onepager a4 to print) with that content plus some illsutrations etc.?
(I use pro)
I mean, if it makes the product (something which I use on a daily basis) better, I don't see any reason not to provide feedback when I encounter a bug
I think it's more about liberals throwing around the word nazi out of context.
Whats the prompt you are using? Image gen seems to be fine for me. Following image was created with perplexity
O yeah I have seen that happen a lot on context that has nothing to do with natzi
tried many different ones. but basically to create an image that contain the following information: [text]
And then the output is some unreadable letters
Weird eyebrows and excessive hair on the chins
do you have an example
also switch to gpt-image-1
i didnt do any prompting lol
was just "make a poster that has the words perplexity"
What if he found it in your memory, "your imagination of perplexity assistant" 
Weird berd moment
Lmao
I have memory off :(
Not the output, the prompt you used.
Also why are you having the image model make a table?
create an image that containts these information.
and then attach picture
it was just to try, but wondering why these spaghetti words coming out
Try using GPT-Image-1, (change in settings) also models tend to suffer when generating a lot of text such as a table
actually i want to create onepager for certain topics. input: lots of information, maybe some images. output actually a onepager pdf with illustration. i cannot do pdfs he said, so i tried with an image. what do you recommend?
Have it create a table, then ask it to create an image
also tried that. the picture i uploaded to perplexity was actually a Screenshot from the table he gave me😂
output was spaghetti words😂
where can i change it? dont find it in the app.
if you want to export information, you can do that at the bottom of the output
You need to use the web version
yeah i can export the table as a pdf but cannot let him create a pdf onepager for me right? just text/tables or images, correct?
Maybe upload a csv file. You can often download tables as csv made by perplexity it's on the right bottom corner
I'm not sure what you mean by onepager, the pdf exports everything in that message including tables
i meant i give a lot of data (text, table, picture) as input and want him to create a brief a4 overview of this topic as onepager (one page with the most important information)
I'm not too sure if it will be able to contain all your information into an a4 page every time, I'd just export it as a docx and edit/tweak the format to make it fit.
But it workes now with gpt image 1
changed on the web
using again in app now
Thank you!!
Cool
When will comet be released for Linux and Android?
Most likely at the end of this year
Not sure regarding Linux
It would be nice but I'll have to use it wine
Ahahahaha
Hi
@obsidian leaf
@burnt iron Stop tagging me or I will report
the hype building up for gpt-5 is real
why is it so expensive to just get the video generation feature ?
@green crystal all for incredible rate limits
$30 per million tokens
gotta show your bank statement to them so they can decide if you have enough money

Dude stop mass pinging
it's very compute intensive, often many times more than text generation
we're pushing back gpt-5 again. it's just not ready. we're close, but close isn't enough.
when it does launch, pricing will start at $30 for 12 tokens. yes, that's real. no, it's not a typo. we think it's worth it.
to make sure everything goes smoothly, we'll also be quietly checking your bank accounts. it's important to us that you're aligned, financially and otherwise.
thanks for being patient. this is going to be weird, and fun, and maybe a little expensive.
more soon
Sam Altman via X
lmfao
Lol
dude I was just looking at X and saw OAI posted
lolol
was confused to why they would ever release something like this
it was instead something about brainrot 💀https://x.com/OpenAI/status/1950572629670412376
Yep
Why does OAI need a podcast
emphasis on the "how"
Yeah if you're elon you just wanna goon to annie because you're sad and lonely because all your 69 kids hate you
Idek if annie was the x gf ai bot name
@timber niche You've made youtube start recommending laufey to me
Why did you blur it furry mat
Moderators, this person @burnt iron is pinging me and I don't know who is he. Can you please help.
yee, saw this along with gpt5 spotted in macos code
totally real and not made in 5 minutes by me
What's that supposed to be
@halcyon coral Please help this man out
uh
Furry?
LAUFEY SUPREMACY
Bro
That gif didn't load
So it looked just like the cat annoyed
Lmao
Hmm, why don't you show us some of your other recommendations?
Maybe because they are furry videos?
noo
Mat is a furry vtuber
but one of them is the openAI podcast
so you aren't denying furry videos showing up?
I am denying them!
Their o lineup is supposed to be for reasoning models
I mean, it was supposed to be a mixed model
Hm
Too late
So a gemini flash competitor
It's a unified model
But way powerful
Wtf is that tech support furry mat
Nah
more like 2.5 pro
More evidence
Hm
Mole rat
this probably forces the model to always reason no matter what
Yep
even for dead ass simple prompts
Mhm
What will trigger it
so when i ask it 1+1 =2
Will it be an option
lots of self doubt will happen to answer it
I'm guessing a smaller model evaluates the query, then determines if extra reasoning will be required
or its baked into the training
Noope

Answer me
what did you say
Are you a furry Yes or No. You get 5 seconds to respond
.
NO
probably this
Bit too late there
it does both reasoning and no reasoning
o4 minin does it
Why and where
Mhm
simple questions are near instant
also kinda like qwen 3
anything else becomes longer the more complex the model thinks the question is
Why can't they make a new line for unified models
and uhh GLM 4.5
Instead of fking up gpt lienup
o3 pro just maximizes reasoning length for everything
Lets make the current naming situation more complicated
and introduce a new lineup
Intel intensifies
now we have o3, deepresearch, pro models, gpt, gptimage 1, dalle, whisper
The U models would go hard
gpt-5-core-max-ultra-0528
Fr
Gpt-5 pro max plus ultra 2tb 5g 100w fast charging 50MP main camera 5000mah battery
And now I'm being discriminated??
Furries get their own superior version
Nah, we never discriminate
Exactly, just like Grok Rudi in voice mode
Gpt 5 furry ver. Let's just say get's crazy
Matsku as a vtuber
what is this
what im surprised about is why elon couldnt just hire a 3d modeler that makes models for vtubers to make one for him
@gloomy cipher sam altmans baby
those would make ani allot better
who is the husband?
Claude Sonnet Thinking
No grok 4?
I thought it was anti "woke"
grok unhinged is fun tho
Hence it adopted elon
It's the florida gay
its literally like its having a constant mental breakdown
I mean, I feel like it's just prompted to throw every swear word at you
Right wing republican that is secretly in the closet
I'd just like a semi uncensored model that doesn't try to immediately vent at me
with the power of a frontier model behind it
gives it a bit extra spice
Ai venting is crazy work
mhm
heyyy guyss
Hello
how you doin man
I'm doing great mate
I'm not letting mat take off that bucket
My mans
One day I will be free
Never
one day when I get bullied enough in other servers
You will live with the bucket
You're cute with that
Mood
That's not exactly what irl people think
Lmao they are chill
free from bullie, free to be every fursona as you wish 💅
Chill people can't not say your pfp and deco ain't cute
You're gonna make me purge my photos at this rate
Nah
I'm sure you have too many photos (of furries) to have the time and effort to delete them
Or we go operation full swing birdo motion on you
Bro is lazy asf
Prob
Furry!
what is catalonia?
A region in spain
Also what on earth is uspen thedral
LMAO LOOK AT THIS GUY
and how do I pronounce that
Some old cool lookin gcathedral
Oh
They're chilling
it looks like he ate too much
Yeahh they did
the trianguler pyramid of life
Don't listen to them berd
Mat eats berds
Do you eat berds
What do you use to blur images
it was images of berds fighting with cats
the first app i found when i searched edit image
on the appstore
and also photoshop when I have the time
i heard there was a channel for pro users. how do i gain access?
@halcyon coral Wait for staff to manually give you the role
Beat me to it
Hi
hehe
Hey Hero!!! How is your day
???
Take some rest
You zombie?
Thinking I'd miss the train
Insane
Baddie what

They have very good restaurants and Leo Messi. 🙂
Yeah it's a crazy time to have a train
True cat lover
I think we have a winner @short narwhal
Lmao
i think ur pfp is enough

I love my cats haha
🤝
Mat doesn't even have cats
Can I borrow one
I'm manifesting
Who's mat
Yeah they devour it
Nopeeee
i had a kitten for a week
@short narwhal this guy
@short narwhal our resident furry
Oh damn
No racism ✖️
Oh...
Hello uncle
Poor Mat
Nope
i think it was better than posting a fursona meme
Nah, he asked for this
at least the BUcket is legit 🪣
wait really? @short narwhal u like getting bullied is that true?
See, no response means yes
nooo
What does yellow dot mean
you took one whole minute to respond - doesn't count
Mr. Mat it's nice to meet you I heard you like getting bullied.
I have a deal, we can swap one of your cats with my cat Jelly Bean @vague pagoda
not by peasants
It's thinking maybe
Real
It just follows your comet color, normally it's blue
LOL
Ahhh makes sense
ehh
You are about the face the wraith of Kesku
disappointed in your response times, considering you have notifs on
My walking carpet that meows is not for sale
mb I'm gonna lock in from now on
Lets play a game then of me pinging you randomly and seeing how long it takes for you to respond.
uhhhh okay
Does mat also work in pplx ?
I will be incredibly generous and give you one whole sunflower seed for each time you get within 5 seconds.
I really appreciate that
how kind of you
Having pplx max I realised it's not about the ai it's about taking productive action getting things will not solve your problems 😭
Have you watched suits ?
At $200/month + VAT, you might as well use the API's directly, with a better scaffold around them.
Completed whole show
Have you watched Dexter?
Is dexter technically a vigilante?
I am watching it I have watched 1 season of suits tho good good
Hmm 🤔 you could consider as such
Actually he is
Watch more..
Dexter: Resurrection
Released this year
Total 8 seasons
Yeah
11
11? okayyy
Dexter 8 season
Dexter the resurrection +1
Dexter orginal sin +1
Dexter new blood+1
Good to hear 👍🏻
Hmm sure
depends on your usage, if its heavy usage with opus 4 and o3 pro, it would def be more worth it if you bought the sub instead of over the API
Incredible translation work by grok
Like where did it pull the sob from the og does not have that
Don't know why you would want those models though. They are both pretty slow. And o3 pro basically works like Grok 4 heavy, where it is just doing the task x amount of times, before choosing the best answer.
Opus 4 isn't too slow? and o3 pro if you are solving complex problems/research
You are probably just used to slow models then.
I know people who would have spent way more than $200 if they used it via the API
What would you count as fast then?
in my experience opus 4 is pretty similar to o3
100 tps+
Yes, if they are using it just with a chat interface. Since their context gets longer and longer.
Intelligence wouldn't be very high then, or the API costs would be quite expensive
I mean, just talking token wise excluding the among of context, I've seen people spend upwards of $1000 worth of API costs on o3 pro
and on Claude code
Kimi k2 is a 1T model, and runs at 300 tps on groq...
There are also the newer Qwen models too.
kimi and qwen are quite a bit different from o3 pro and opus lol
Its much more expensive on groq though
and its not a reasoning model
Not compared to other models in general.
Reasoning model does not mean it's good.
It just means they do RL on CoT
Usually reasoning models perform better than non reasoning ones.
Nope, depends on the task.
Ask it to make ascii art, and compare GPT o3 or o4-mini to just vanilla GPT 4o
LLM's tend to mess up on ASCII art in general
Reasoning models are a good way to reduce the cost of training performant models, but that doesn't mean if a model is not reasoning, that it's can't beat a reasoning model.
Same with MoE vs Dense
i think its safe to say the reasoning significantly boosts models their abilities lol
non reasoning will still exist for very simple cheap high volume tasks
Nah, it's just that the latest models use reasoning because it's an easy way to improve an already trained model, without needing more data.
There are other ways to get a similar performance boost.
Train your own model then, talk is cheap
I actually think AGI will be achieved by a Minecraft mod
I am simply comparing the performance of opensource models where they literally tell you the techniques they use and the benchmarks they got from those techniques.
open source uses non reasoning ones as base yeah
same for propietary ones to a degree
but there's quite a bit added and changed
o4 mini used 4o as foundation
but i dont think 4o would be capable of reasoning with images, files and tools at the same time just by telling it to
With a dataset that focused on it, it would.
Reasoning is increasing the dataset too, but through RL instead.
Didn't openai release GPT4.1 literally for that?
The problem that reasoning models have with tool use is the time it takes for them to call their first tool, compared to none reasoning models.
I'm gonna test this pretty quickly, since 4o should easily be able to do that.
the o series models are as quick as non reasoning when calling tools because it all happens in their reasoning chain
I'm gonna give it an image with items inside of it, and a csv file, which has some of those items in a list with a price, and ask it to give me the total price of the items in that image. Should be a pretty simple test.
wasnt it for long context?
I don't believe so. Mostly for more agentic tasks.
More context could just be a part of that.
It also has better tool calling, code gen and other stuff like that.
apparently its for a more cost efficient and longer context model for coding
Yep, but specifically for editor integration, which required good tool calling. Probably the reason you can only get it through the API.
Yep, I think they added it when people were complaining about it.
capped at 32k for plus and 128k for Pro though
Can anyone tell me which is the best model to use for the problems? Like which would be best, which can give a answer with good accuracy?
For STEM problems, use o3
I am asking for like mathematical problems
o3
o3 is somewhat unstable I guess, it doesn't listen to prompts correctly!
STEM = science, technology, engineering, mathematics.
Try out sonnet 4 then
How can I use it? Does it have 100% accuracy while giving answers? I am a pro subscriber already
No model has 100% accuracy, you can switch models here
click that computer chip looking button
I have used the claude one in the reasoning. Should I use claude 4.0 sonnet for best?
Use the reasoning version
does someone know if theres a way to sort the #1124060874785685554 by number of reacts/upvotes or whatever i thought there was but cant remember
Is the claude ( reasoning version) best for STEM?
I prefer o3 but Claude sonnet is a strong option as well
Ohh, but why isn't o3 stable? It really doesn't listen to prompts correctly, even R1 1776 was one of the best which is removed now
I haven't found o3 to be unstable, might be your prompting
Check dm plz
@ask
I'm so upset of R1 being removed. Best model to check for chinese text in relation to android queries
dammit
@short narwhal i found what you do for a living
@agile jay how did that perform?
It did it easily.
People underestimate the none reasoning models for normal tasks.
................
it was image generation model
Which one?
Amuse
very close to 100% with tools
the only thing i like in perplexity is the UI, but not the quality of results, that UI always looks like i have to buy this product. after purchase it feels like why i bought this product
Yep, only problem is sometimes the model is too confident and doesn't use them
Oh, Amuse would generate the image and then crash because of some API was missing at the end. It was running in emulation.
a few times it deemed something simple enough that it told itself that its fine without tools lol
it'll do it by hand

Yeah....
gotta tell it to do so
Yep, system message can change the behavior a lot.
Bro why the image generate is so bad
try switching to gpt image 1 under settings
everything seems fine
I see so still can't run Ai stuffs on arm cpus
I can, just not through that client.
why I can't see "max" models only in site
You still have access to R1?
everyone does?
im go ask #1140622008086970420
Hloo
Max is the subscription the 200 dollar one
@gloomy cipher https://www.instagramez.com/p/DLvE4zUxFLW
You need a max sub to use it, makes no sense in showing you them if you don't have the sub, esp in the mobile app
It shows regardless.
yes makes sense
Thats not what Rydox is saying.
but people use app more, maybe they need to show it
Ik it shows on the mobile app
I can't see it on pc
I'd try updating and clearing cache
Isn't pc version electron wrapper?
naw, it's missing for me and @final echo at the very least
that is intended, r1 got removed
you can find it on perplexity site it doesnt show in app like o3 max
Even if R1 still pops up for you, it won't work. you will get routed to another model
ah ok
Lemme know when gpt 5 comes out.
You aren't in the OAi server?
ah
2 mutual servers
openai discord is like... too formal... bruh
there is also a reddit community oai server
maybe thats were the action is
where all the moots at
moot?
mutual followers?
Oh oops
fr
also trust urbandictionary
true
It's just a custom system prompt
nothing new
I think they're really overhyping it
this word better in hindi dictionary
I can't read hindi
You could learn it
k
For the bit
That seems like it would be a pain
to learn the language
Ah, my fellow berd that I want to uh "unalive"
Hindi
Comet invite possible ?
I swear Duolingo got the best marketing team 😭
Like bro they literally post the weirdest content on yt
I have got one to spare. Dm if you want.
Have you ever watched Duracell’s marketing after Tom Sandoval cheated on his ex-wife?
No, why ?
Now that’s incredible marketing, will never buy any other batteries for the rest of my life.
"incredible" how, can you describe which type of ad your refering to or what type of ad was it ?
Just got Google AI pro subscription for free. It doesn't seem worth it at all. 😉
Did anyone try the new ChatGPT study mode? It's actually sick.
Waiting for perplexity to give us something similar like it hopefully on comet. 😩
It's just a system prompt, nothing new
You can put this into a perplexity space to achieve a similar result
This is the system prompt they are using for chatgpt study
Is that so? Will try it tomorrow.
Veo 3 and 2tb storage make up for it 👍🏻
Agent is pretty dope as well
So when is the bug of not seeing the model’s thoughts be fixed
Comet or Agent Mode
Noice eddy!
Just the normal search
hello
New comet widgets hopefully soon
Hi everyone!
I was wondering if anyone here would be interested in a little invite exchange – I have an invite for the Dia browser and I’m looking for one for Comet. If anyone’s up for swapping, just let me know! Thanks so much in advance 😊
Can I talk to perplexity like in ChatGPT or is it only research?
Yes u can
U got pro or free?
Free currently but I am thinking about switching to perplexity
Would you recommend it?
Instead of ChatGPT plus
I have never used gpt plus
But stay here for a while there are others who can tell u better
@dry kestrel
@short narwhal
I mean I just switched because plus ran out, I like it because you get access to multiple models, I think the only different is token size?
The use of different models ist also very interesting for me. Is there a limit?
You can, but it’s not really meant for that
300+ uses a day
There is a context window of 32k, same as ChatGPT Plus
beat me to it
deep dive ➤ #💬│general message
check Tutorial II
ngl we should pin this
i got an email from perplexity that I got invited to comet, but when I click the link I just get to the main perplexity page, can anybody help?
Berd that reaction was not intentional
Was it logged in? To ur account
And u are doing this on ur pc? Right
And was it the right account
Yes got the email to my account and logged in with that account, all on PC
BREAKING 🚨: Perplexity is gearing up for the GPT-5 release as well! GPT-5 Reasoning will be available to Perplexity Pro users shortly after the model release.
︀︀
︀︀And the release is closer than we think 👀
Ahm but there is no such thing as gpt5 reasoning
Right??
Definitely seems like OpenAI has it ready, just cleaning it up and maybe wants to do a really clean release.
It has been spotted in the code for the ChatGPT desktop app
this aligns with what Sam Altman has been saying as well and with what we are seeing with these other leaks
<@&1098365634498285709>
and can I please have my free bitcoin
lol
its just a mere 117k usd
Surprise its max subscription. No thanks
Me likey
Can we half it
Um no, I need to buy more birdseed
I don't think it will but I guess we will have to see. It may just be because they are a max tester.
It may also depend on the reasoning effort setting
and the exact version since theres a nano, mini, and full one
I'm guessing the one on Max will have maxed out reasoning and be the full model.
Where money goes people follow. I doubt. Money run the world bro
It will all depend on what OpenAI prices it. They are not going to overcharge on something you can get from ChatGPT for cheaper
Didnt even feel o3 pro in perplexity.
Expensive model :/
What makes you think that it’s max only?
BREAKING 🚨: Perplexity is gearing up for the GPT-5 release as well! GPT-5 Reasoning will be available to Perplexity Pro users shortly after the model release.
︀︀
︀︀And the release is closer than we think 👀
the X post makes it seem like a MAX only feature
Doesn’t to me
due to the user being a Max subscriber
Yeah people gotta just read the description
GPT-5 isn’t even out, the subscription doesn’t have it
Erm actually
I’m sure testingcatalog could get it to be visible on a pro account as well
I mean, they said "GPT-5 Reasoning will be available to Perplexity Pro"
In the description
Yeess
They said max
One more arrow
"a little" 😭
lol savage
I can see it if I squint my eyes
Hello
Hlo
the right leg doesn't go around the left leg anatomically but gets "blended in". its convincing though. i used to do animation (2d and 3d) so i catch these things.
there's some other generative artifacts but the motion paths pretty good
There's rumors about it coming out tomorrow because all releases have been on Thursdays
yep just in case something is wrong they have friday to fix
I don't think that's the case though because last moment announcing won't bring them much views as just announcing it right now
Hmm, Im guessing on the 5th, if it was Thursday they would have tweeted already
august 5, gpt 5
Yeah
2025
5pm
They should have releasted it on may 5th
ah we are on the same page
- the verge said early august, end of july wouldn't make sense
I hate ios26
wth
Keyboard blocking everything
true
yeah i know watcha mean... im on ios too but im on laptop
Chat is dry
lemme wet it
new models landed in LM Arena
I liked last one
Horizon alpha is now available
it is a stealth model on OpenRouter
Imagine if it was gpt 5 alpha lol
Is agent down?
Even image upload isn't reliable atm
Suddenly my agent requests went to 0 lol
Ffs right when I actually need agent for actual work...
I have personal super stupidity.
Thnx for sharing low effort link
Good thing I know spanish
This went on for a while until I pressed the stop button. It always does this whenever it starts with 'OHHHHHH.'
Whatever guys in the replies just saying it's o3 routing, not gpt5

Good Morning 
ngl horizon alpha doesn't seem like GPT5
Good morning fellas
@dry kestrel bro when do you sleep
You are awake at all times 😭
Good thing I’m not in public
Also
Just a little trick, if you guys want to use study mode from openai in any other ai, just use this prompt, is working very well
I am not a big creator but i already made 9 with my ai arts, maybe I am lucky for now
9 what?
Never
Is working very well, is on my language
million
No iam asking what that is
Is it sharing pictures or something?




