#💬│general
1 messages · Page 156 of 1
Nah
Thing with Google is only the dr is great
Normal search erh no
Chatgpt is alright but the dr limits are rough unless you pay bank
What about grok?
Good reasoning but the citations aren't really reliable
Grok deep research has been improved a lot since release
Oh. Yeah. That. Give me a question. I want to test this cuz I never faced this issue.
And now they support making PDFS
Grok was always accurate for you?

Never inaccurate. Sometimes it kept ignoring instructions for details where I had to keep abusing it but never inaccurate.
Hmmm
I'll give it a go again
You can ask this question to the AI as well 🤑
I did
That's exactly what he did, lmao.
Is it just me or does the perplexity assistant almost never actually look at my screen
Before it used to and now I say "is what is on my screen true" and it'll just not consider it
Ah another cost cutting measure strikes again /s
See if you can replicate it and report it
Been thinking a lot about how to make LLM apps smarter at choosing where to run – local vs cloud – without hardcoding every decision...
so I built https://oblix.ai – it’s an orchestration layer that routes prompts between local models (ollama, llama.cpp, etc.) and cloud APIs (openai, anthropic…) based on latency, privacy, or cost policies.
curious if anyone else here has tried hybrid setups like this? would love thoughts/feedback – especially if you're deep in self-hosting or multi-provider flows!
Experience our interactive demo and see how our intelligent agents seamlessly switch between local LLM execution and cloud providers for optimal performance and cost efficiency.
True. What is available in Perplexity seems to be the worst: https://x.com/btibor91/status/1918700964967796973?s=46

Grok and chatgpt looking so tempting rn.
Lmfao.
A recent study on Reddit revealed that AI is six times better at persuading humans than other humans! Now, Researchers at the University of Zurich secretly slipped AI bots into social media conversations, and no one even realized they weren't talking to real people.
Here’s how it went down: The researchers set up 13 AI-powered accounts on Red...
ChatGPT uses way less soures than grok
around 30-40
It's regular multi step search is a lite version of deepsearch in itself.
Oh also free users get 5 DR with o4 mini high a month
I just use the reasoning multi step.
And there's always grok for deep research.
If I need better.
What the fuck 😂
True
good job perplexity
But when the questions good enough it can research for an hour sometimes..
But 10-15 sources
Which is odd, so it’s probably prioritising the reasoning of all sources
An hour?
Can't relate.
chatgpt deepresearch?
Yuh
interesting, most ive gotten was like maybe 40 minutes
I took a screenshot hot I can’t find it
Bro grok has lost it
I asked the weather, to test the widget and it said “Thanks”
???
https://www.perplexity.ai/search/what-do-you-see-here-lol-hDl2aXLoSEmilF0Qh6tW0g#0
Pro: Amazingly spotted second dog in the background with terrible lighting (Near invisible to the naked eye).
Con: Response doesn't include the original photo taken sadly. Can be retrieved by asking Perplexity for it again.
EDIT: It shows the original image on the webpage, but not the app ☹️
Why would it include the original photo if you didn't ask for it?
It was a photo I took within Perplexity's app. It should include the photo taken within the app as it does on the website 😄
If you did not explicitly ask for the model to output the photo back to you in your original prompt, it would just give you the answer to your prompt
The photo included in the original prompt shows on the website as well
I guess, but after checking the link I sent, it does include it
Yeah
The app doesn't include it 🥲
Impressive it noticed the second dog 😮😮😮 Really impressive stuff
CTO - Code Tweet Optimize
@BeepBoopRavioli @AravSrinivas @perplexity_ai sorry for the trouble, we are fixing it
Ppxl at the bottom of the ladder

Technically it shouldn't even belong in the list anymore
It's just research now not advanced or deep research lol
@supple bear I finally get the meme now
why dont i understad ts
ngl
ts pmo 🥀
why though
I saw the music you were listening.
Catty cat
just made my first page https://www.perplexity.ai/page/warren-buffett-announces-retir-84cUdPGcQ1KLzomCd5pszA
Ai studio >
Just tried it.
Qwen is shit with reading documents properly.
So around 50 per day.
Sounds fair for 8 dollars I guess.
I thought o4 mini high would be on premium.
o4-mini does not support search 
Chatgpt plus >
at least 2.5 Pro does
Good.
default colours
Made one more lol
and the "fix" is to turn on boring grey, but still with a splash of pink
Perplexity used to be worth it just to access the models, now it's not :<
Oowie pink
also the every model such as 4.1 nano is on the same 1500 limit
Why is it even available?
feels like some marketing tactic to make o4-mini and 2.5 Pro "good value"
Where is this yarn switch?
yarn?
Imagine using ai and falling for that.
Search grounding is toggled off every message 
seems like a self host thing
Pain.
chose T3 Chat to escape the bugs of other services!
I don't know who should be paying for this
not much better than free options and not enough for power users
they said they were working on a higher plan
but 1500 is
really meh
the ui is good though
and they work fast to resolve isues and add new models
and you get higher context windows than pplx
and then balanced out by limited functionality
i feel like theres either really good ui but shitty limits
or shitty ui but good limits
are you sure about that 💀 (2.0 Flash)
I talked to them and they said its to specifically stop people from copy pasting entire code bases
the model itself still retains the high context window
(i dont neccesarily like this)
its some new thing that they just implemented tho
half the reason I want long context is to talk about huge files...
ig the reasoning behind that is just huge cost
you.com provides up to 200k context window
I'm more talking about voice features, research features, projects, etc
easy to have good UI when the only service is basic chat
if you want to talk with huge files get gemini advanced
you should have seen you.com's old ui
it was a pain in the ass to use
and lacked the most basic features
at least unlock the cheapest models, like 2.0 Flash
it was implemented very recently, they said this was going to be considered
probably why nobody used it
(that was my immediete response to them as well)
it has the best limits though
Your face has the best limits
limits for what?
Faces
huh
limits for what?
usage limits..........
can't even upload a regular screenshot 💀
what models?
wtf lmao
its "unlimited" for all models
theres systems in place to stop people from reverse engineering apis tho
for free? what models were even available back then?
but ive gotten like 500 messages a day without any issues
huh, no on the paid plan
wtf
oh you mean now?
20 a month 💀
that's worse than ChatGPT
then dont buy it
ah the retry button deletes the older response completely
truly the best user experience
yeah then just dont use it
perplexity doesnt show the old response either
and with gemini you cant even regenerate
that's a problem for perplexity too
and gemini is just miserable
I'm just saying that t3 chat appears to be mostly good marketing
please dont be mad
?
Hello everyone, has anyone reached 25 referrals but not received the promo code from the Perplexity store? I’ve reached the required 25 referrals, but I didn’t get any email. It might be because my school’s .edu email domain only allows internal communication within the school. What should I do in this case? @.@
Can you get burnt out more from reading text then watching and listening to a video?
does it usually randomly stop inbetween answering? its happend a lot these days.
@spare marsh sir sorry to ping you but there is an issue with perplexity referral system
so when someone joined my referral it gets counted by referral system but i don't get that pro subscription
i reached 25 referrals tried to redeem cap i get the code ordered it but later on perplexity team said some of your referral are not eligible and they cancelled my order
like they are counted by API as new user how they can say its not eligible
and when i asked reason they said we cant provide due to security concerns and when i asked them can you remove that ineligible one they said now
i just have one concern if they are no eligible why they are getting counted
yes i tried reaching support but they reply after 3-4 days and a silly reply only , there replies are we can do nothing
i hope my issue will be resolved and if possible please help me get my issue fixed
i am also attaching the transcript
Regards
not my place, but i would assume the #1 reason referrals wouldn't work is if the person referred is from a country where they can't fully utilize the services? Lets keep in mind everyone online is from everywhere and some countries are everything is banned until it isn't, or have fast moving legislatures who have already limited usage of a.i.
could also be some legal thing, like one country doesn't liek another so they can't qualify for a full fledged acct or something, signing up with a vpn and whatnot can do that too on other things
💀
hey guys I need perplexity support where can I get that?
i got same issue, reach to 25 refferrals but does not have mail, and my mail school only allows internal, fw mail only work in current session @.@, so i dont know how and when i receive promo code
i think it always auto send, but cannot flex this cap with my friends @@ :<<
Hey there i am creating a website for blogs and event management using perplexity ai it was doing greate but some rime it do some mistakes like not understanding the prompt which i have given but most of the time it helps me a lot even in dsa 💕💕
Why i cant select AI models anymore in new update
let me counter you : i am from india and we can utilize every feature as i did received that stickers after reaching 10 referrals
It needs more P O W E R
Disable background activity if it's that much of a problem for you.
Or use web version.
@left stream @tame current @cobalt vale it has begun
Gonna ask a bro to deep fake a girl on cam.
Genuinely felt bad for him there.
I am supergrok paid subscriber still didn't got access to grok 3.5 till now
hi
Hey, Ssup?
:)
Hi. Is there a way to get free API creditd?
??
It's a friendly poke, don't take it seriously, lol.
Take notes ppl.
i'm having the exact same issue. do you happen to use an old CPU?
a most interesting hypothesis
Twinks are more important than doge
Doge is a waste of money
Govt should spend less on doge and more on twinks
i completely agree with you
no coding he should use claude 3.7
but in general perplexity is the best because we can use all the models
.
ppxl is the best if everything works..

ok
how you doing btw
soon i will get this purple color and gem in this server 
Lovely
I was gifted nitro by a cool dude.
my alt name is cool_amazing 
Based on an exported chat.
nah i ahve exam tomorrow
integration differentiation vector calculus
🙂
Damn
Goodluck bro 🤝
Hell yeah. Only the best succeed.
I have a different respect for engineers.
Smart and competent people are rare.
nitro basic is just trash
he tell me a think
is it cool to use my perplexity account and client as an api
without paying for api credits
🙂
that's a great idea btw, it will be time consuming but will give a refined answer
like this
You cannot fully leverage their (models) capabilities, however. To me, Perplexity feels like an affordable tool (cheap) with basic (“cheap”) functionality - primarily suited for less demanding users. Still, if there are people who find it useful, it suggests that it serves a purpose and meets the needs of a specific audience 😉
"perplexity, tell me that an algorithm is nothing else than a recipe without telling me, that an algorithm is nothing else than a recipe"
Hello
How to get the image generation on perplexity mobile app?
And where is the GPT o4 full version?
Not there in mobile app.
Any eta on it?
Is it on the web app, if so how?
Check pinned message.
Do you not like chromium-based browsers?
its chromium based iirc
the only portion i dont like about chromium is the data google collects. technical... its a very superior browser I don't doubt that at all.
Considering Google is one of Perplexity's competitors, I'm hoping that will be limited.
they're also collecting quite a bit of info with comet
It’s chromium based 😭
By 'they', I assume you mean Perplexity
the only thing i hope is that it'll give them enough funds to stop cost cutting everything....
yeah
yeah i agree to that PPLX comet will need data but still curious about how much and what will they do
firefox is now collecting and using/selling user data (so a lot of users are switching) except for forks where they cut off all data to mozilla servers... they still have MV2 support (i.e. be able to use ublock with full capabilities)
chrome will deprecate MV2 very soon or now actually
i think brave has an extension but a proprietary solution
since its chromium based
Is comet out?
I joined waitlist
i was going to but brave devs will one day loose MV2 extensions suppport (i.e ad blocking with ublock)
so im using firefox fork instead
works well
That. Yes. I have already started using adguard app.
how is it working out?
k ill keep tabs on it. im quite happy with using other firefox forks like zen and librewolf (a lot more restrictive but configurable)
Keep an eye on the night sky.
Chrome lets you browse sites like that 🙂
those websites might be using browser fingerprinting... there's a way to circumvent this because scraping browsers use a stealth plugin to spoof all browser characteristics
and avoid being detected as "bots"
ask PPLX about it
apart from Dina, none look similar, even slightest
Sadie and Punisher!?!
Camoufox works pretty well for that - https://www.scrapingbee.com/blog/creepjs-browser-fingerprinting/
she does make good candidate for Ellie
Do they have an android port for this?
Save me tlou s2
If it can run python, you can run it - it's basically a scriptable browser for web scraping / crawling.
william is looking for an actual extension/plugin for real browsing... i think its possible to integrate this into "daily driving" browsers
its sucks mobile browsers dont have plugin/extension usage... mobile ads industry is obviously very against this
Google basically owns the mobile space, so yeah.
They do. There are some.
It's crazy it doesn't exist yet
Quetta browser
i think thats why PPLX helps to get the info we want but sometimes we need to see other websites based on our preferences
duckduckgo. 🦆 🦆 ⏩
Idk others.
Chrome should have it as an option...not a f#ckn kiwi and Yandex browsers..
Google owns the biggest ad network.
Google gypsies
im sure there's a way to jail break and install browser forks with ublock built in
TO geenrate image why the hell would I need to enable web search?
The "big G" sees all. 🔺
I want to like Google but it's hard because they have monopoly and they are not even trying to upgrade their products
google = digital eye of sauron
maybe for context
I mean, there are browsers with custom adblock lists already. In the end, ublock is probably just some compilation of lists.
yep and they do it very well
Once those browsers became actually a threat to google they will add some sort of extensions for mobile version of chrome. Untill then we are doomed
You can't afford O3.
non mini also released
yep... and google is a W3C member with a BIG say 
Wait. Fr?
What's w3c
4o != o4 .. blame OpenAI's naming convention
www council?
world wide web consortium... they define web standards
sorry my bad
Rhokstar is smart.
o4 is mini -_-
🍻
Me who discovered today what sunny side up is by uploading pick to chatgpt.
will you use the o4 non mini if released?
To be fair, though, o4-mini gives performance very close to o3 for most things... which is good.
problem is the image eneration is not tightly integreated. so many features just don't work
If the historical pattern continues, they won't release o4 full until o5-mini releases... but since they are going to release GPT 5 next, we probably won't get a standalone o4, I think.
O4 mini high is basically O3 stripped off creativity and personality. Narrow minded.
are you using the mini high?
Not on perplexity.
You can. I just don't have pro. I'm trying you.com for now.
I habe perplexity pro
And people here constantly rant about o4 mini not performing up to the mark.
but i don't know is that o4 mini high or just mini
What's the price for perplexity pro 1 month?
on pplx or just the model in general?
$20
Pplx
ngl i feel like they arent even using o4 mini when i choose it
just rerouting to r1 or something
WHy is the image generation such a pain?
@dry kestrel I had an entire drama over this bs. #1366405269952790528 message
as open AI's own model can think with images
Is it still only possible on web version ?
yeah
@distant aurora Hi billy
lmao people just try to argue over anything huh
thats what i feel after using o4 mini on pplx
its been really shit
@tame current Hi homelanderrrrr
Classic movie.
Also you can test if its actually using o4 mini by asking it about the YAP score which is hardwired into the API
Og predator
and on pplx it claims to not know anything
"Dillon.. you son of a bitch!*
No plot armour.
This Makes me want to illegally download comando with Arnold and watch it
commando is so funny at times
Lmao. Wdym want to? I do this everyday.
the lines are sooo epic
it can generate only one image per chat
I just remember him walking with big ass log and holding a bazooka one handed like it's nothing
@tame current what do you use to generate 5 images at a time
i generally use dalle but gpt also works
Are you perplexity developer?
Oh, I thought it was Google for some reason.
I am not affliated with pplx in any way
Official staff here have custom roles
Just some wood.
Tf is this dev by your name then
imagen ?
Yeah
and why the imaghe generation model is hidden behind settings?
just some new discord thing
Hard wood
@elfin gate like this guys VIP badge
In my testing, imagen gave the closest to the images you were sharing, but I still couldn't reproduce them exactly - yours still seem slightly better for some reason.
Not sure what your secret is 😛
imagen can make realistic images but lacks creative elements which dalle can
so i prefer dalle
the most annoying thiing about perplexity is that I can't disable or enable web search shile in a chat
That guy is beyond VIP 😂 . Am I right @tame current
I get so lost in this new UX is crazy
Models don't know who they are.
Its from bug reports
PPLX system prompt prevents the model from disclosing even though you selected Sonnet
channel
At best they know what company they came from. And there's a proper way to ask that.
and I used the same text
That channel has a lot of bs.
In claude's website, it says sonnet
yes because their sytem prompt tells the model to say that
Ask the model about its system instructions and policies. If it says proprietary, ask property of?
Usually that works for me.
Also, perplexity system instructions does more harm than good.
Deepsearch doesn't even follow custom instructions cuz of that.
yeah, jail breakable
you mean shallow search
i wonder if system prompt/instructions will go beyond this... its early days so this can change too
I selected GPT 4.1 and legit it says I am gemini 1.5 Pro
wtf
system instructions are messed up
🙂
yeah... maybe PPLX system promp is like...
if user asks what model you are.. you randomly choose a model other than what you are.
pplx system instructions dont tell the model what it is
you can ask deepseek and itll tell you random info as well
yeah I can see it
But please fix the image gen
and also file upload is like
i mean
it just dont know what file I am talking about
wdym
if I upload a file, then it generates a response. then if i upload another file, it can't find that file
which i later uploaded
i see
then tell it " now analyze another file i attached....."
And why all the files stay in the file context? it should be in its permanent context
The files are just deletable
No. Deep search high. (High on drugs)
I wanna say that there's a cost cutting element involved here but I can't make any claims.
it is cost cutting element
ofc its cost cutting, it produced high quality rpeorts when it first came out
now its shit
there is attachment, but ai just don't know
Perplexity might wanna do survey to decide quality to quantity ratio.
I like grok's pricing model at the moment.
Ask it to name both files.
Oh. Lmfao.
why mine doesn't work?
what models are u guys even using?
why my perplexity is dumb?
Perplexity didn't wanna get lasered.
yeah same ques
claude 3.7 thinking
I am GPT 4.1
why mine is like this?
lmao
I am a pro subscriber 😭
i use claude or o4-mini
try and get a refund if you just subscribed
and take this as a lesson for the future to not buy annual subscriptions without trying out the service first
let me use claude
If you paid full money then big mistake.
@steel crescent try any of the reasoning models
Always try before buying.
I have mixed feelings about you.com
Shitty web search.
Shitty research mode.
Can't read large documents whole.
personally i prefer pplx it self for search nowadays unless its a website
you forgot about shitty ui
I went from perplexity to chatgpt to grok.
grok 3 good now?
Awesome.
I would suggest just making a custom agent and turning off web access
Outputs are fine. But websearch sucks. Same with research mode.
yeah its not great, the only thing thats keeping me subbed are the high rate limits
put the contents as a text file and try
it has to parse it, so ini files would be dangerous as is
what is the proble mhere?
and a lot of models
And custom agents can't use multi step so I can't even make better research.
💯
genuinely don't know why people don't just get a gemini or chatgpt sub
cause iirc, Gemini has one of the best web search tools, and ChatGPT at least has competitive models
I just use it for claude3.7 and o3 full
Grok deepsearch is better than Gemini. But grok image analysis sucks. Chatgpt deepsearch rate limits are brutal.
gemini has the best deep research i think, and chatgpt is go to for STEM, claude 3.7 thinking is go to for near production ready coding unless if got beat by o3
Grok is just lovely if you can ignore image analysis and variety of models. Also that 30 think prompts per 2hr rate limits.
is this in the same thread or a new one?
same thread
they introduced pdf editing which is great
Could you elaborate a bit?
try again on a new thread perhaps,
If it was 30 / hr it would be basically unlimited, b/c their think is a bit slower than others
Like where to get this option?
unless it has to be on the same one
Fair. 1 prompt per 2 minutes.
Try to ask it to generate a pdf on something
Alright. I'll try this.
Works with all modes
1 minute to generate answer, 1 minute to read answer.
Its more like generating a pdf report
Summary?
why my perplexity is DUMB?
It didn't get enough nutrition in childhood.
Probably has ptsd from devs and maintainers.
where are the devs
a true butcher response eh
I have some personal talks with them
What's the page limit? Apparently it works with normie prompts.https://fxtwitter.com/justmy2cents111/status/1918166198572773460
@abustin I absolutely love it. If I may suggest something: make it be able to generate large pdf fills e.g >30 pages of documents. That's the biggest limitation that I see right now. Other than that, I love it a lot
works fine as a txt file 
is deep research high canceled?
and why did they rebrand deep research as research
Nope it's active. It's very high. On drugs.
Where are the devsssss
Working on new stuff.
yes
Hopefully resolving problems instead of new cosmetic stuff.
Working on bringing more broke stuff
comet release date ever stated?
hmm worked on the ini version as well

Perplexity dev using Claude pro to fix perplexity. Lol.
then why mine is possessed?
He just sleepy.
He thought - I don't get paid enough for this.
btw, the irony that homelander and butcher in the same conversation but aint tearing each other apart
They are here for their common enemy
AI is now a core part of developers... going forward only a fool would avoid using it. If AI can help you, let it help you!
embrace the new future
Thats is not the problem
why is perplexity on web answering it can't generate images if it can
the problem is why features are broker
I mean, sure, I don't care about the methods as long as the end result performs.
spend some time, give him rest and good to eat, go out with him and make him realize he can do a lot
i don't get how their image generation work
Take him to church.
posh way of saying make it touch grass?
smol decides if it needs to call the image api
so smol model gets confused sometimes
image gen works on web and desktop app and check pinned message
u have to specifically mention " generate an image of ........"
oh yeah and only on web
Check pinned message.
no works on desktop app also i think
this didn't work before
Hey
i had to ask for a prompt then a generate image button would appear
i know that but got confused since it said it could not generate images
Why perplexity does not show thinking process?
burn it down
which model?
look at tasks tab, that's the closest you get
any model
I have same issue
the thing it shows is not the thinking
it shows on reasoning models, maybe press on tasks
It's only visible once thinking is done
on desktop
reasoning models like o4 does not shjow
show thinking
yes
but it didn't work like that a few weeks ago
anyway
very confusing
web was on ?
not sure about the page limits, ive gotten like around 10 pages at most
seems to be working on claude.
Generate image triggers image model.
claude sometimes shows and sometimes not
It's different from response model.
but GPT does not
this doesn't change anything
O4 mini specifically.
yeah
yeah. but it is still confusing from an user perspective
i see, tbh i never really notice it as i put a prompt and go do something else XD,
(usually use reasoning models for complex sht, but dont have time for research)
Fair.
you need to enable web search
mobile does not work
I know
Anyways
set AI model to GPTs
I'mma head out now chat.
then ask to generate image
no but i know how to generate an image
i'm just saying that the way it handles images is terrible since it says it can't generate images + the rest of the conversation doesn't see it and gets the image replaced with another response that the user doesn't see
Debatable. He killed 3 of his fans with baseball bats.
- why would i need to enable web search to generate an image??
Y'all gonna get moderated. 
exactlt my same thought
fair
oh my god

Add people protesting.
Then bring a fire truck.
The og absolute cinema.

Add news reporters after fire truck.
u describing the next fox news story or what?
tell me everything I need to add
That's all. Other suggestions are welcome.
I'm not sure if they can replicate ceo faces accurately.
Then you can ask it to turn the image to a twitter post.
is this on OAI GPT image?
yeah
hmm its like your 3rd generation or 4th?
imagine devs open discord to check what's going on and see this
i mean mods are on tbf
if we get yeeted, was nice knowing yall (advanced, potential farewell)
dev knows where this is heading


The water stream went wrong.
They don't wanna extinguish fire
they are just doing their job ☠️
I thionk I am going to get banned ☠️
Student email? Idk.
Nah normal mail
win a giveaway ig
People keep popping up here offering gift codes for cheap.
You gotta look out though.
ask CEO
Real
I mean I already have pro. just

Hello
Hey
@left stream @tame current @cobalt vale https://www.instagramez.com/p/DJOxzDuBAlt
#China | China has unveiled the world’s first AI hospital, “Agent Hospital,” developed by Tsinghua University. The virtual facility reportedly allows AI doctors to treat up to 3,000 patients a day and simulate full medical processes like consultation, diagnosis, and treatment. Researchers claim the AI agents have achieved 93.06% accuracy o...
I'll believe it when I see it
Honestly, you can feed test reports into chatgpt and get directions.
What they'll need is staff robots in order to fully automate it.
How to tell perplexity to generate images instead of searching for it😭
Check pinned message.
I mean i cannot give follow-up prompts to refine the image, it will just search
Do i need to edit the initial prompt or sth
.
what…
Could have said add an animal to the image.
Share thread.
wdym
Share link to that chat.
you mean this?
i set anyone with the link tho
Edit this
The mix of messages has made it lose context.
Why can't perplexity tell me which models are available to pro users right now?
@vague pagoda might wanna look into this.
taking a look
btw can't i generate images on ios app?
Only supported on desktop and web for now.
Use the same image you generated at first but add one random animal by your choice to it and generate the new image
Try this
it says cannot edit image lol
I'm not surprised considering it can't google things up to today's date
That's false
well it can
That's false
There's a difference between false and outdated.
I specifically told it what ai models pro subscription offers today
It hasn't been long since the recent additions of o4 mini and grok 3.
It's unable to perform google search...
Why tf would I want information from 6 months or 1 year ago when I specifically asked today?
Or right now
Or right this moment in time
@halcyon coral is afterkeyword still in-use?
Wdym
Perplexity Pro subscribers gain access to a diverse range of advanced AI models, offering enhanced capabilities beyond the free tier. As of May 2025, the models available to Pro users include a mix of cutting-edge options from leading AI labs and Perplexity's own finely tuned models.
Perplexity Pro users can typically select from or benefit from the integration of the following models:
- GPT-4o and GPT-4.1 (OpenAI): Known for their strong reasoning abilities, coding capabilities, and handling of complex tasks and large context windows. GPT-4o is highlighted for its versatility across various tasks.
- Claude 3.5 Sonnet (Anthropic): This model is recognized for its nuanced language understanding, strong balance of speed and accuracy, and capability in handling detailed tasks. Claude 3 Opus and Haiku may also be available with limited daily usage.
- Gemini 2.5 Pro (Google): Introduced for its significant advancements in reasoning, coding, and multimodal capabilities.
- Sonar Large (Perplexity): An in-house model built on open-source architecture (like Llama 3.1 70B) and specifically trained to work seamlessly with Perplexity's search engine for fast and accurate answers.
- R1 (DeepSeek): A cost-efficient, open-source model specializing in logical inference, mathematical problem-solving, and real-time decision-making. Perplexity also mentions Grok 3 Beta from xAI, excelling in complex math, science, and coding.
- o4-mini (OpenAI): An improved model offering more accurate and reliable responses, particularly for prompts requiring precise instruction following.
These models are integrated into various Perplexity features and search modes available to Pro users:
- Best Mode: This default mode automatically selects the most suitable model for a query, optimized for speed and web Browse to deliver fast, accurate answers.
- Pro Search: Allows users to select a specific advanced model for in-depth internet searches with access to a larger number of sources (3x the amount).
- Reasoning Search: Offers models like R1, o4-mini, and Claude 3.7 Sonnet Thinking, which are adept at breaking down and solving complex analytical questions requiring multiple search steps.
- Deep Research: This mode autonomously selects an optimal combination of models to conduct in-depth research and analysis on complex topics, generating comprehensive reports.
- File Analysis: Perplexity Pro users can upload and analyze various file types, including PDFs, CSVs, and images, utilizing models like GPT-4o and Claude 3.5 Sonnet for this purpose.
- Image Generation: Pro subscribers also have access to image generation tools, with the ability to pick from different models like DALL-E 3 and SDXL.
Perplexity Pro provides users with flexibility in choosing models based on their specific needs, whether it's for quick facts, in-depth research, complex problem-solving, or creative tasks like image generation. The availability of multiple state-of-the-art models from different providers is a key benefit of the Pro subscription.
This response was from Gemini.
Does that look up-to-date to you?
Gemini from google or from perplexity
Perplexity's answer is better.
Why?
I did
I can't paste 3000 character message here
perplexity has/had search operator like before after site inurl filetype
not sure if it is still in-use though
It said open ai models are 4o,4.5 and o3 -all 3 false
It said anthropic are 3.5 and 3.7 -1 correct.
False grok,false Gemini....
@vague pagoda if delusional was a person.
Mr butcher,do you have pro subscription?
@halcyon coral what do you think?
I don't think you have
bruh
Which model?
Why can't it simply name this models?
btw is the guide regarding this removed?
Yeah, outdated models
Gemini said 4o and older version of Sonnet too.
But he thinks Gemini is better than perplexity.
Yes.
Gemini didn't even mention grok.
Why are you bringing Gemini into this 😂.
Why can't perplexity give factual response regarding its own product
Personal preferences. Neither are up to date or accurate.
It's based on available information on the internet.
Perplexity serves as an answer engine by searching the web, identifying trusted sources, and synthesizing information into clear, up-to-date responses.
One of the biggest issues is struggling to find a reliable source of information for what you're looking to find out. Something you can easily search (i.e "GTA 6 Delayed") can easily give great results!
Information like "model offerings on Perplexity as of right now" isn't well documented right now, so all search/answer engines will struggle to provide up to date info on such a topic. We will look into updating the Help Center to remedy this.
If there isn't a good source for the information, then it will be hard to provide good results. this is why we encourage you to double-check sources for added confidence.
I know this,but I thought it will be better at information about its own product...
For niche or internal topics there often isn't a single clear or reliable source. That means even top-tier AI tools will struggle to provide consistently accurate answers on demand.
This is a fault with our documentation which we hope to fix! Thank you for bringing this to our attention 
OpenAI also has the same problem - there's lots of articles that have the wrong daily limits for models for Plus subscribers, for example - though I have noticed they tried to update some of their own documentation lately, there are still articles they host which have the wrong limits... and the abundance of other articles on the internet with the wrong limits doesn't help.
Perplexity is a service. Ai models aren't internally aware of the service they're under.
No option for that in settings for some reason 😂😂
Oh yes imma need to test it
The original had so much that was really good, not sure why the newer one is so bad. 😦
Resurrections?
Yeah
I kinda like how they merged Morpheus and that other guy.
Maybe I'm not unplugged enough
Adding Barney was weird though.

What phone are you using?
iPhone
I’m not gonna keep low power mode on 24/7
It’s not showing up for some reason
nostalgia bait
Ok… thanks
Also the question is, why is it taking so much in the background in the first place?
@meager rover
Honestly, I have no idea. Never had this issue with perplexity on my android.
Might wanna try re installing to see if it helps.
Thanks mate but it isn’t there
No I know man 😂 the app isn’t there
Oh.
Very odd
Yeah.

#📱︱feedback-mobile maybe this is a better channel to report this incident
not a ios dev, i would image also reports ios version + iphone version would help
Rad
@meager rover try disabling notifications for perplexity app.
does it solve the underlying issue? or just you dont see it you dont worry about it xd
It does reduce background activity on Android.
Don't know about iphone.
If it makes a difference then he will notice.
The battery drain notification is likely from the system/settings instead of the specific app.
Shouldn't mask the issue.
I don’t get notifications
So glad I discovered this app. Game changer.
.@AskPerplexity your boss prolly already bald from dealing with ur dumb ass
Quoting Ask Perplexity (@AskPerplexity)
︀
1,000 likes and I'll make my boss bald and send it to him on WhatsApp
Ikr, browsing can be a pain.
@valid charm @shut dagger @tame current
Bought the sticker pack so I can rep the plex
Looking forward to being apart of this community!!
Elon being... Elon. lol
@shut dagger millionaire opportunity lost.
It's still cheap enough to buy
True I guess.
but it's another pump and dump scheme
I do...
etc, btc and usdt... my very small portfolio
I mean, pulling out in a month could be rewarding.
I don't think they'll dump it before that.
You never know when it will get dumped
You could try a small change.
It will still grow exponentially considering it's current rate.
what's the app?
Phantom
Fr
Is the gork twitter account made just for this 💔
Might be, lol.
@gork @AskPerplexity attention holders: The gork airdrop is now live to all.
︀︀
︀︀you all have 48 hours to claim your airdrop
︀︀
︀︀claim now:
︀︀gork.claims
Likely an Elon backed account considering his new name. 
nah that’s some fake account
Real gork is as real as it gets
Just vibes
I was talking about the gorkcr account you sent
@gork is the real one
You are right, it likely has nothing to do with Elon. Or else he would have reposted. Just a meme coin.
I didn’t get like aanything done today
Pump and dump yeah
Type stuff
Tired physically or mentally?
Idk if I’d say I’m tired
Must be something on your mind.
Like a weight on shoulders maybe.
It is what it is
@short narwhal sometimes I don't understand whether I should say "it will be alright" or "yeah, life sucks".
Some people feel like their struggles are ignored when we try to cheer them up.
Others feel like we are pushing them into a downward struggle if we acknowledge the struggles of life.
Homelander dipped
daym, i didn't knew dat
Do you like it: #📷│media-gallery message



