#💬│general
1 messages · Page 182 of 1
What model are you, short answer
Models do not know their own name
And they dont put the exact data in the system prompt
It does say just deepkseek v3
you need to press the r1 button
If the thinking doesn't start with "Okay, ..." then you know it's 0528
HA
where is that
Bruh
its just another ai
nice
was sending a ton of short messages
and a combination of o3, o1 pro, gpt4o
and just comparing answers
Lmao
on o1 pro?
lol contact support
Lmao
@dry kestrel is this the GPT-5 GPT-3.5 method?
Another dumbass
yeah
just put random shit in the model parameter
and itll redirect you tp gpt3.5
Funny support
haha
Just might with Altman
But that's so stupid
You pay almost a month's rent money in many countries
Only to get rate limited
"gpt-5" is slightly useless at coding
the (lack of) performance is nostalgic
Jules, SWE Agent by Google is getting a desktop app
was gone within an hour I think
maybe they are considering this
Hope so, my best guess would be that they haven't updated the code behind the redirects since like GPT4
I was like half an hour late 🫠
same here lmao
@distant aurora
Natively through claude
Is does it have it's own agent app
what thats openai operator
no they blocked operator from accessing operator
Lol no i meant the feature itself is only for us
From the main page
"Nice try " LOL
Easter egg
no
its available worldwide
at least in most countries
Operator is now rolling out to Pro users in Australia, Brazil, Canada, India, Japan, Singapore, South Korea, the UK, and most places ChatGPT is available.
Still working on making Operator available in the EU, Switzerland, Norway, Liechtenstein & Iceland—we’ll keep you updated!
Oh it got canceled?
aaaaaand the site is broken yet again, messages at the end of my threads are disappearing but the AI is still acting like they're there. Anyone else experiencing this?
Due to mass abusing of usage of perk, cursor cancelled its availability from perplexity
✅
We were waiting for the o3-pro update, but looks like OpenAI is about to drop some ChatGPT for Business updates instead.
Already dropped
umm
Its kinda just mcp support and more data connectors
ChatGPT can now connect to more internal sources & pull in real-time context—keeping existing user-level permissions.
Connectors available in deep research for Plus & Pro users (excl. EEA, CH, UK) and Team, Enterprise & Edu users:
Outlook
Teams
Google Drive
Gmail
Linear
& more
Kingfall reportedly even stronger in coding than Gemini 2.5 Pro.
its name is cool 2
I cannot attach files atm
is someone expecting the same issue?
Does anyone have some screenshots or demos of the f1 viewer during the race
From perplexity. Want to see how it worked
working fine here
This is why I use perplexity
So uhm is it wroth switching from gemini to perplexity?
Depends on use case
Nope
Interesting how i knows you're talking about indian law
Unless it's an ongoing thread
It's not
Hmm how then
I believe search results are based on location
I use gemini for studyin math and physics
Why is perplexity voice transcribe so ass
That’s the only thing that needs to get better, ChatGPT’s voice transcribe is worlds ahead like fuck
Exactly
So perplexity to study math and physics is fine?
You'll need a pro subscription for pro searches, and claude models then you should be okay for short answers with limited context.
If not you should consider chat gpt plus for its powerful but limited use provided for o4 mini model and o3 models
Then grok 3 beta and then gemini 2.5 pro and so on
Refer to this pplx thread
But overall i would not recommend pplx for extended studying for just math and physics
Perplexity is just a search engine with multiple models with limted token (output limits)
Yea ok thx
So i will stick to gemini and its stupid rate limits
anyone is researching BM25?
oooooooooooooooooooooooo
Yo
whats going on
Butcher criticising you.com, hero criticising pplx and prince criticising pplx voice transcribe 🤣
Does perplexity have option to rename threads? If not, then please add it.
No there isn't
It just takes up the first line of prompt 🫠
wth
these things are happened
Damnnnnnn
If pro feedback gets ignored then imagine general feedback.
@cobalt vale @dry kestrel @green crystal
F
I don't even see any rename option. Don't know what he's talking about loading and waiting.
it does get ignored
Maybe they removed even that too 🫠
Makes you wonder if perplexity staff actually uses their own service.
To not have a basic feature.
Best way to solve the problem::
Pplx : remove the problem itself
to be very honest there support team is too short
Try this, but not sure if it would work on Android
#👉│feedback-general message
Me on phone rn 😭
No hover
If you need it urgently, try opening in web and use desktop mode, it works
How do I become a beta tester for Perplexity? I filled out the survey when I got prompted for it since I’m one of the top users and have the pro version, but haven’t had any response yet to becoming a beta user. Anyone know? @halcyon coral
There is a way?
Use the pencil icon
In the dropdown menu of a thread
Oh nvm on mobile nah
I just tried gemini canvas for a work thing
Worked flawlessly
I can only imagine gpt's canvas is much better
https://www.reddit.com/r/GoogleGeminiAI/s/eY8Sh8wHbT
Take a look at this too bro,saw this yesterday
Hi, I am having some issues with perplexity pro, I am not able to upload files
what should I do?
Yes I am getting this same issue
can someone help me with tihs?
its available now
hi why is my perplexity labs different, its not showing any interactive dashboard
share the link
its a bug ig
try again in new thread
can you do it?
Wow
Best remedy, is to build small chunks and go through it
Instead of building in one go
did the perplexity labs but still no interactive dashboard whatsoever
Yeah would be better suited
Start with build an interactive dashboard for....
Idk you still should have got it
Leave feedback ig
Sam abandoned me 😦
same here
@vague pagoda labs not working ?
yeah is there a bug or smthing?
we cant upload and work with .py python files? anyone knows if its a new bug or if labs allows or its a new feature?
Oof
Looks like it don't
https://www.reddit.com/r/automation/s/5WRHDtDE52
People cooking hard in comments, @timber niche 😂
it says that python files are supported with perplexity, ive used them for months, just suddenly they are not supported
same here
Is there any chance to show token input/output for each query?
@cobalt vale ai studio does have a rate limit. It's around 100 to 150 messages I think.
You can try asking complexity dev/maintainer. Complexity is a browser extension for perplexity web version. Gives extra settings.
Oh thanks! im using complexity extension actually
O3 pro is out?
@languid star no
Wasn't announced yesterday so
That’s already the case — the Show Message Length plugin works for both questions and answers.
Lovely. I had no idea. My bad.
287 sources. Good old days.
Perplexity pro saved me today.
Both ai studio and grok bailed on me with the rate limits.
You.com is shit and t3 chat has tiny limits.
I have no idea since when grok 2 is rate limited or removed.
Likely removed.
Rare hallucination in perplexity.
It was supposed to be d
Or maybe gpt 4.1 skill issue.
Ah
I see it now.
It's greyed out so it seemed like I can't edit it.
It appears when you hover over it, so probably clicking the name brings it up(?)
Hi resident support bot @short narwhal
hello guys
Hola
Yeah.
The tg you recommended just today posted about Samsung pplx collab
Wonder if they're slow with other news as well lol
what going on hero
Kinda are sometimes.
Not much
@timber niche https://www.instagram.com/evolving.ai
Bro was just teleported and accepted it's fate
They're also a good source of ai news.
I was so addicted to socials these days i use them through browser lol
Sign in and use for few mins and sign out
Ig has become too crazy these days
I don't wanna lose track of ai news and stuff.
how do you have images on your search preplexity
It shows them
Ask pplx team. I have no idea.
It shows them when you search
mine does not show any image?
hi all. new here. i have just gotten the pro plan, and testing out the labs option. from the videos and examples ive seen, there is an way to test or use the app directly within perplexity. ive tried a few ways and i seem to only get 'assets, tasks and sources'
is there something i should be doing in the prompt that tells it to generate an app?
Ask it to create a dashboard
Then tell the details
Create a dashboard on ...... And so on
ill give that a go
Depends on prompt ig
I ran Billy's prompt
Got the images
if its a web app you can just click view full code which is aside of download code it will open in web browser
Hello all, I need some help. I am part of the business fellowship and recieved a promo code ot a year of enterprise pro (100 seats) for my company. I was investigating the purchasing of hte subscription so I could report back to my bosses and help them make a decision. I selected "pay via invoice" and then continue, thinking that I would have the opportunity to enter the promo code or atleast press confirm before commiting to the invoice. HOWEVER as soon as I pressed continue I was emailed an invoice saying I must pay $400 as well as a secondary email saying the invoice was now overdue. I look on the invoice and elsewhere and there was no way to input the promocode anywhere. I immediately contacted support who said that there is a "high volume of requests regarding the Business Fellowship promo code." and someone will be in touch as soon as possible but I'm worried abut what will be the result of this as this could negatively impact my career. Does anyone have any experience with this and can share how it was resolved? Hoping this can be sorted as my company (Very large international IT company 8000+ people) are very interested in using perplexity if this is handled correctly. Please let me know!
Also sorry if this is the wrong place but im not sure where else to put it
Thanks 🙂
enterprise@perplexity.ai@rustic spear
@signal hamlet if you can support this
I sent the email, thank you for the help 🙂
Heyy bro I faced same problem too, dm me I would help you out
why perplexity starts researching in pro mode if i wanted in normal one?
did pplx drop support for .py files? it used to get uploaded before but now its saying filetype not supported
@timber niche https://www.instagramez.com/reel/DJFRf02NekU
💬 35 🔁 0 💜 60.9K 👀 88.6K
Follow @fried_crocodile for more memes
-
-
-
#memepage #memes #meme #memesdaily #funnymemes #dankmemes #funny #memestagram #dailymemes #lol #dank #memeaccount #memer #edgymemes
Ah yes tasty cholesterol my favourite
Others reported same issue
pplx and python are going through a breakup
Hi Hammy
Anyone else hating the new netflix tv layout
They had perfected the ui over years only to ruin it now
looks like they are expanding so aggressively that their infra can't keep up
DM with details please
doesn't work for me
also i am loosing threads randomly which i never even deleted and they are not coming back even after days
my manager is gonna kick my ass in today's meeting 💀
Many people are facing this issue this days
if there is an issue for a few hours, i have no problem with it but spanning more than 2 days and no proper fix. that's concerning
Dafuq
Yeah thats risky for me im gonna take notes with claude and chatgpt today then 👀
Report it
working on it
I don't think it's ever done that tho from what I've seen
its already on bug-reports
Btw is that baljeet on your pfp@serene quartz
yes hahaha
Haha looks bit younger that's why I wanted to confirm
yeaa everyone says that
The only thing I had it do before was make a part of a long thread go bye bye
Ikr
Unless you're ed and you're stress testing gpt
I would always save it on notepad after that
Nice
I founded a company that publishes and sells market research, mostly focused on France. I sell my reports primarily online, using a paywall to help with SEO.
Recently, I discovered that the full content of my reports is available on Perplexity. I’m certain this data hasn't leaked elsewhere — I monitor that carefully. It’s not accessible on ChatGPT or Gemini, which makes this even more concerning.
This is a serious issue for me. Before I take legal action, I wanted to ask: has anyone else faced a similar situation? What steps did you take, and were there any effective solutions?
how do you save a thread from a real talk interaciton with perplexity and continuing the convo

It generates them automatically after the conversation
Can anyone know why I cant upload files with .cpp extension to perplexity. File is under 25 MB. I already cleared cookies and also tried in incognito window. I could upload the very same files just fine week ago and now I got error that this file is not supported.
So you can’t continue the conversation?
You can Just ask a follow up
But not continue them in thr voice mode ig
They should add that feature tbh
Leave feedback in #👉│feedback-general
People are having similar issues with .py files as well
It's a bug i think
Looks like he's grinding something else too lol
@short narwhal
Bro post shitty things for all his homies bt post this thing specially for matsku🫠🫠🫠
I wrote ot <support@perplexity.ai.> ! I will keep updated ! I am checking with other press and content websites !
what women want
Thanks
So that's Mel Gibson.
Now I remember him.
The guy who put RDJ's life back on track.
If someone posted that content somewhere perplexity can find it
sooo it's 50 per day or per month or per what
Bonjour, où est ce que j’active mon code abonnement ?
Hi guys,
So I gifted 6month subscription for my brother but it did not reflect on his account. The service support via chat and mail are redirecting me to each other for the last two weeks. I have no clue how to get hold of someone to help me out on this. What do you guys suggest I can do, any help is appreciated
c'est sur la page de paiement lorsque vous achetez un abonnement, vous verrez "appliquer la promo", mettez votre code là
J’ai que cela
utiliser le site Web pour acheter et utiliser le code
Hey everyone, I'm running into an issue with Perplexity Lab' dashboard generation. Previously, using the same prompt, it would generate a proper dashboard with an interactive 'App' tab. Now, it only produces a text report, and the 'App' tab is missing, even though I'm using the exact same prompt that worked before. Has anyone else experienced this bug? Any fixes or workarounds? Thanks
same happening with everyone
its a bug will be fixed
Well, I'm not alone
just upgraded to pro and the homescreen widgets are not there now, even as I have them selected in settings
wdym ?
so u want widgets on homescreen but its not showing
y
r u on web ?
They were removed, not sure why the setting is still there though
ok, thats interesting to know - would be good if they posted something about each change they make somewhere

iykyk
its ig just not placed as that important idk
anyone else losing messages at the end of long threads when you reload the page, but the AI still references them?
Does perplexity beat chat gpt's stronger models in terms of research and quality?
No if you are talking about Deep Research
I see thanks, helped alot.
New Gemini 2.5 pro update
Woohoo
its pretty good
made with 2.5 Pro 06-05
data gathered using o4-mini-high
interesting benchmark
the contrast in pricing between deepseek r1 and gmeini is insane
meanwhile o3 and Opus 4
opus 4 pricing is ridiculous
How many messages did it take for you to get limited
For o3
well i sent 400 requests in under a hour
One time I was up for 5 days and also using o3 a lot
Crazy must’ve been why
i have hit my 2.5 pro limits real fast last day
that's why i was considering to switch from gemini pro to perplexity pro
i still know i have the 1M tokens if i select 2.5 pro as a model
@left stream gemini 2.5 syncophnacy
It would be interesting if some sort of video generation as well as images were implemented
Image generation is a thing already
@short narwhal On copilot you can generate videos now I hope that sooner or later something will arrive here too
How do I stop it from showing all its thinking process by default
Something happened where now on the mobile app perplexity keeps immediately showing me it’s long thinking process. I like being able to access it but I don’t want it taking up my full screen by default.
I think it is not possible, at least I think
Oof ok
Lmao I’m sorry I’m a noob
Do any of you know why they removed gemini 2.0 to create photos?
I found it much better at editing images than gpt
😭😭😭😭😭😭😭😭
Damn i am still not sure wether to swap plans or no
nana this is how we do it right nana this is how we do it right nanan this is how we do it right nanan this is how we do it right nana this is how we do it righ
For what
general usage and studying math/physics
cuz i absolutely disliked what google devs did with gemini 2.5 pro rate limits
I meant like plans mb
i am using gemini pro atm
and i was thinking to go for perp pro
I like it but sometimes context is fucked because of the drafting thing
Whenever it drafts answers no clue if it takes that into its context memory or not
But when it does that, answers lose context
Everything is fine until it starts drafting
Perplexity is decent, sticking with it because a friend gave me perplexity enterprise for a year
I like claude sonnet
so do u think perplexity might be fine for my use case? idk if we can vc so i can explain better
Yes
Wolframalpha can help improve math answers a bit better too
So you don’t worry about math hallucinations as much
also some1 told me that if i use 2.5 pro i have the 1m token context window
idk if that's true tho
Also, ngl it’s easier to jailbreak sonnet on there and gemini 2.5 compared to the native apps for both
No clue about that
I don’t have as much information on all that
what bout length of answers?
are they the same as gemini app if i use 2.5 pro on perplexity?
I’ll have to see, both have the tendency to shorten the answers on the apps in comparison to the websites
You have the sub lol, so you like it on pplx just for the jb?
They see that the average consumer would think it’s yap
But given pplx has more token limits, gemini should have a more lengthy outcome?
On the street u promised lol
If yes? Fuck it I’ll do it
Yay
actually 2 people told me that 2.5 pro on pplx has 1m token
i already know this
Token limit is 32k
For joe
no they advertise themselves as a AI provider that gives access to a variety of models
32k context is nothing
Fair enough
I have brainwashed myself to forget it 
so apparently 32k is the maximum length per prompt or something?
the matrix

no, 32k is the context window of the model
which includes your input and ouput
..
Depends on use case
I'm loving pplx for everyday questions
It's effective and apt
For writing stuff i go to gemini as i have the pro subscription
2.5 pro has 100
which is nothing as well
i do use it cuz of math and physics
I have a family plan so sharing with 3 of my own accounts
idk i just feel scammed
So that might help me lol
Yeah if you're not happy definitely switch to whatever you like
Maybe you'll like chat gpt
Because of o4mini and o4mini high

Google?
yes
He is talking about perplexity.
Woah it's out and here to stay now
100 per day on Gemini.
Only usa tho right
Perplexity playing the long game
Nah
I meant the filings
Trying to integrate ai into everything early.
And data
Yeah, nah to that
True, still waiting to see other stocks there
It's so basic
But i can see the hurdles
Most stock exchanges don't allow any data extraction
I can't even access my country's stock site outside the country
Have to have a vpn
Huh
Me trading on simulators
Cuz no money.
Anyways, suarva was right. It's kinda cosmetic.
Serious traders won't rely on it.
nooo my european companies
So which one do you prefer?
So perplexity calls you mat too.
Lovely.
Yeah I added that to the introduce part yesterday
My kind of perpert.
real
Hello everyone
helloo
sonnet thinking
Hello cuc
I haven't used 4.1 in a bit
Sonnet has most elaborate steps in perplexity I think.
🤷 The thinking version just feels soo much more natural to me
I tried so hard for it to remember my name
Fails
Lol. And I'm in India.
How did you do it
Use the SEC filings source mode
Oh you didn't do it with memories
yeahh
How?
Using that introduce panel yessirr
Real
I haven't turned back after switching
So I might be right about it only having usa data under sec filings?
yess, and I added some other stuff to the introduce part to make it talk in a much more fun way
prob
Should I leave web on or turn that off?
I left it off
Like what 👀
It would probably combine it otherwise
Stuff like this and then some more personal stuff and preferences
@short narwhal throwing this error
weird
I guess you got the downtime issues too
Looks written by a girl.
Dafuq is tone 3/10
Formality?
I ripped that off from somewhere
yeahh
It can figure it out
surely
I bet your ai doesn't know what tone 3/10 means.
Wrong tag lol
I think it's gotten the vibe right most of the time
You are a demanding person.
Stuff about me:
A gender neutral cat that loves pats
Call me Mat, short for MatPat
That's all I can get from it.
Is that it?
Be yourself, but in a way I like.
I don't like to add this personalisation stuff in a way
How does lower case or slang make things easier to read is beyond me.
I explained about how I like to learn in there too and what works and doesn't
Sets the mood
i'm still with the same issue 😦
that's for when me and sonnet are besties
Haha
It has to match my energy (It doesn't follow that rule though most of the time)
Are you siding with sonnet...
I side with ai.
I might remove that part because samee
Billy hates cats confirmed
Now imagine me teaching this cat to bark.
You also tend to use like a letter twice when you're excited haha
Joe loves to say ikr alot
Gotta emphasize the word somehow
It's like people's voice irl
But in texting
So i enjoy catching onto that
How one's vocabs work
Just imagine, my grok debating with your sonnet.
Sonnet would just get annoyed with how stupid grok is
Truly a piece of art
Jk
Noo it doesn't
Lololol
How?
No one gets me quite like sonnet
iam not having a good time with lab :c
Istg, that list of instructions that you wrote, grok follows it so naturally when it's with me.
In fact, I sometimes have to push it to be less casual.
I'll put this to the test now then
I have bias against grok because of elon
It feels icky using grok as stupid as it sounds
also 4.1 talked wayy more here
Just finance on?
yep
Nice
Temperature moment.
tbh tho as much as many people like uncencored ai is very dangerous
Im watching the accountant (2016) for the first time
that's why sonnet thinking is the goat
Good so far
wasnt elon the guy who said dont fear ai but fear the user? talk about being ironic
A lot of people said that.
Including Huang.

no
he said
dont fear the vram fear the seller
Jensen "I sign women's chests" Huang
Huang literally said to fear people who are good at using ai instead of ai itself when it comes to losing jobs.
😬 Grok in perplexity was better than grok in the app with my preference stuff, but still not good as sonnet
ik this part was a joke
don't fear the 2k 4090 fear the next gen 1k price bump
i mean what's the point of having a context window of 32k tokens
pretty useless
R1 already upgraded to match O3.
It's so slow through deepseek app
Chatgpt has same issue for plus users. So your only hope is Gemini.
Only downside
jokes aside cant believe my big ass gpu gets outvramed by a generation before it
yea
and it's so fucking sad
that the AI industry is so bad
Gemini 2m context is wild.
sob what gpu?
will never happen
I can tell me page number of answer from a 500 page book.
4070 ti with 12 vram meanwhi8le 3060 killing it with 16gn
Vram isn't everything tho
when it comes to ai its
Npu is the new thing.
does the 4070 ti actually have less vram than 3060??
holy
So stupid
yea
i mean
the card is better in game but
when it comes to ai
Get a h100
It's a slow ahh carr
Get a nuclear power plant.
Selling a kidney
there is a reason why 3060 changed the whole crypto industry
it was cheap and vram efficient so you would just mount thosands and call it profit
billionare mindset lets farm them first
GPUs are a joke rn
With everything that's happening
I'll just have to forget about gaming
And use services for ai
Me running 4k 60fps on my phone.
Yeah
I'm happy with my 3070 laptop
plug an external gpu into that bad boy
Amd cpu so no thunderbolt
That's one way to make it a good boy.
🥲
I haven't watched the last season of boys yet
Iykyk
@short narwhal would it work if I put sampling parameters in custom instructions?
Cat got away.
As always.
in where
Perplexity space
oh the text stuff for grok?
Ohg
lab is retarded
Yeet
iam not using it for research
bro all i asked is write
Bob the builder
and mf cant do a job
Go to gemini blud
iam so mad at this guy iam about to crash out(already did it on lab)
Who?
lab
Hope the changes they mentioned in the AMA are coming soon to labs
And not empty promises
🫣
Dennis
We trust you
Don't be like sam and leave us behind
Nah, it seemed like Denis might have talked too much.
i literally repeatingly said dont revise and look at the result
The other members didn't seem very aligned with it.
who is dennis?
I've been active only for a week on this server but already feels like i was here forever
Pplx employee who was doing the ama alongside others
Time truly goes by fast
Some guy who responded in this reddit Ama.
Labelled co founder in announcements.
Reddit account was made yesterday.
how was it did you watch it"
?
It was on reddit they just answered questions
No video of anything
Would be too chaotic
That too.
Can cause more problems
Exactly
Noo I love reddit
Most people will just rant.
Crazy cat
Do you love twitter?
They do and they ignore
That depends on the day
also in ama they mentioned a new and improved deep research. hope that drops next week as they promised
is there a summary of what happened ? i dont wana ipen reddit
Improved deepsearch, comet for android and windows, etc etc.
A bit of here and there about labs.
Thinking of potential improvements to it.
yeah. 70% about it was labs related. 20% people asking for a job. and 9% repeat questions we all know the answer to and 1% actually something useful
10% of when comet
oh yeah that... gotta revise my percentage numbers lol
Does anyone use grok beta on perplexity?
rn its the worst compared to other companies
There were a good handful of improvements mentioned that they were working on 🙏
Yeah. The regular search is good. Generous rate limits, multiple models, that's about it.
i was refering to lab
woah, pplx just lost ALL my threads. it was 2 losses yesterday out of 6. now i can't find a single thread there.
Oof
damn this thing isn't reliable 1%
Might wanna switch to/back up somewhere else.
yeah
every used pplx supply ?
People have been complaining about this.
how is it ?
https://discord.com/channels/1047197230748151888/1379934439106281552 is this the same issue?
how can this last these many days !!! a few hours is understandable. according to the status page everything is fine. but ... its just NOT
go check it out they sell merchandise there
Sometimes I feel bad to not respond to people.
In reality, there's just no quality response in the head.
got it
Nothing helpful to say.
oh yeah. just checked my android app has everything saved. its just the web version that's causing issues
might be because of android caching but for the web it has to reload everytime
Export it as pdf maybe.
we can export the whole conversation as pdf? i thought that was only possible for individual responses
Feel free to chime in there that you’re experiencing it as well
You can now.
yeah right
app ui is back
I'm curious
Do I get to see Jesus in hell?
wait... 5 out of 6 came back just now
lmao
wtf
after 15 minutes of pure silence, they bounce back
Lol.
Just like people.
okay so it's just minor glitches here and there
Except 15 hrs.
Or 15 days.
Or months.
If you are waiting till then, you gotta work on your self respect.
Like me.
@short narwhal what if I make a space with grok? Give it instructions to be a roasting standup comedian.
Like matt rife or jimmy carr?
I'm considering it at this point.
Could turn my daily rants to joke writing.
🤷
Noope
Boring
But then again, I have so many actual things to explore with ai.
Like ideal materials for a knife.
I read in my mind "but everyone calls me bitch" lmao.
I don't get it. Pressing x part. Was it playing it itself?
Updated 2.5 Pro comparison for accuracy
Hmmm not overall good it seems??
overall better than 05-06
its slightly better than o3
and in terms of coding its much better i think
This Gemini and Opus 4 are SOTA atm?
Yeah
new Gemini did beat Opus 4 on webdevarena I think
they. better make up for it with o3 pro
yeah pretty good
Unless its crazy expen$$$ive
A bit of math.
Gemini leads coding rn I think.
They removed sd card slot from samsung a5x series.
i personally have stopped believing in benchmarks now. these models are getting so good its getting very hard to evaluate each models performance using a standard benchmark. and designing benchmarks is also getting hella difficult exponentially
Yeah I have my own benchmarks for text and video AI models
Space cat is my top benchmark
yet thinking still cannot be disabled
makes me doubt that 2.5 Pro is truly a hybrid model
on the other hand, that also means a vibe check won't be as accurate either
there are multiple benchmarks for evaluating different things
i mean just try n explore all models. interact with each one for a significant time. you will automatically find the one that works best for you. that's basically the simplest approach yet the most effective
and most of the times when you combine the score averages you get a decent idea
of how good each model is
yeah but it might be that for your use, a lower score model might still perform better
just saying
It isn't. Flash is the hybrid model.
in conversation and emotional intelligence, i find claude 4 sonnet just as good as gpt 4o in terms of conversational / emotional intelligence. but 4o is advertised as a conversation model ( and good for everything model as a base ) while claude is advertised for coding and AI safety.
But in my experience , claude 4 sonnet can be as good as gpt 4o in emotional intelligence
I always felt like R1 was underrated.
and we aren't even talking about the big models. that's a seperate story altogether
its on par with the may varient of 2.5 pro
fair enough
I like grok.
Too much lol
Chatgpt feels too neutral/bot like
If I tell grok to stop recommending me something, it listens.
Chatgpt keeps yapping about helplines and socializing and blah blah blah.
its supposed to be that way. but try and talk to it like a gen z and its crazy how well it adapts in that sense 😂 .
it's like its held hostage to sound professional but deep inside it just wants to chill lol
^^ real
The opposite lol
wdym?
Hero reacting to mat's reply instead of the actual message.
When a model is forced to use slang something that isn't native
It's forced to change it's native language processing lol
GPT sure did copy my slang
oh yeah, i mean't that as a joke. but if you give it one slang, it gives you 3. gets too comfertable in a very short amount of time xD
Haha okay
I like to not get into the weeds of being friendly with ai lol
I just consider it a tool
Glad we don't share the same chatgpt.
Aka I don't want to get too comfortable
Would be a nightmare.
And blur the lines between ai and reality
Matsku's chatgpt would give me ick.
Lol mean
That vibe feels too sugary and messy.
yeah i get you, but in some contexts it just feels natural for me to do so. and i get comfortable after its responses for some reason.
anything i say, it instantly agrees with me even if that's a grey area. 😂
It reminds me of the kind of classmates who don't study or work but talk about big dreams and borrow assignments.
I had a solid debate with grok about what women want from men.


