#💬│general
1 messages · Page 80 of 1
Whats the software that you run?
You could also run smaller models at their full size but once you get up to like the bigger models like 72b or 1 2 3 then it's like using claude
Msty I would link it but I don't feel like that's classy
Okay. I'm going to need a tutorial on this
Well that's the point it it's simple enough that you can kind of poke around and learn at your own pace
But if you need help learning the interface perplexity can help explain it
Hmm
Most of the annoying stuff that will make you pull your hair out it simplifies
There are more advanced front ends for local AI but that one is all in one and it's simple click and install and ready to download models
Okay. I'll look into misty
Just remember if you're downloading new models you're going to want to look for some that are in between probably 16 and 20 gigs total size and don't go below 4K quantization
That's the difference between your computer crashing or things being super slow. Also keep the output tokens low like 300 took me weeks to figure that out
you want to keep like 20 or 30% of the 24 gigs free for its memory so get the biggest ones while leaving about 4-8 GB of your graphics card
So are you just feeding these models information?
Yes lots of it
Since it's local I basically told it everything about my life
I've done all sorts of weird projects it's wild how good it works
I just have a RTX 4070
To be able to give you suggestions for improvement?
Or just information about your life for it to learn?
They can give you like suggestions of like what to make in your house whether your chicken salad went bad it can help you get a job it can help you get through your life problems yeah all sorts of crap
And it's not a Google's business
Even through images?
I think that it does have an image uploading feature
Okay
But you would need to find a front end like silly tavern that has flux integration so it can generate pictures if that's what you're wanting to do but I'm working on that one for my own ideas
I would not suggest messing with silly tavern
It took me a year just to figure out the menus
And by year I mean yesterday
Since we're cooking up ideas. I'm thinking of a personalized agent that can help with a multitude of things
Let me show you my idea
But seems like you are on track
It's a few steps past that
I don’t get why perplexity in mobile phone is far less detail and informative then on laptop, or desktop
Opus has been removed?
Hey @barren saddle!
That's right Opus has been removed as Perplexity always stays up to date with the best LLMs. Currently the best models from Anthropic, Claude 3.5 Sonnet and Haiku are available.
hello
Hi 👋
i have a problem with a api key i dont know where to find it. I Dont understand what he wants
Sorry but idk how maybe ask it in #⚡│ask-community
Thank you!
No problem!
Is anyone else having this weird issue with Perplexity today? After the model completes its response, the stop button icon stays visible but doesn't work when clicked. Can't send any new messages either - have to refresh the page every single time to ask another question. It's happening with every query for me. Anyone else experiencing this?
To generate an API key, you need to go to 'settings/API.'
ive got to pay right?
With Perplexity Pro, you get $5 monthly to use on pplx-api.
Hi
Do any of you use nextcloud or owncloud? I'm currently on proton drive, I'm thinking of switching, is it worth it? I don't have another device, server or home internet capability. 1 laptop, phone with mobile internet and I'm not permanently housed.
I mean, nextcloud is self-hosted, which is basically the whole point
from what you describe, you want the exact opposite, lol
proton drive is probably way better for you
I use nextcloud for some stuff, but only because I have a server at home
Claude 3 Opus too old and too expensive
Claude 3.5 opus coming soon this November
Will this month we will get Claude 3.5 opus?
Your profile perplexity, click the word api, it generates api button and copy it and key in your api key that shows in your picture
I don't see the API keys pop up, I think it's because my payment is via mobile and not direct maybe?
$5 api is free for pro users, go to the website at your profile, click the api and click the generated api to get the key
Please note free for $5 you need to inform @signal hamlet to get it otherwise you can’t use the api
yeah it's not on there it's prompting me for a CC
add it to the pile lmao
perplexity supports pdf with how many words per file without losing context?
Does anyone know if there are any other services that are like Perplexity with Opus? I ended up canceling my subscription today since all the other models are pretty crappy compared to Opus. I don't get why I'd pay for it without Opus if CoPilot lets me do the same thing for free.
Perplexity stopped usijg one of the models, when I checked it jusy wasn’t using one
Doesn’t it annoy anyone else on iOS that the app doesn’t open links if your browser?
Or do you just all use Safari?
Even if I used Safari, I think it would annoy me that links I open don’t appear in my history
Anyone else having issues loadiing perplexity?
i cant
yea
idk why when I open a thread it says: "This Thread does not exist"
new recent bug
because we've had this bug for 5 days where threads become inaccessible when they're too long
and it's literally making it unusable
now , whole website is broken
Having issue pulling up my spaces and logging in
but it's still not fixed
oh fantastic
good work
somehow like this
MY PERPLEXITY CRASHED 😭😭😭😭😭😭
hopefully they're pushing an update or something
SAMEEEEE
Well im glad im not the only one getting issues.
'Hopefully'
downnnn
i think mb they are updating
so cannot access
yeah it said updated website
Keep getting access denied with logging in through Google.
amd then it sid this
and then it said this*
They prolly have to manually update it 
hey you never know
could be a new bug instead
Hopefully bringing Opus back 
Is the site down
nothing would surprise me anymore
raise of AI XD
I thought I was just going crazy for 10 minutes as it forgot a 6-hour long project conversation and then remembered it again 20 minutes later
the android app doesnt work either
For a company that is very AI centric they are not very proactive
Sure would be cool if the Admin's would say something.
Update workflow:
- Turn off server
- Take out old hard drive
- Put new HD in with update
- Turn on server
i guess the whole perplexity somehow having bug now
we just got perplexitied
It's the New York times ddosing them
past week got stability issue problem you can visit bug-reports #1140622008086970420 for all the fun
literally no website AI ever making this website more bug than perplexity
404 page should just say are you not perplexed?
may be it is human error 😄
Someone is posting stuff there now :|
Was studying for an exam and forgot to take ss 🙂
the AI became sentient and realized taking itself down was the best option to preserve mankind
@signal hamlet cmon man any news?
I got the info that it's updated to the latest version but I'm now free user and there is no library ...
Not the only one?
same here
If they are updating now this is such a bad practice, do not release something on monday
down?
same here
dude monday is WAY better than friday like last time lmao
its down
It told me I lost 6 hours worth of project that it does not exist freaked me out
at least they'll be in to fix any problems
its not working
very fair. both days suck. lol
I got this after trying to sign in on the web just so you know, I downloaded X search and set perplexity as my default search engine and now I cannot log into perplexity through Safari however, my PWA version of perplexity which is running through the Safari browser it works just fine, but using the actual Safari app, I am not able to login to your perplexityI have disabled all ad blocking and tracking and this is still the case. I even used Apple sign in and it didn’t work and I am trying to type in the Apple relay email manually and I am still not getting a sign in code.

Deployment issue
Prayer circle for them to not blindside us anymore 🙏
sorry, i have a problem. is down the system from desktop? thanks
It was a deployment issue. We’re reverting back. Sorry. Will be up and running very very shortly.


MAAAN
They need an API to send these updates to the discord.
Spaghetti aaaah codebase
THISSSS
its over
Perplexity is down
that is how discord chat should be XD
Servers are down? I cannot log in
why i'm enjoying the situation ...
There's to hoping this is them in the process of fixing the bugs
Deployment issue. annoying but happens.
We're lonely
welcome to the club
also joined to complain
just wish there would have been more communication from the perplexity team honestly.

they are really bad at communicate to the community
This is their biggest problem imo
Hi everyone, please be patient.
Pplx will be up and running soon 🙏
reached our capacity is such a funny way to put it. like waht do you mean, you are the capacity
Haha yes!
web3 is not real .... so ... hahaha
yeah sure
i mean would have been more patient if you guys would have chimed in sooner honestly.
same
my assignement due in 30m ☠️
pplx, create a bot that tells if you are down 👀
Website is up
hop to chatgpt 🙏
Hi
we are so back
They went down right as I canceled my subscription? This is poetic.
cluade !!
its back gg
It still seems to be working for me right now
time to write a case study
I love spam
chat are we back
But I still can’t log in through the safari app
🫡
they are cooking
Was this a promotion tactic to make us join the discord
YOU did this lol
Waiting

Not buying pro again
WHAT HAVE YOU DONE
Don't know if it's a happy moment for Perplexity - they maxxed out their use frequency, or if it's a technical issue
got me haha
ok, that was quick fix , good job
don't forget to fix another bugs too ...
Good luck fwends
There are no bugs, just extra features with a bit of spice
I'm going to take this opportunity to ask if anyone has an alternative with Opus support

yeah but don't tell us you have a capacity that gets reached if we ask too many questions, we are supposed to believe you can handle anything. just say technical issues sorry we'll be back shortly or whatever
gl you got 25 mins !
Its back 😄
and chat become lonley again , they all back to their space
you guys will be missed
🤘🏼
People lost there friend for few mins
but dont worry, your friends (AI models) are back.
Im going to use perplexity to search for an alternative to perplexity, but with opus.
cya all
chathub has opus.
but does it work as well
It works yeah ofc
I tried You.com, but I had to request a refund to Apple since they took my money and I didn't get my subscription
You get free trial on chathub, feel free to try.
Check your DM,
ok, back to my space ! have a nice day
I just need Opus. Everything else is crap
Sure it has it, feel free to check it.
also offers o1 mini so yeah.
Best models
and it also has Perplexity Sonar if you would miss it 😂

Fast enough solved 😄
You use Opus for writing?
Did you have any luck with changing your email?
Nope.
Ah whats your use?
Mostly for 10-Q and 10-K research
...is the PPLX CEO unironically shilling on twitter to the new presidency?
good LORD.
there's low and then there's whatever this is
Why not provide for your country? And Perplexity has the money to do it.
They basicly did it to the normal people too, with the voucher/coupons for 1 year free...
making an offer in private is one thing
publicly at-ing them is just tacky
It stays a company, it shows its providing.
Hello everyone, may I ask question regarding subscription here? Or is it against rules
Hello, my thread has disappeared with all of my many threads. please assist me in how this happened and how I get them back
whoever it was had the bright idea of pushing whatever change caused it on a friday, so it's been days but it happened over the weekend
I'm hoping the botched update they had to roll back was meant to fix this issue
just... gotta wait for them to actually push it again.
They don't address problem either
they reverted it
like they tried updating the site
it failed
they rolled it back
I just hope it's fixed like
very bloody soon
because it's extremely frustrating
high quality answers, something republicans truly care about
Hey! Sure, it's allowed 🙂
https://perplexity.ai/pro?referral_code=TQRFFK3J
$10 off subscription
The coding sync session is running now. Feel free to DM if you want to join. We work on independent projects, 30-40min sprints, 5-7 syncs and mentoring in between. 100% free, no ads. That helps me to stay focused on my own deliveries.
thread "updated to the new version" by itself mid-using it and now i can't see A LOT of my last messages
can anyone confirm if the "thread does not exist" error is still persisting for you all? I just refreshed and tried accessing an old thread that was affected by this and it finally loaded and I can see and continue that thread. Don't know if this will be permanent or if I just go lucky.
They ""Fixed"" it
except NOW IT DOESN'T SHOW YOUR RECENT MESSAGES.
oh brilliant, okay
you can easily see this by asking a follow up question (in a thread that was long enough to cause the bug)
reload
it will be gone
before it used to hide old messages in the convo if it got too long
now it's hiding new ones.
brilliant.
just absolutely smashing work guys
I hate that there's still no news on fixing it and if they ever will fix it (just like scrolling to the end button in the mobile version is inexistent)
I'm pretty sure the timeline is something like this:
- Originally, once the conversation was long enough, it would start hiding old messages
- They release an update to try and fix it
- Leads to threads not being accessible anymore, lasts over the entire weekend because they pushed it on a friday for some reason
- They fix the threads being inaccessible, but now the original bug is flipped - now NEW messages are hidden, which is even worse!
I think the devs could probably do with being introduced to a crazy new concept called testing (and not in prod)
correct. My new thread got bamboozled too. Can't even use perplexity properly thanks to inexistent support and devs ❤️
and the "every model becomes 4o" bug is still not fixed
yeah, like
this is quite possibly the worst support/experience I've ever seen lmao
how do you mess up this bad multiple times in a row, and take weeks to fix it
this is what happens when you don't rely on your customers for your income but investors. however the bubble will burst eventually, and when it does, it'll be a dog-eating-dog world where everyone will fight for the customers. for now, they can afford to piss in our ear and tell us is raining
I, however will remember when that day comes and where my money will go
seek peace in the storms beyond your control focus instead on what you can shape within
You are wrong. Sonnet 3.5 is their MOST ADVANCED... BEST MODEL, MOST SMART EVER. Clearly your problems are user error
"We have testers, they are the end users"
Anyone have any decent prompts for a Pitch Deck review?
"adopt the persona of an experienced investor in this industry. Share all the feedback, questions, praise and critiques they would offer upon reviewing this deck. With each question and critique, provide an ideal prompt to get/create the information to address that."
How many o1 searches do we get a month?
Thanks @molten blaze 
will there be any black friday sales for perplexity?
Hii
hello pepoll
i think pplx might be having service issues again
I am
ohmo
why does the problem occur when i use it

Is Perplexity down? It had a log out issue and on all my browsers it logged out
yeah
the same thing happed around 9 and half hours ago
happened*
but got fixed
Who is admin here ?
Yeh it had an issue lol
I might've lost my thread rip
nah dont worry
oop
it will be back hopefully
same thing happened some time ago
perplexity down?
seems
seems like it
it is showing me as a freeloader
Same was real confused
I can't load my threads lol
the app is broken at the most crucial moment in my life
Its funny when things go down to realize how dependent you’ve become on them and how your usage of tech has changed… it used to be outlook/o365 or gmail would go down and no one could get any work done but now im like im set for work as long as pplx is gtg
Seems it's working now
Is the website down?
oh thank god i was ready to jump off a bridge
goated
Maybe not
so true
it is
That does exist?
Yes, we can choose the type like web,Academic. In the last version, we can choose the logic (o1-mini)
yeah reasoning is long gone
I don't think the o1-mini for me is better than Claude 3.5 sonnet
o1 mini yaps more I think, but you can just tell it to be more concise.
With complexity you can use it.
thank you so much, you're a life saver
Yw
Chathub is also here if u need opus.
out of curiousity, is opus better for any particular task?
be prepared to lose what you lean on for nothing outside your control is truly yours
Writing and such.
Then yes
Nobody stops you from getting a own workspace machine, and run local LLM, which web search LLM.
ahh okay interesting, thank you for the tip!
Your welcome, its not in perplexity anymore sadly.
Rip 🥲
So I got pro for free because of revolut, but for what can I use perplexity? what makes it different
deep dive ➤ #💬│general message
So, its a tool to get more accurate information more quickly with the help of AI/
So you can use it to get more references for example
## Top LLMs (Overall)
| Rank | Model | Score |
|---|---|---|
| 1 | ChatGPT-4o-latest (2024-09-03) | 1340 |
| 2 | o1-preview | 1334 |
| 3 | o1-mini | 1308 |
| 4 | Gemini-1.5-Pro-002 | 1301 |
| 5 | Grok-2-08-13 | 1290 |
| 6 | Yi-Lightning | 1287 |
| 7 | GPT-4o-2024-05-13 | 1285 |
| 8 | Claude 3.5 Sonnet (20241022) | 1283 |
| 9 | GLM-4-Plus | 1275 |
| 10 | GPT-4o-mini-2024-07-18 | 1273 |
Llama not even in the top 10s… Why are source-available/open-source models falling behind?
Yi-Lightning is open-source though? (Never heard of it before)
i had a lot of hopes for the nvidia lama one
o1 is still my favorite.
are you a big fan of o1 too
are you one of us
I love Perplexity.ai, I love everything about Perplexity.ai, the idea of the website and its features. But what I don't like about it is when they degrade the api's for eco mode reasons or rather to cut on costs which feels like every ai model on perplexity is like GPT-3.
Unrelated though, but I've noticed that LLMs are really bad at writing stories…
I'm not, I think o1 is mid
way too expensive for what it can provide
sure, it's good at SOME very niche applications like math
but it's not better at coding than any other model
this, especially algorithms
You fool, you are in the palyground paying $29,000 just to use o1!?
nah, I'm just looking at it from an objective standpoint
I don't need o1 for what I do
so I just don't use it
Which LLM is best at reading assembly code?
If you ask me ||o1||.
anyway, I think the concept behind o1 is interesting, but it needed like, an extra year to cook or something
Orion will fill your expectations.
what we got is very underbaked, 10+x more expensive than other models for very little gain
Sorry for the inconvenience luigi.
we'll see
I do not believe in Altman's hype ride
>o1's the best, still not (anything like it) on Perplexity
We love o1!
I don't trust the o1 in PPLX lab, that's just me.
reasoning models like o1 are more of a limited application
very good at SOME things, probably like, agent stuff in the future once it's more polished
but as is, not very useful
There's this one: https://huggingface.co/spaces/yuntian-deng/o1
I mean I'd say o1 is like, barely worth using
unless you specifically need it to help in one of the areas it's good at
Luigi, your opinion will be different only if you had more money.
Don't blame it on the LLM.
...nah, it really isn't, but you do you man
Nah, o1 helped me to solve some coding issues, no other model on Perplexity barely even solved it (full of mistakes and bugs).
At least we know Luigi is not a fan of o1, that makes o1 SOMETHING!
Fun fact, o1 have replaced CS tech jobs.
I want to try Perplexity API, But need some test Bear Token, I can not find the Token when i login the Perplexity.
where apply Perplexity API Bear Token?
Alguien puede decirme como puedo crear imagenes?
can someone PRO SEARCH for me on perplexity on how the process of getting deported works when visa is rejected in USA airport thanks
You can find a guide here https://discord.com/channels/1047197230748151888/1225557012532695060
anyone here code a search engine with similar architecture as this: http://infolab.stanford.edu/~backrub/google.html
Trying to build each component from scratch to dive deep into search
Do you want a particular model?
Hmm, opus please thanks.
Opus 😅
Or the default will work too the "gpt-4-turbo"
Appreciate you MR. mydpi
hello
remember, you can customize AI responses via your AI profile or Spaces instructions. you can "improve" how it responds 😉
Smol cat
New messages in threads are STILL hidden
can anyone tell me what perplexity is?
Everyone's got a problem with not keeping the messages?
deep dive ➤ #💬│general message
i just wanted a short answer but thank you 🙏
Yep, every thread that gets long enough has that issue
Does perplexity allow extensions like this?
@worldly trout Complexity is unsupported officially, but it works smoothly on desktop
Ah okay, so they don't ban people for modification like this?
Grok requires pro sub right?
Nop
Yes
Thats all models u get with PRO @drifting stone
Wrong url
😂
Perplexity should use groq lol
would be crazy fast
😎
Do we know if preplexity will enable us to load our own documents to search them via API anytime soon?
I sure love seeing perplexity's ads while my threads are still broken...
it's been nothing but tech issues for a month
Hello
Hello to everyone who's moping about long threads not working properly
Well but it changes the actual behavior of the site right? Well anyway that's good, just wanted to make sure.
Like for example how does complexity allow you to use o1-mini if the base perplexity doesn't?
It only works with perplexity pro
Oh so complexity doesn't add it, it's just part of the base application (with pro).
UI 😛
spaces is the GOAT
Does anyone know why my spaces doesn't follow my custom instructions?
Hey guys, I run Red Brick Labs - building AI tools for people. Looking to get some contractors to help with work. If you understand LLM's, front end and DB's - please reach out
The coding sync session is running now. Feel free to DM if you want to join. We work on independent projects, 30-40min sprints, 5-7 syncs and mentoring in between. 100% free, no ads. That helps me to stay focused on my own deliveries.
Did something happen to spaces? I had a few built out but I cannot see them any more from iPhone or safari
I hate that I keep wasting my pro uses on the spaces that don't listen to me. I want to research products with a certain flow and it just completely ignores my prompts
Anyone else noticed that the scrollbar is missing in the Mac app? Makes scrolling to the bottom of a long thread slow and inconvenient.😐
Hey everyone,
I got a Rabbit R1 which came with a free year of Perplexity Pro (valued at $200). It's about to expire in a few days, and I'm wondering:
Will the Perplexity Pro subscription automatically continue to be free after the trial period ends?
If not, will I have to start paying for it to keep using Perplexity on my R1?
Is there any way to extend the free period?
I've heard that the R1's Perplexity features will keep working without the Pro subscription, but I want to make sure I understand correctly. Any clarification would be super helpful!
Thanks!
Will the Perplexity Pro subscription automatically continue to be free after the trial period ends?
No
I've heard that the R1's Perplexity features will keep working without the Pro subscription, but I want to make sure I understand correctly. Any clarification would be super helpful!
I'd assume it would use free search then, so yes.
Free model, and limited pro searches.
R1 is trash, basicly just running android apps.
Should work.
have you guys fixed the mac os app?
still google sign in problem and it has a lot of features the webapp hasnt for mac, i like the one you sent me it works great but not integrated in the mac system
I've been trying to follow the #1140622008086970420 page but want to 100% clarify if anyone knows. This is my understanding: the "thread does not exist" bug is fixed, the "not displaying new follow-up responses in long threads" bug is not yet fixed. Is this correct?
bro 🥸
sam, we know you tuned this into the model lol
what even is that tone i didnt tell it to respond in this way or anything like that ahaha
"but hey, if any "technoking" moves start happening, Ill keep you posted!"
invalid
uh
i didnt get to copy it
is this against the rules
no right
yeah no lol
its making them profit
surely theyre fine with referals
dont see it
:/
oh god i dont even have $10
nvm
accept what you cannot change but be prepared to adapt your path
Bruh, I'm using perplexity pro and it gives me less than 10 sources each time, but felo ai and genspark give me around 20 and they're FREE
no way
20 sources and theyre free?
are you affiliated with them lol
ok this is a ruthless knock off of perplexity 💀
let me try it tho
You.com also gets alot more.
@simple portal But perplexity verify and has trusted sources. it does not take random.
There is also Phind.
definitely someone who started a company during the dot com bubble
oh wow lol i wonder if they rly needed that
ok yeah you.com is at least not exactly the same
People believe they want to get ride of the "GPT" later.
looks like its got some cool features
So incase they get more, and beter models, and they dont add "GPT" Anymore in them. Chat.com will be amazing use.
rebranding to "chat" is kind of weird though
since llms do more than that
Bruh, 3 pro search in each day reload...
i think its primary source of profit will be the api
Are you serious
are you mad that perplexity doesnt give u their premium service for free? or talking abt felo
Previously the Pro search was 5 that reload every 4 hours
bruh
For 20$ a month u get basicly unlimited Anthropic,OpenAI,XAI,Perplexity AI.
eh itds not the same as the chatgpt subscription
I think im gonna wait for the black friday discount 🫠
but yeah $20 a month is not much and these models are expesnive af
theyre losing money on them rn
It is
highly energy intensive process
Ofcourse.
If u look good u can get 1 year perplexity for under 30 bucks now
Who knows, Knows 😂
Amen, Im craving every black friday deals lol
He was talking about Perplexity free i believe.
Yes, thats what I thought
yeah prob not reasonable to expect more for free lets be honest people buy hamburgers for $20
and starbucks lattes
and this is a cutting edge new tech which is extremely energy intensive and complex & $20 gets u a ton of usage for an entire month
vs 2 frapachinos
Also u have pretty much unlimited usage.
Each 2 min +1 Usage back.
lol
300 was already alot, and now its 600.
im sure they're just building their userbase too in this stage and offering more for free initially while graduallly
restricting features behind a paywall
would be expected
They already gave alot away for free.
yeah thats also due to the rate of improvement in the models
but yeah they do
its because we're in a tech bubble lol
investors are pouring everything into these companies without them being profitable
because they know the winners will dominate the entire tech industry
whoever those winners are
& their best shot is to invest a lot in a lot of companies
see which one is the best.
Those are most known.
Phind has a instant model, and is the fastest i seen with search.
0 shot
ok
its pretty fast but the model seems a little weak
Yeah mostly low params, for so super fast model.
How biger how slower.
also 0 shot probably
no intermediary reasoning steps
uh
it just cited 1 source over 23 times
in an answer
I setup xSearch on iOS such that I can search multiple ai search engines SIMULTANEOUSLY
可以一次全部使用吗?
U can use all at once?
Anyone knows how many reasoning beta answers we get in a day? Answer quality is great way better than normal pro search but not sure how to pace it (seems like we only get 10 or something?)
O1 迷你版是 10,所以是道具 10。
O1 Mini is 10, so it's prop 10.
You can always use O1 Mini using complexity extention.
It refills +1 each 4 hours.
Awesome, thanks much!
Yep. I open up "spotlight" type my query with a shortcut phrase at the beginning or end of the query, and then it will open up uncovr first because that's the first engine, and then it opens up a separate tab for each successive engine in the list.
Unfortunately, not all websites are compatible with this, but I still feel like the fact that all of those tabs were opened after just one action is much more convenient than having to open up each tab individually
Perplexity just sent me a notification
We Might Be Wrong About Uranus
broken again?
yep
dead
yea
mmhmmm
had a few good hours, then went to “all threads deleted” and now whatever this is
just read the text wtf
I thought it was my internet, lol.
this is God’s sign to do my offline flashcards instead of browse on AI
sorry y’all, i brought this on us with my procrastination
nice
Great!
procrastination back on 😎 thanks jesus
perplexity is down. Reached max capacity
website has been revived.
144 sources
Sorry to bring up politics, the page is about jd Vance
Okay, so I counted 65
The 144 appears to be generated by an ai model and not just a number like in perplexity
BUT STILL
I wanna register my edu mail with my school provided, but I can't get a pro, why?
what do Perplexity Sonar Models mean? what is sonar model
solar slares and stuff
Are there any way to remove a source from the Android app or is it the web app only?
I was using a space for playing satisfactory and it's really only possible with uploading the photos but it will get so confused if you keep piling them so being able to remove sources's lifesaver
Pretty crazy
if they're not baking the sponsored sources into the answer then it's all good
Do paid users get ads?
yes they do 🚀
lol 😂
Why? Give it to free users
Paid with ads is some Amazon trash. At least the ads don’t seem very annoying, I don’t even look at the suggested follow up question, kinda forgot they existed
I just wish they were using it to cover the free users, and then provide ad free forever for paid users
with ads, maybe for once the follow ups section is actually useful
imagine you.com appears as a sponsored result when you're asking about perplexity on perplexity itself
😂
Haha, that’s would be great
I bet you this is stage 1 of ad rollout
Something small and less noticeable
💵😎💵
I hope they reconsider the ads for paid users, I think that ads will be an important part for them providing a free tier, but if ads start to degrade the experience than that is a no go
I keep changing the model and Spaces still doesn't follow my instructions
Try entering one line at a time
Doesn't seem to matter to much. I'm trying to have it format for product research but it ends up just throwing the "buy with pro" but without the formatting
do not let minor inconveniences distract you from your path @harsh stag
I don't think they should also give these to paid users.
This is the message of the day!
Powerfull mesage dude, in life you should never let inconvenienced distract you from your path
Sponsored questions?
And will premium subscribers have to see the ads?
do i need to care about how much big one thread gets with all the info, is there a limit for characters a thread can have?
Hi everyone, can anyone tell me is perplexity api is allowed to send attachments like images? because no where i found that such endpoints, but their AI said it supported the attachments.
Hey Aravind! 👋 I've been diving into some fascinating research on using fractals to boost AI performance, especially in language models. Check this out:
Fractals can be a game-changer for AI models by tapping into their self-similar patterns and long-range dependence. Here's the scoop:
• Self-Similarity: Helps AI grasp and mimic the repeating patterns in human language
• Long-Range Dependence: Fractal parameters can improve how LLMs predict downstream performance
• Fractal Attention: This could be the key to major breakthroughs in LLMs
I've got some ideas for how Perplexity AI could level up:
- Add fractal parameters like Hurst exponent and fractal dimension to the model
- Experiment with fractal attention mechanisms for better pattern recognition
- Team up with math and CS experts to push the boundaries of fractals in LLMs
What do you think? Could be pretty revolutionary for AI and language models, right? 🤔
If you want to dig deeper, check out these sources:
• https://www.toolify.ai/ai-news/unleashing-the-power-of-fractals-in-language-models-1774225
• https://www.geeksforgeeks.org/unveiling-the-potential-of-fractal-machine-learning/
• https://arxiv.org/abs/2401.04141
Let me know your thoughts! Maybe we could brainstorm some more ideas on how to make this happen. 🚀 #FractalsInAI #LanguageModels #AIInnovation
Citations:
[1] On The Potential of The Fractal Geometry and The CNNs Ability to ... https://arxiv.org/abs/2401.04141
[2] Unleashing the Power of Fractals in Language Models - Toolify.ai https://www.toolify.ai/ai-news/unleashing-the-power-of-fractals-in-language-models-1774225
[3] Unveiling the Potential of Fractal Machine Learning - GeeksforGeeks https://www.geeksforgeeks.org/unveiling-the-potential-of-fractal-machine-learning/
[4] Helping enterprises solve data-related problems with Generative AI https://fractal.ai/helping-enterprises-solve-data-related-problems-with-generative-ai/
A Computer Science portal for geeks. It contains well written, well thought and well explained computer science and programming articles, quizzes and practice/competitive programming/company interview Questions.
The fractal dimension provides a statistical index of object complexity by studying how the pattern changes with the measuring scale. Although useful in several classification tasks, the fractal dimension is under-explored in deep learning applications. In this work, we investigate the features that are learned by deep models and we study whethe...
Hi all, do we have a way to search for content created and publish into the Discover Section ? It's like there is no Search Engine to do it - ( i am in the Pro version) | thanks !
Stoicism i see
yes sir
Great i like it too
What is the best perplexity model available right now? It appears that gpt-o is currently the most effective option for me for some reason.
I’ve just discovered Perplexity and enjoyed it so much I’ve subscribed, but I’m disappointed to see on joining this Discord community that ads are going to be trialled. If ads start appearing on my app I’ll be discontinuing my subscription and going back to Claude and Chat.
So quite interesting, i have perplexity pro and i chose GPT-4o. I have asked if its chatgpt and it said it isnt - how does choosing model work then ?
Hi, guys. I know it has nothing to do with the group's topic, but would anyone know which search engine ChatGPT is currently using? Is it Bing? Which perplexity uses?
🤔
no matter whatever model I choose, I just get replies from gpt 4o now. Does anyone know how to fix it?
Hi
well u cant do anything about it tbh.
they will never fix this bug
Any math geniuses that can give insight on what the 2024 presidential popular vote margin of victory will be as a final result? Trump +2? +3? Even +1?
do you know if it gets resolved on its own? If yes, how long does it take if at all?
its random sometimes it go back to the original model, sometimes it still gpt 4o. what u can do is pray that the CEO of the perplexity care about his services
I don’t believe that the pplx api supports images or attachments, I believe it’s text only
Hello, guys. I have Perplexity AI Pro. How do I generate some images?
I am a pro subscriber for months now. For the last 6 hours, it doesnt matter which model I choose, its all GPT 4o for me which is highly unacceptable since I majorly pay to get access to Claude Sonnet 3.5
@signal hamlet can we expect a quick resolution to this as I was working on a project with extensive usage of Claude and its now in a limbo despite being a paid subscriber
Math? use o1-preview
You betting on it?
Can you share a thread? How do you know it's 4o instead of Sonnet?
is gpt-4o incapable of something you were using claude for?
ive had some models disabled at times as well as a pro subscriber
i think they do it from time to timed
ay
🚨 The AI Bubble Is About To Pop...
But not for the reason everyone thinks.
While everyone watches OpenAI vs Google, history is repeating:
Someone's quietly building the next AdWords.
(The parallels to 1999 are impossible to miss)
The implications are staggering: https://chrisbora.substack.com/p/the-ai-bubble-is-about-to-pop-heres
u can just what model are you it will just say im open ai from gpt 4
im using claude 3.5 sonnet and the answer is openai gpt4? yeah right
https://discord.com/channels/1047197230748151888/1225549032718602332
Need a better way to identify
hello, i am trying to ask answer but the answers getting trimmed out and not receiving full answer
bro its different, like all the answer literally gpt 4 instead claude 3.5, usually claude 3.5 sonnet have shorter respond but gpt 4o using long respond. its easy to tell its not from system prompt
Okay, one last test. Do a before and after and see if the "gpt4_limit" changes
https://www.perplexity.ai/p/api/v1/user/settings
there
Yeah you're right, that's super weird. Have you tried logging out, clearing temp files, cache, cookies, and then logging back in?
(Works from my end)
Sonnet version
GPT4o version
its still not working
its not just me alot of people got this bug too
Yeah, very strange. You could also try a different type of browser just in case also. But it's starting to sound like an A/B test gone wrong. I'd start a support ticket if you haven't already
If you do a follow-up question to my Sonnet thread above, it still reverts back to GPT4o?
i try following up for the sonnet version, it got gpt respond
even i try one of popular browser microsoft edge it still give me gpt response
So ads on perplexity? What does that look like??
They need to add ads as soon as possible, this is what will make them truly different then any other llm.
Check out #😎│cool-projects if you are interested in a project I am working on.
If I may ... 😉
This test you're doing with the page (https://www.perplexity.ai/rest/user/settings) is in no way conclusive. All models with a quota of 600 are grouped under the generic "gpt4_limit," whether it’s Claude 3.5 Sonnet, Claude 3.5 Haiku, Sonar Huge, etc. Only Claude Opus and o1 mini have their quotas counted separately ("opus_limit" and "o1_limit"). This has always been the case. Therefore, it is completely normal that when using a Claude Sonnet 3.5, a quota is deducted on the page (https://www.perplexity.ai/rest/user/settings) under the line "gpt4_limit."
Okay, you're right. The label is misleading. Then we're left with only analyzing the results and guessing the LLM. Which is what was done earlier
That thread is just not relevant
(as in, in this context it does not apply)
the models will known if they're made by anthropic or OpenAI
this has been proven several times in the thread about this issue
I’m in Europe, so I don’t yet have the ‘pleasure’ of seeing them 😅
Oh don't worry, I'm sure you'll be included soon enough 💀 😅
Clearly, no one will escape it. 😓
If PPLX results don't turn into the giant glob of ads that google search results provides, I'll be slightly happy
If it’s as planned in the follow-ups and clearly identified as such, it might be acceptable. I’m waiting to see it concretely 👀
Please note:
Before getting into the details, we want to highlight our guiding principle: the content of the answers you receive on Perplexity will not be influenced by advertisers. Users come to Perplexity for a more efficient, uncluttered, and unbiased search experience, and that isn’t changing.
Yeah, getting this issue too. Started happening last night. Been about 10 hours since then and still having issues with getting the correct model to respond.
i recommend report this bug on this channel https://discord.com/channels/1047197230748151888/1299441439620464640 so they dont ignore this problem
Yep, just did. Thanks! 👍🏻
Nope, I’m not paying for ads. Either ads or subscription, not both.
Another one has just started.
By popular demand, we’re expanding our campus strategist program to Canada 🇨🇦
If you’re interested in applying, or know someone who would be a good fit, please reach out!
Ads as well as paying $20 a month. No. I won't be renewing.
They plan on using ads even if you have a monthly subscription?
That is not going to go too well.
Yeah, honestly pretty crazy move
How to avail non-profit discount for Entrprise pro license
I'm sorry did they officially say Pro would have ads?
I'm trying to find any source whatsoever of this and can't find it.
Look in #📢│announcements
I get for free users. I mean for Pro. There’s no explicit mention of it.
which engine on perplexity gives best programming results?
you mean the model correct? I use perplexity to code a lot, and I love the Claude 3.5 Sonnet. But sometimes I use GPT-4o.
Unrelated though but fonts without anti-aliasing look so pretty…
What happened to o1 mini on perplexity
它已更改为 BETA 推理,这取决于提示,有时 PRO 会根据问题自动进行推理。
It has been changed to BETA reasoning, depending on the prompt, and sometimes PRO will automatically reason based on the question.
||您在使用实时翻译软件吗?||
是
Anyone knows if thinking claude can be used on perplexity? I tried creating a space and pasting the prompt but the prompt is way too long
Upload it as a file.
and each time in your prompt tell it to use the file in space for extra reasoning as example.
Or tell it in the system prompt.
I heard that they now limit Pro Searches to just 3 per day
I love Grok 2, less censored more fun and rewritten using GPT 4o to be logical
Folks, I'd like to change my registered email address on the platform. I can't find the option.
:(
if you wish to change your email address you should write an email to support@perplexity.ai
Quick question - Another user I know generated this result on his search of US elections .
This is what i generate. How is it possible to generate results in that former way ? I think if I'm able to get results in that format I would be using it more often because it's much better presentationally.
I like the detailed search function but sometimes when I have to see things live all this information becomes unnecessary
https://www.perplexity.ai/elections this is a special page made by the team for the election, now that it’s over they might not include it in perplexity results
Oh shi, so that was a bespoke result. I thought Perplexity was able to create those maps and tables using data from various sources.
hey 🖐️
Hi
Hey! Is there any reason why Claude Opus 3 was removed? It is an excellent model for deep reasoning and Haiku can't match it.
perplextiy can't generates images anymore?
Hey @heady mica!
Opus has been removed as Perplexity always stays up to date with the best LLMs. Currently the best models from Anthropic, Claude 3.5 Sonnet and Haiku are available.
Today's sessions has been started.
Do you know why they recently limited Pro Searches to just 3 per day for free users?
are the ads also for Pro users?
Hope not
Perplexity, web add Gemini 2.0?
Where?
I believe it’s the they are testing it as Gemini Experimental 1114
Gemini killing it
well its just opinion of the people. its just vote. like every llm come out will be the top of leaderboard in lmsys
We will see where it stands when the CI lowers
Yes, as often happens. But my first impressions after trying it are quite positive. It doesn't matter if it's first, third, or lower.
In the very least it would be 4th place
Would be surprising if it did out best old Gemini 1.5
So right above grok, or higher
The API will be released soon, and the 32,767 token limit will be increased quickly. (According to official information)
Quick question: I am using llama 70B via API and I've been using 2k tokens per request all day but dashboard says nothing spent. is there some bug in the dashboard atm?
Click "View dashboard"
Under api keys
you will see how much you used there.
I've been using perplex api for some time and it all been working fine but today all I get is empty screen. figured I should report it
It indeed does not update
+1
Nice
Hello, guys. My app isn't working
Hi, guys. I know it has nothing to do with the group's topic, but would anyone know which search engine ChatGPT is currently using? Is it Bing? Which perplexity uses?
🤔
Gemini has to be the worst AI available, utter garbage
How so?
In this ad announcement, this isn't going to touch paying users right? The minute I see an ad as a paying user, I'm gone.
Bye, if you are in the USA then you will see ads, even if you are a pro user
Crazy I know
Not in the US, so safe for now, but lmao that's wild. Paying users shouldn't see ads, period.
Lucky, haha. I’m fine too tho, I will just block the sponsors with an extension
But of the ads get more intrusive… perplexity will be trash
Hello, guys
Yo
Umm...I think I need some help.
What’s up?
I can't upload some document on my Perplexity.
Even if I bought the Pro, choose the docx.
But it still said cannot deal the document
I’m not sure if docx is supported, I will test If o can upload that type
I think I can give the screenshot.
I cant too in Spaces now
How do I upload a file into a space?
Are you too?!
Yes, i think its globaö issue
I even tried the pdf, txt, but they are the same
I'm very happy! I managed to change my email through a ticket opened in the support channel within the platform.
This error has been preventing me from uploading any files to Spaces for 6 hours now.
Well, it's seem not my problem...
Hmmm, I get the same problem, in a normal chat docx files work, but in spaces it fails
Bug report time
Yap
Yes, yes, same
I've been having this same problem since early today, I've also reported it.
:(
Wait, so it's still work until the early?
Why dose perplexity have such a hard time with files uploads
Always doing file upload related bug fixes
I live in Brazil, here it's 10:25 PM Thursday.
I live in China( Taiwan), now is 9:26 a.m. , friday.
Have you tried google Gemini via aistudio.google.com
It’s free and has 2M tokens of context
Also notebooklm is pretty nice
OMG! I like to joke with my colleagues who live abroad in different time zones that they're already in the future, while my future won't be until tomorrow. It's a bit funny.
It's so cool how the Internet can break down borders
I can't use it.It said I don't have the permission.
Oh, you don’t have pro permissions
You can leave the discord and rejoin with the links in perplexity setting
OK, give me a minute...
Anyone having this problem can expand on the issue in this bug report, so then they can take it seriously and fix it
Thanks
No problem
I know, it's really cool.
Huh, it looks successed.
Haha, I guess it’s fixed, for now…
Without any process
Yeah...
But I feel grateful for your help.
Of course 🙂
To prevent some people from China(mainland), I think it's much better.
Hm?
你用翻译器吗?
沒有,英文就這麼爛...
没多大意义 xD
怕被當台獨啊
"为了防止一些人从中国(大陆)来,我觉得这样更好"。
这就是你说的。
首先,為我的爛英文道歉
再來,就怕某些極左翼份子說我台獨
没关系,我不是中国人,但我能翻译大多数语言,写作也一样。我用英语写作,它就会被翻译。
不要担心别人怎么想。
如果您经常使用 discord,可以下载 vencord,并使用免费的谷歌翻译插件进行翻译。
您可以翻译选定的信息,并将您的文字从中文直接翻译成英文。
I know, but ...
I don't support the DPP.
I favor the KMT because at least we won't be threaten.
用中文说,因为我不知道你到底是什么意思。
Thanks for your advice, but I still insist in using my bad English.
我又不支持民進黨
我比較喜歡國民黨,最起碼他們當家時台灣不會被威脅
不過蠻意外的是這裡不像網路梗那樣問我中國的一些問題...在我印象中他們很容易搞混...
事實上,這是一段複雜的歷史問題
反正現在大陸認為民進黨是支持台灣獨立的,所以他們用各種軍事演習和經濟制裁等方式“威脅”他們
所以,你希望中国和台湾再次加强联系?
但相對而言,國民黨希望能夠找到一個平衡點能處理好台灣和大陸跟台灣和美國的關係
我明白了
在我看來,所謂的統一是歷史必然的
但是要經過很長一段時間進行磨合
絕對不是通過幾張紙或是戰爭就能達到的
所以我說了,這是要經過很長一段時間的磨合
這又設計另一個複雜的層面,因為雙方在政府機構和政治體系上都有很大不同
沒處理好就會像當初東德和西德那樣
當然不用
這又很複雜了...
反正對我來說,最理想的統一方式是雙方在很多方面都要做出巨大的讓步,但這很難
我會中文,只是英文不好
我不會打簡體中文,但我看得懂
現在我使用傳統方式,這樣比較好嗎?
倒是我上bilibili 時有些人看到我打字很痛苦(因為是繁體字
您能更清楚地理解這一點嗎?
因為簡體是從繁體進行簡單化後變成的
我當然明白 xD
所以一個字當中很多部分被省略,所以有些人看不懂
是的,就是這個原因。
因為在書寫的時候更簡單也更好學
新一代都是簡化的中國人,對嗎?
呃…其實用你想用的方式就好了
因為我都看得懂
我想問你的母語是什麼?
荷蘭人?
其實用英文就好
比利時和荷蘭的語言。
我看得懂一點,剩的用軟體翻譯就好
我是說你來自哪裡
我不介意一直翻譯。這樣我們就能互相理解了。
因為軟體翻譯還是有問題
就像這句我看不懂
比利時人,ok...
蠻酷的
呃…抱歉我用中文的聊天方式
這一句話其實是我在問你來自哪裡
是的,我來自比利時歐洲。
有點羨慕
因為我都不會用這些外掛
就像我說的,您需要 vencord,它是修改過的版本。
它仍然是 discord,只是經過修改,讓您可以使用外掛程式。
Hmm...有趣
也許我應該找時間和設備測試
Vencord 完全免費,外掛程式也有 google 翻譯功能。
我使用 DeepL pro,以獲得更好的中文翻譯。
但 google 翻譯也做得不錯。
看到一堆中文以為進錯板
😂
real China hours
What is happening lmao
毛主席万寿无疆
看到一堆中文以為進錯板
華語萬歲
Does Perplexity do a real plagiarism check before writing an article? When we tell AI to write a plagiarism free article, it is never fully plagiarism free. (Most articles are usable though)
You can always produce the prompt in which it gives the article outline and specify that key in the prompting as well
Making it generate original, key details and not to include anything that may come across plagiarized
That helps
But you really shouldn't have to worry about that... But there is key prompting guides. You can even direct the ai to be more aware of plagiarism. So even if the AI is aware you do have to state that in your prompting...
But yes it does it's and will even include information near the bottom of it's search to indicate where the resources can be located
👋 Dear community members! Am I correct in understanding that if I purchase the 'Professional' subscription for $20, I will be able to send API requests specifying the LLM model GPT4omni mini in the request instead of 'llama-3.1-sonar-small-128k-online'?

