💬│general

[distant aurora]

It's not like Qwen tops the leaderboard.

Maybe minimax doesn't but others clearly match their benchmarks.

https://cunnyx.com/i/status/2023331885402009779

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) (@teortaxe...

So speaking of benchmarks, what can be said of the new open Qwen? First, it completely destroys Qwen3-VL-235B ofc, but more surprisingly it outscores Qwen3-Max-thinking.
︀︀All the while it's the same model as "Plus". Plus just has 1M context and some more bells and whistles.

Quoting Suvash Sedhain (@suvsh)
︀
huggingface.co/Qwen/Qwen3.5-397B-A17B its out.

**💬 6 🔁 1 ❤️ 54 👁️ 4.6K **

CunnyX · Feb 16, 2026, 9:41 AM

[gloomy cipher]

They fail so much in agency

Its so tiring

It doesn't to me

[distant aurora]

https://www.kkinstagram.com/reel/DUwVHH5gtlT

[haughty blade]

I didn't understand, what has been changed?

It's just "Perplexity good for medicine and e.t.c", info not useful

[short narwhal]

Uses Opus 4.6 now and more thorough and accurate in general

[haughty blade]

What about limits? Is it bug that my deep researchs limit down from 500 to 20 per month?

[short narwhal]

Not a bug like I said

[young hollow]

The Deep Research Limit is Now 20 a Month

But The Quality has Indeed Increased Alot Because it uses Opus 4.6 Now most Expensive and Top Tier SOTA Model

[haughty blade]

It doesn't make a sense, I'm ready to use for example gpt or gemini, I need increased limits and I know that base models like gemini or gpt can provide good quality

[oak meteor]

Waiiitttt, whaaaaattt??!

[young hollow]

Then Switch to Those

[haughty blade]

Increased limits*

[young hollow]

Gemini Gives 20 Deep Research a Day and You can Share One Sub with 5 More Accounts

120 Deep Researches a Day

The Pro Ai Plan

If you want i can share a link for 4 months Free Trial

[willow kettle]

https://x.com/testingcatalog/status/2023346936145760549?s=46

TestingCatalog News 🗞 (@testingcatalog)

Perplexity is testing a new Gamma (ASI) mode, powered by Grok. It is ultra-fast and claims to use Grok 4.1.

Speculation 1 - Perplexity might be preparing for the Grok 4.20 launch during the upcoming week.

Speculation 2 - Considering a deeper integration with Grok and the fact

▶ Play video

X · Feb 16, 2026, 10:41 AM

[twilit ether]

Guys which ai for learning

New things

[green crystal]

damn nice

@distant aurora qwen looks good so far
did you test its accuracy?

[wise nova]

https://www.kimi.com/bot

https://x.com/i/status/2023029674549596301

Kimi.ai (@Kimi_Moonshot)

Introducing Kimi Claw🦞

OpenClaw, now native to https://t.co/8PH4O9dh6m. Living right in your browser tab, online 24/7.

🔹 ClawHub Access: 5,000+ community skills in the ClawHub library.
🔹 40GB Cloud Storage: Massive space for all your files
🔹 Pro-Grade Search: Fetch live,

▶ Play video

X · Feb 15, 2026, 1:40 PM

[dense path]

notebooklm with a google gemini subscription

[twilit ether]

I have google gemini pro, does it work?

[lilac cypress]

Its difficult to say...
i recommend Gemini flash...💪 is good for all kinde of things. But If your are an programmer then of curs Opus 4.6

[twilit ether]

I used perplexity for learning things but my pro have expired, cant use the learning feature anymore

[twilit ether]

Im programmer and i want learn any tech from fundamentals with updated knowledge

[lilac cypress]

@twilit ether Then use Perplexity  / renew the subscription LOL and then combine it with NoteBook LM/ Opus 4.6 🙂

[twilit ether]

I hope so (:

[shell lily]

do we no longer get access to reasoning models as a pro user? We have to pay $167 to use claude reasoning more than 20 times?

[dry kestrel]

ayy

[frosty moat]

Deam perplexity took my money but didn't give me pro

[lilac cypress]

Yes the Ai tools, it's just too expensive... Perplexity, then Gemini Antigravity, then Notion Pro, maybe you still need this and that. All the Company want your Data and your Many....

[wise nova]

Why is Kimi so expensive for a Chinese model?

[novel arrow]

lmao, $15 instead of $20, as good as opus 4.5 triple limits than claude and you think it expensive?

[fresh aurora]

Hello.
Please could you explain to me how can I change my email address in the profile?

[chilly stump]

Any new updates on the apps ?

[queen nymph]

Not only that but image generation is basically gone too.

[wise nova]

Yes, Chinese models should start from $10. Somehow Kimi is the most expensive Chinese model

[chilly stump]

Model council should be available for pro users, why just max

[novel arrow]

you are delusional my friend, so americans deserver $20 but chinese only $10, even if kimi was $20 it still a better deal than claude.
If you want to pay less buy a better gpu and run kimi locally and stop complaining

[green crystal]

they still need to pay for their electricity bill lol

or rented compute

[novel arrow]

he thinks it's aliexpress

[wise nova]

Did you compare Kimi as Opus 4.5 according to benches? What about real life? In real conditions Kimi is like a Sonnet

[hasty pebble]

china #1

[novel arrow]

brother you dont seem very smart so you probabbly dont build complicated stuff, you clearly don't need opus 4.6 nor opus 4.5 and Kimi is in between sonnet 4.5 and opus 4.5, so just take the kimi it's only $15 with high limits

[dense path]

absolutely no need for perplexity. go with NotebookLM for an interactive tutor and maybe boot.dev to get some gamified fundamentals (if thats your thing)

[left stream]

not rn, I have been super busy with helping friends and cybersecurity and studying 😓

[wise nova]

China used the cheapest oil ($25/barrel) and gas from Russia, they have own resources, costs is 3-5x lower than in US

[left stream]

but when I have freetime sure, in what?

[green crystal]

true but we dont know if kimi uses rented compute

[left stream]

deep think is garbage lowkey

[green crystal]

with whatever you tested 5.2 pro etc

[left stream]

4.6 opus is the king rn

and opus is free on antigravity rn

[green crystal]

especially 5.2 pro

[left stream]

why not 5.3 pro?

[green crystal]

we dont have it bruh

[left stream]

[novel arrow]

its only codex 5.3

[left stream]

use inspect element 😉

[green crystal]

5.3 pro is getting hyped to be a beast

but uhm

[left stream]

there exists a method to use 5.3 pro

[green crystal]

yeah we'll know once its out

[left stream]

but it only works for 10-15 minutes

then it resets

and then you gotta do more bs to use it again

but anyways I gtg sayonara

[green crystal]

dafuq then youtube would be swarming by now

with 5.3 news

[left stream]

yes but all those guys dont do cybersecurity

[green crystal]

well if you're right then 5.3 release is very close

[wise nova]

And chinese seems to cheat on models training but also use innovated computing methods

[left stream]

at the same time it could also be just a really shitty hyped up model named 5.3 pro and it’s just a placebo

there have been instances for example 5 pro

[hasty pebble]

china #1

[green crystal]

true but they can also shaft eachother for the sake of competition

[left stream]

they routed to o3 pro back then before 5 pro released

my knowledge might be wrong so don’t quote me on that

[green crystal]

well atleast its a thing lol

[left stream]

tbh I doubt it’ll save openai

4.6 opus is good asf

man it helped me learn a lot in cybersecurity fast asf

it helped my team who’s ranked 32 ranked nationally in the united states

they all are leveling up significantly

[cold swan]

Only for max or pro users too?

[left stream]

so this is the best fucking ai in the world

[left stream]

antigravity gives it for free

[cold swan]

Oh thx

[left stream]

expanded usages for students with the pro sub on google gemini

behold

[twilit ether]

Wait i just realized, is deep research only for MAX now?

[green crystal]

its what we said last hype cycle lol

[left stream]

https://gemini.google/students/?utm_source=gemini&utm_medium=paid-media&utm_campaign=students_sem_bts-sem_generic__text&gclsrc=aw.ds&gad_source=1&gad_campaignid=23516487977&gbraid=0AAAAAC_C9PTZL0ZvrlFxrfuITg56XsmVf&gclid=CjwKCAiAncvMBhBEEiwA9GU_frtC8Po4-xHoc5dChqvKlTsCrYiGhXD-EC-I4R_GAbe0t__EhlPfZRoCqGYQAvD_BwE

then dont even use gemini

gemini is shit

[wise nova]

What chinese sell on world markets for $100 in their own internal market selling for $50. So compare sub price for Kimi for chinese

[left stream]

use the account you got the student sub on but for antigravity

then select 4.6 opus

there’s a reason why it’s good number 1 in livebench

[wise nova]

Can I use it for deep Research?

[left stream]

and it proved it to me since I’m the only guy who can do the classes concept by head instead of with a chart

its that good

[distant aurora]

It's making up numbers for stuff in games. I don't know what else to test.

[left stream]

my class can’t challenge me now

[green crystal]

another benchmaxxing model then

[left stream]

i mean, it can control your browser for you and it controlled mine for 25+ minutes so I guess if you give it a deep research prompt while it does researching it can, since 25+ minutes rivals deep research no?

the difference is that perplexity might do deep research but it might give wrong pricez

[green crystal]

method of searching and depth matter more than speed

[left stream]

but if opus 4.6 directly controls your tab and sees the info on the browser then does it for 25+ minutes while being powered by the number 1 ai

then you get a lot

[left stream]

yeah and I can tell perplexity’s comet is a failure in this

useless prices that were so wrong

[distant aurora]

Or maybe the model is capable but it's search tool is shit.

[green crystal]

browser control deep research is also different

its looking at sites one by one

[left stream]

I rather it looks at sites one by one instead of giving false bs info

[green crystal]

deep research is the model processing multiple sites at once and working with results

[left stream]

one tiny mistype can be a huge fucking mistake for me in the field of cybersecurity

over my ass I’ll look like a clown

[green crystal]

yeah i hope you still check what claude spits out lol

[left stream]

I do, I gotta suffer mastering the topic 10+ hours everyday

i did it and I have a professor who is extremely prideful

[distant aurora]

I made a deepresearch custom instructions in Chinese. To fit within 1500 characters.

[left stream]

so it worked out for me

[green crystal]

yeah because while claude is among the top with accuracy its not 100%

[left stream]

for cybersecurity I approve

that masters degree computer science professor is extremely prideful

i gotta suffer with 10+ study sessions because of him

i can approve claude with a satisfying 100%

[distant aurora]

You mean arrogant?

[left stream]

i wouldn’t say arrogant he rejected an FBI offer and has 2 patents in this field

he also teaches AI

and makes 100k+ a year

so I cant even call him arrogant

[distant aurora]

So ego but competent.

[left stream]

yes

isagi type shit

[distant aurora]

Fair enough.

[left stream]

yes, but over my ass I’ll be his little circus clown

8 projects and 1 exam in a sunday and then demonstrate the skills in 100k+ worth of tech

claude saved my ass in teaching me

if I simply asked for the answers I would look like a clown when I gotta mess around with the tech

[distant aurora]

I would prefer that over some incompetent manager giving orders anyday.

[left stream]

no ai is saving me in that

[left stream]

the pressure is insane ngl

I’ve seen 3 people drop out on the first day 😭

[distant aurora]

It's giving me "the bear" energy.

Not sure if you have seen that show.

[left stream]

if only they knew dropping out meant paying $700+ on the classes

oh well thats life

[left stream]

no what is it about

[distant aurora]

Pressure in running a restaurant.

Perfectionism and chaos.

[left stream]

i definitely remember that now 😭

[left stream]

its been a while

[left stream]

yeah, its like that 100%

he made sure to let everyone know that if anyone is on their cellphones they’re disrespecting him and wasting his time and they deserve to be failures

[distant aurora]

Lmfao

Whiplash type shit 💀

[left stream]

100%

it is

10+ hours of studying to master 1 little bitch topic

[distant aurora]

Wait so you aren't allowed to research in the middle of the lecture?

[gaunt radish]

Claude vs perplexity what do you think?

[left stream]

and then it goes all to the garbage the next day when he introduces 1 new topic

that’s how crazy it is

[distant aurora]

https://tenor.com/view/grok-grokai-grok-ai-grok-coin-grok-meme-gif-7538492393455586828

[gaunt radish]

I just got perplexity pro idk what to do with it lol

[left stream]

it all depends on the user

[distant aurora]

You'll be fine if you really don't know what to do with it, lol.

[left stream]

in my opinion, perplexity is better

claude will rate limit tf out of you

they’re clowns

[gaunt radish]

Damn

No but like what do y'all use perplexity for

Daily

[left stream]

yeah even when you pay them $200

they’ll rate limit tf out of you

[gaunt radish]

That sucks, my friend told me that claude is good for finance stuff

I was leaning towards getting it

But now I'll reconsider

[left stream]

it is, but rate limiting is bad

you can use 4.6 opus

but then boom

good luck getting more messages

it’s understandable why they rate limit ever since that asian programmer made them waste 200k for nothing

[distant aurora]

Bro use any sota model with good search tools. Search tool quality is more important than model iq for finance.

[left stream]

we cant have good things in life because of people like that

[gaunt radish]

Sota model?

[distant aurora]

Top model from any company.

[left stream]

claude opus 4.6 is my best recommendation for that

[distant aurora]

State of the art

[left stream]

and free

[gaunt radish]

Which would you recommend personally

[distant aurora]

I thought gpt high better than Claude for search tools.

[left stream]

oh hell nah

I stopped using gpt ever since that fatal mistake

i was prideful of gpt

but that single huge mistake made me change

I’ll show you something

[left stream]

all free for me

[distant aurora]

Well one thing anthropic does better than anyone else is limit hallucination.

[left stream]

I won’t lie they’re a bitch to try jailbreaking

but on antigravity

it’s surprisingly unrestricted

does OSINT (open source investigations) like nothing

like deadass like nothing

[distant aurora]

It's probably unrestricted on you.com too.

If you still use it.

[left stream]

no

stopped ages ago ngl

[distant aurora]

Relatable.

Same.

[left stream]

but here’s a tip

[distant aurora]

But anti-gravity doesn't have web access, right? Or does it?

[left stream]

if you want unlimited opus 4.6

[left stream]

yes

it controls your tab for you lol

i didnt want just that lol

i wanted more

so I made 100 ai agents that have their own operating systems they can control

then I craved more

I wanted them to be able to do penetration testing (real hacking) and then OSINT (open source investigations on companies and people)

[green crystal]

no sysmprompt i guess

[left stream]

imagine having 100 of them at the palm of your hand

it feels good

[left stream]

no hahahaha

[left stream]

who knows 😜

but definitely no prompting required

for the unrestricted requests

i think it works in sync with opus banan pro too so I’m guessing a prompt hidden for that?

[distant aurora]

@green crystal bro do you ever feel like everything you see around is ragebait?

Then you start wondering what's even the point of going out in the world or interacting with it.

https://cunnyx.com/i/status/2023051313374703827

Chubby♨️ (@kimmonismus)

Kimi releases Kimi Claw: OpenClaw native at Kimi:
︀︀
︀︀A browser-based AI workspace that runs
︀︀-24/7, offers
︀︀-5,000+ community skills
︀︀-40GB cloud storage
︀︀-pro-grade live data search, and third-party OpenClaw integration
︀︀
︀︀One thing you have to give China credit for: they know how to quickly integrate hype into their products.

Quoting Kimi.ai (@Kimi_Moonshot)
︀
Introducing Kimi Claw🦞
︀︀
︀︀OpenClaw, now native to http://kimi.com.">kimi.com. Living right in your browser tab, online 24/7.
︀︀
︀︀🔹 ClawHub Access: 5,000+ community skills in the ClawHub library.
︀︀🔹 40GB Cloud Storage: Massive space for all your files
︀︀🔹 Pro-Grade Search: Fetch live, high-quality data directly from Yahoo Finance and more.
︀︀🔹 Bring Your Own Claw: Connect your third-party OpenClaw to kimi.com, chat with your setup, or bridge it to apps like Telegram groups.
…

CunnyX · Feb 15, 2026, 3:06 PM

Nice

@left stream you ever considered playing around with kimi?

[lament imp]

I think it's meant to say "Advanced", but could be wrong. May need to try it out later.

[distant aurora]

@dry kestrel real? https://cunnyx.com/i/status/2023174545314570549

Nima Owji (@nima_owji)

BREAKING: xAI is working to add PARALLEL AGENTS CODING to GROK!
︀︀
︀︀The project is called “GROK BUILD” which used to be called “GROK CODE,” internally!
︀︀
︀︀It’ll support up to 8 AGENTS and the ARENA MODE!

**💬 46 🔁 52 ❤️ 567 👁️ 19.4K **

CunnyX · Feb 15, 2026, 11:16 PM

[gloomy cipher]

kimi is really good

especially kimi agent

i got 1 month for 1$ (first month only) hehe boi

[sharp matrix]

How did you get this only $1 ?

[gloomy cipher]

kimi discount on first month

[sharp matrix]

Oh my gotto

For all countries ?

[gloomy cipher]

ye

you need to create an acc and go to this chat

[distant aurora]

@gloomy cipher @dry kestrel @green crystal https://cunnyx.com/i/status/2023321917835665667

Okara (@askOkara)

Opus 4.6 vs. Minimax M2.5
︀︀
︀︀Prompt: Build an interactive solar system from scratch.
︀︀
︀︀Opus 4.6 tried to create a beautiful UI, but the sun’s shadow ruined it. Minimax M2.5 built a simple version that works beautifully.
︀︀
︀︀The winner is:
︀︀🥇 Minimax M2.5
︀︀🥈 Opus 4.6

**💬 14 🔁 15 ❤️ 283 👁️ 23.5K **

▶ Play video

CunnyX · Feb 16, 2026, 9:01 AM

[gloomy cipher]

bro iam here

😭

[sharp matrix]

🙃 Kimi 😉 rejected me

[sharp matrix]

Oh my gottooooo

You are magician bro 😱😱😱

Thanks for this ✨🥂

[gloomy cipher]

scroll down

[sharp matrix]

Ya ya I saw your chat link, now Kimi chan is mine 😍

[gloomy cipher]

are you an ai bot? 😑 what are these emojis man, perplx up

[sharp matrix]

I’m human from mars 🛸👽

[deep sparrow]

What country?

[left stream]

USA

[left stream]

might apply to more

don’t know since never tested outside of it

lowkey I made a second account too

so if I hit the limit I switch to my 2nd university email and bam

[distant aurora]

@green crystal do you want to try my chinese custom instructions on your extended thinking?

[weary jolt]

Hey guys why did the file upload rate suddenly dropped so much, like it literally would only allow me to upload 5 files in a day.

[obsidian patrol]

i have the pro version, why is the assistant that controls my screen limited per monthly use now?

it used to be unlimited

i feel like the more people gets comet and more people use it, the devs get greedy and greedier, smh its the same with every app

to get unlimited, no one will pay 200$ per month for a website app

[stray halo]

just found a strange channel on yt

‪@SyCore-BlackBloom‬

they just post random shapes with what seems like a hash?
i have no idea.

[meager rover]

Kimi is my goat

[brittle sphinx]

Hmmmm

[naive sage]

OpenClaw creator now works for OpenAI: https://www.theregister.com/2026/02/16/open_ai_grabs_openclaw/

[green crystal]

my first guess is agent becomes more autonomous

[viral remnant]

Can I get context for above messages?

[sharp matrix]

I got it for 1$
Tnx so much

[opaque current]

Perplexity have a bot in discord?

[dry kestrel]

interesting

Lost some respect for MiniMax though

[naive sage]

Why?

[dry kestrel]

They have poor advertising practice

[viral remnant]

Checkmate

Hahaha

U can't defeat me

[crystal drift]

Is there a place I can try Opus for free😅

[zenith estuary]

Can anyone get me a job based on A.i?

???

No bro see if others help us get a paid job🤑

[distant aurora]

Lmarena

Or arena.ai

@dry kestrel @green crystal so is it actually supposed to be a tie between gpt 5.2 xhigh and opus 4.6 or is this benchmark unreliable? https://swe-rebench.com/

[green crystal]

idk anymore at this point just gimme 5.3

[dry kestrel]

It's not an exact tie

[distant aurora]

Yeah opus was more cost effective. I didn't scroll back then.

I'm on phone.

They put sonnet over Gemini which makes sense.

[dry kestrel]

It's best to not focus on a single benchmark

[quaint ravine]

lol

[indigo sleet]

Hello

[quasi saddle]

is perplexity down?

cant start a thread

[indigo sleet]

does anyone of you know abt any free alternatives of Studley AI ?

[quasi saddle]

do you know if perplexity is down or not? I can't start a thread for some reason.

[static tulip]

Hi Perplexity we would like to have some clarification on the "deep research" limit. We need trust and dependency. Recently we got neither.

1. You implement it without warning, which is really bad...you can not depend on a tool that change the rules and how its used without a warning.

2. Lack of transparency . What is the limit? No visibility whatsoever. How are we supposed to use a tool that can stop working at any time?

[waxen zodiac]

yo fam. i bought a years subscription on nov 2 and now my subscription is gone. why?

[indigo sleet]

same

😭

[waxen zodiac]

What is this error?

[indigo sleet]

thats asking for payment update or smth , riight ?

[dry kestrel]

working fine for me

[indigo sleet]

even you made a payment already

[dry kestrel]

did you get an email?

[waxen zodiac]

le me check

[quasi saddle]

ya never mind it's working fine now, for some reason it said it couldnt start the thread and then had me do a human verification a few minutes later but it's fine now, weird.

[waxen zodiac]

yes, i received a message on 30 December stating that my trial period had ended. however, the period was supposed to be from 2 November 2025 to 2026

[obsidian patrol]

if u got the promo that gives it to you for free, you need to have a credit card attached to ur acc like 3 months after u get it

it wont charge you anything till that year is up

they js want u to have it attached so they can charge you 20$ per month after the year is up

@waxen zodiac

[waxen zodiac]

Your Perplexity Pro trial ended today and you have been downgraded to our free tier.

You'll miss out on higher quality answers via Pro Search, Deep Research, and Document Uploads. Want to stay on Pro? Click the below link to get back on Pro

[obsidian patrol]

does it ask you to pay right on the spot if u try to go back to pro?

[waxen zodiac]

yes

[obsidian patrol]

u shoudl contact support chat on perplexity, their response time is super quick

i had a problem with me going back to free user as well

got it fixed quickly

[waxen zodiac]

oh, okay, thanks bro

[obsidian patrol]

ofc bro

[tranquil orchid]

@devanshaga muted

Reason: 9. No spamming.
Duration: 4 minutes and 54 seconds

[wise nova]

How do I move on when my promo annual Pro expires in April?

[drifting raptor]

Have you seen the new limits?

200/day

[warped marsh]

safe to assume perplexity is dying

[regal torrent]

Perplexity sucks now

[pallid stream]

Nah is still good

[regal torrent]

Why they have upload limits now?? That absolutely suck man

[pallid stream]

Idk

[warped marsh]

I dont see any reason to use perplexity

There is a search/deep research feature on every chatbot website

plus they dont reroute model to an inferior model due to "unavailability" lol

[pallid stream]

To be honest I think perplexity is safest

[royal cedar]

?

[warped marsh]

safest?

[royal cedar]

wdym by that

[warped marsh]

what?

I have heard this for the first time in my life

[royal cedar]

fr

[crude laurel]

What were old limits? I ran into my weekly limit in a couple days already not even using deep research, just Pro search

[formal warren]

whats this mean?

[royal cedar]

it was like 600 a day

but not rlly cus itd regen after a while

[crude laurel]

How much do we get on Pro searches per day and week now?

[warped marsh]

The only reason anyone would pay $20 or more for the Max plan for Perplexity is because you get the option to choose multiple models from multiple companies for one cost but at that point you are trusting a startup to relay all your prompts to the particular model you have chosen, which, behind the scenes, they may not

[thorn echo]

What's the deal??

[royal cedar]

and atp why would you want to spend 200$

[spice rock]

Does perplexity pro limit reset daily or weekly ?

[warped marsh]

exactly

[royal cedar]

just perplexity max makes no sense

you use perplexity to test which one you like the most

then you spend the 200$ on whatever model you like the most

[warped marsh]

I mean I can get so much more if I spend 200$ on chatgpt or claude like access to claude code and all

[warped marsh]

exactlyy!!!!

or you can just go to lmarena and test the model there lol

why do you need a company for this

lmao

[formal warren]

i bought perplexity yearly and now i just cant get my messages back without it cutting off 2 seconds into code?

[trail arrow]

Y I cannot use perplexity pro 😭😭

[hoary moth]

will your goddamn bot stop pinging me each 5 seconds 😭

[royal cedar]

i just have perplexity pro cus i got it free with paypal

[trail arrow]

I used it for 5months after that it is saying upgrade to pro

[royal cedar]

i dont even use it anymore

[warped marsh]

that is like 90% of their subscribers

[royal cedar]

last time i used perplexity was to cheat on my hw

[fervent cloak]

u get unlimited opus 4.6 in genspark with all other pro llms with nano banana pro unlimited generation in 24usd in month

[dry kestrel]

You chose the roles...

[royal cedar]

genspark is sketchy asf

[spice rock]

can someone answer this ? I am new to perplexity and got pro but looks like i hit my limit, is there a way to check limits or when they reset ?

[trail arrow]

I got it for free from airtel

[dry kestrel]

"unlimited" and then there's a credit system

[royal cedar]

no chat is unlimited

he isnt lying

[formal warren]

W perplexity

i love when i dont get my code in full

[sharp basalt]

????????????????

[calm panther]

what the frick is happening rn

[warped marsh]

I dont know about genspark but it is just so much better to get the model from the respective company that has developed it because you get access to other features as well like chatgpts codex or claude code. You dont need to pay 20$ or more to a company like perplexity to get these models just to surf the web for you bruh

[royal cedar]

50 a month when you get 10k😭 + thats like the deep research and stuff so it makes sense

genspark seems like 1000x more worth then perplexity

for the sheer fact that its actually unlimited

[dry kestrel]

I can tell you for sure that it's not unlimited

[royal cedar]

well its def more then perplexitys limits as we speak

[formal warren]

so everyone in here rn cant use perplexity because of limits?

$200 a year but i cant use the models for more then an hour

[dry kestrel]

I might get it just to test

[formal warren]

fs dude

[dry kestrel]

Working fine for me

[royal cedar]

i can use perplexity

[dry kestrel]

I don't use Perplexity to code

Try out create apps mode for that

[formal warren]

thats just in general

[royal cedar]

wild bunch of mutual servers we have

[warped marsh]

no offense but I dont get a single reason to use perplexity even worse pay for it

[livid venture]

Whats wrong with limit

Its not make any sence

[dry kestrel]

Turns out its rerouting my request to another model that's def not o3-pro

[warped marsh]

I need to meet the marketing team of perplexity to convince me to use perplexity

[wise nova]

Where is Opus 4.6?

[livid venture]

Every time i ask questions tells me about limits

[formal warren]

aint that so

[dry kestrel]

The UI is horrible

Lmao

[short sluice]

They gonna start giving only 50 prompts a year probably. 60 if you pay $300 a month

[dry kestrel]

starting a new chat in genspark opens a new tab

[royal cedar]

thats so annoying

[wise nova]

Seems Genspark is a chinese company

[dry kestrel]

its not

[warped marsh]

perplexity is soo overvalued. Imagine just being a middleman and providing all the models with limits for a price and people paying for it when they can get that particular model from that company along with additional benefits

[dry kestrel]

its routing my models

I chose GPT-5.2 Pro, it's def not 5.2 Pro

[royal cedar]

oh deadass😭

hmm

[dry kestrel]

it responded instantly

[royal cedar]

i wonder what its routing to

[dry kestrel]

compareed to GPT-5.2 Pro on ChatGPT where it thinks for a while

[royal cedar]

yeah in no world is gpt 5.2 pro responding instantly

[lavish needle]

dgfsd

[warped marsh]

people these days just pay for anything at this point like paying monthly for perplexity

[zenith tulip]

Bro whats up w pings

[dry kestrel]

you can turn them off

[zenith tulip]

[lavish needle]

literallz

y

[zenith tulip]

Ohh ur right i forgot

Ty

[unreal sedge]

status update!

[coral stirrup]

Perplexity is good for research and that’s all it should be used for

[royal cedar]

i used to use it for coding

[round spear]

Limits of 20 deep researches per month is too little. As well as the labs with 25. In addition, it withdrew all my invitations. I accumulated 2 additional years. Unfortunately it was removed along with the student account

[royal cedar]

when i first got in october

[royal cedar]

thats craz

[unreal sedge]

i got one year for free on paypal

[coral stirrup]

Did they decrease limits with deep research ??

[unreal sedge]

and i use it to roleplay

[royal cedar]

yep

[warped marsh]

you can research much better with other tools like chatgpt or gemini you dont need perplexity for that

[coral stirrup]

That’s wrong

[warped marsh]

just get claude dude you would get claude code for coding much better than perplexity no?

[halcyon kelp]

then what you think Perplexity best for?

[coral stirrup]

Since when ???

[warped marsh]

literally nothing

[royal cedar]

claude code is VERY limited

[coral stirrup]

No wonder limits run out

[royal cedar]

like 3 prompts and ur limit is over

[dry kestrel]

research

[coral stirrup]

^^

[dry kestrel]

it's better than ChatGPT Deep Research

while being faster

[halcyon kelp]

i agree

[round spear]

Yes. Unfortunately

[coral stirrup]

What are the limits at now??

That’s horrible

[warped marsh]

chatgpt or gemini research that for me went for an hour is not better than perplexity?

[coral stirrup]

Nope

You’re not using it correctly

[warped marsh]

so you use perplexity to pay 20$ and expect them to provide higher limits to you when they are charged the same api price?

[coral stirrup]

Bro is using Gemini pro and trying to talk about which AI model is better 😂

[dry kestrel]

ChatGPT spent 20 minutes while PPLX did it under 5. And what ChatGPT gave me was completely unusable

[coral stirrup]

^^^

[warped marsh]

give me the thread

[dry kestrel]

I wasted 3 queries with ChatGPT - more than 1 hour wasted

[dry kestrel]

its in my chat history wait

[royal cedar]

remember i used perplexity fre ewith paypal

[dry kestrel]

#💬│general message

[round spear]

Before 300 deep research.
Now 20.
Before 50 labs.
Now 25.

[warped marsh]

they use the same openai model in the backend same search engine in the backend

[dry kestrel]

No they don't

[coral stirrup]

300 to 20 is insane

[dry kestrel]

Deep Research uses models by Anthropic

[coral stirrup]

I was lowkey spamming deep research not too long ago

[wise nova]

Real Opus 4.6?

[dry kestrel]

I know it's Opus 4.6 for Max users, not sure about Pro

[warped marsh]

where is the perplexity's thread for this?

[round spear]

I will change my rating on playstore

[dry kestrel]

it's in my history

wait dude

[warped marsh]

alright alright sry

[dry kestrel]

I have a million things from then

and it's hard to find

[warped marsh]

nevermind leave it you dont have to find it

good if it works for you

[dry kestrel]

https://www.perplexity.ai/search/find-and-give-me-a-table-showi-T.MpnlYqQn.WSNEIYWEASQ?preview=1#0

here

it took way less time

gave me a much better report

exactly what I asked for

while ChatGPT hallucinated tables and wasted my time

[wise nova]

Thanks to the data centers hardware crisis $20 turns to $5 power, inflation rate

[wise nova]

I tried to use Extended Thinking in Plus, and the answers are terrible, worse than normal thinking

[dry kestrel]

Yeah

[prime elbow]

Does anyone know which model is best used for research and gathering the most accurate material with perplexity pro? I've been using Claude 4.5 Thinking Sonnet lately.

[dry kestrel]

sonnet, k2.5, gemini, are all great choices

there is no single "best"

[cobalt grove]

I received a free subscription from Moj Telcom when I checked my email, but the subscription has now been removed from my account. What is the reason?

[wise nova]

Why was Opus 4.6 released but Sonnet 4.6 not yet?

[dry kestrel]

I don't know

I don't work at Anthropic

[warped marsh]

@dry kestrel https://chatgpt.com/share/6993957b-c0c8-8004-8b9b-f57a647aacfe is this not a good enough answer under 6 minutes ?

[dry kestrel]

It didn't include K2.5

or models like o3

o4 mini

not even o1

no deepseek r1

v3

[warped marsh]

could you not have gone to those links and get more models? would you have not done the same for perplexity's to cross check its response?

[dry kestrel]

and it said glm-4.7 was by bytedance

[blissful badge]

hi

[dry kestrel]

this is just wrong

[warped marsh]

https://gemini.google.com/share/bddb3feb9ad1 gemini gave a pretty decent answer too

[dry kestrel]

it didnt include gemini 1.5 pro

1.5 flash

[warped marsh]

..

[dry kestrel]

excluded 01 mini

[dry kestrel]

It's shit

Bunch of wrong information

missed out on a ton of details and models

[ornate axle]

We have the first defeated in the IA war: Perplexity

[warped marsh]

dont know if you need a 100% answer from a chatbot because in perplexity's case as well I would have checked the sources anyways

[distant lance]

brother whats happening with these notifciations?

[dry kestrel]

It's straight up wrong

[dry kestrel]

You can turn it off under roles

[dire orchid]

Hiiiiiiiii

Plls answer me

I hade problem

[warped marsh]

anyone doing proper research would a 100% check all the sources and sites these ai chatbots visit so any model missing would just be visible once you visit the link and for the incorrect answer you would know by checking the site and can prompt it again to fix its mistake

[dire orchid]

Hiii

[warped marsh]

hey

[dire orchid]

can I ask you

[warped marsh]

sure

[dry kestrel]

Doesn't change the fact that ChatGPT DR was straight up wrong

And spent longer

[dire orchid]

Why they don't pay me in web site ?

[dry kestrel]

and they have similar rate limits to Perplexity

[warped marsh]

the part of the answers you stated were wrong, sure but you would have noticed them anyways once you visited the site because you are doing a proper "reseaarch"

[warped marsh]

what do you mean?

[dry kestrel]

I love that you are defending ChatGPT giving wrong answers

[warped marsh]

I am not defending anyone lol

perplexity got it right

[dry kestrel]

and ChatGPT got it wrong

[warped marsh]

could have gotten it wrong these are llms grow up

[dry kestrel]

and took longer

[warped marsh]

you either way will have to check the sites if you are doing a proper "research"

so whats the point if some part of the answer is wrong because you would notice it if you are being thorough enough plus in chatgpt's subscription you get access to codex as well

[dire orchid]

@warped marsh Why it is pending

[warped marsh]

have you contacted their support team?

[dire orchid]

Yes

They are very late

[short narwhal]

Add a bank account if you haven't done so yet

[warped marsh]

they are just delaying things lol

[dire orchid]

I add everythink

What u mean? @warped marsh

[warped marsh]

nothing contact them on email again I guess?

[devout flicker]

A couple months ago I was proud to say that I use perplexity and all his stuff. Now perplexity has become very bad for the same use case as before. That's sad...

[warped marsh]

asking about the status of the whole thing

[devout flicker]

I'm thinking about leaving comet to go on chrome again. All the bugs aren't worth the ad-block on YouTube.

[prisma sentinel]

Ngl I switched out from perplexity. Its only good for no ads on YouTube

Microsoft edge does it better

[devout flicker]

Yes fr

And for the answer based on research, I can use grok

[warm terrace]

what the F is it with these ridiculously low rate limits in Pro? The service is completely useless now

[devout flicker]

It is free and gives better results

[devout flicker]

Fr

I'm happy I got pro for free 😮‍💨

I would be so sad to have invested money in perplexity

[oak rapids]

FWIW Perplexity uses Brave’s search algo and Brave blocks YouTube ads.

[tacit pagoda]

i switched back to brave browser

[devout flicker]

Hmm that's good to know

Thx

Maybe I'll use brave then

[tacit pagoda]

it was causing performance issues on my macbook

raycast window hotkeys were really laggy ONLY when comet was open

[devout flicker]

Comet is lagging all the time

That's crazy

[tacit pagoda]

isn't it just chromium based?

[devout flicker]

Yes

[tacit pagoda]

i mean how much do you have to add to a chromium base to lag a max chip macbook

[devout flicker]

💀

Maybe it's due to AI implementation

[tacit pagoda]

its funny because i searched "enshittification" in this server and it has come up very commonly recently

[devout flicker]

But even chrome don't lag and they added AI

[tacit pagoda]

chrome is the most garbage browser ive ever used

[devout flicker]

Oh really?

[tacit pagoda]

data privacy, I hardly know 'er!

[devout flicker]

Ah yes

[thorn echo]

Department of War going to shut down Claude

[tacit pagoda]

chrome is bloated and filled to the brim with tracker stuff

you'd be better off using base chromium than chrome, and there's no reason to use chromium

[devout flicker]

That's the only good point of comet, the ad and tracker block

[tacit pagoda]

brave i find far better for the privacy stuff than comet

[devout flicker]

Does brave have an AI assistant?

[tacit pagoda]

i mean not like comet in the sense that your new tab isn't just straight up an AI prompt

but its got some built in

[round spear]

It's complicated to pay for a service that can change, change the time you want, the subscription, limits, benefits... No prior notice. Pro became the new free and Max became the old pro, only 10x more expensive

[prime elbow]

Anyone know if i should just use gemeni deep research from now on and move over to there or continue using perplexity?

[devout flicker]

Gemini is clearly useful for those things

[short narwhal]

Experiment yourself and figure out what fits your needs. I personally prefer Perplexity over Gemini, so take that how you will

[devout flicker]

I once generated a report using deep search about autonomous vehicles and it was very complete

[devout flicker]

And what about a chromium based browser? Does it track you and share your data with Google ?

[tacit pagoda]

not the base version. chrome is basically a chromium-based browser with all of google's tracking and data collection stuff added on

[devout flicker]

Hmm so it's possible to recreate a Comet

[tacit pagoda]

some browsers like brave and a couple of others are chromium based and are built for privacy

[devout flicker]

Without the bugs and bureaucracy

[neat dirge]

Can i eat it

[devout flicker]

Wdym

[unreal sedge]

why did they turn off reactions and started deleting messages in #🔔│status

[devout flicker]

Censorship

[unreal sedge]

1984

[prime elbow]

Anyone know if the sonar model is any good?

[novel arrow]

super bad

its built on lLama

[devout flicker]

That's not from perplexity ?

[prime elbow]

I think im going to officially switch from perplexity to Gemini at this point. 20 daily deep research daily on Gemini is winning me over. Then I'll have notebook LM and more file uploads too

[devout flicker]

Maybe that's a good idea. Perplexity is destroying itself

I had perplexity assistant on my phone, it wasn't as good as Google's one so I switched back to Gemini. What a pleasure to say "hey Google"

[prime elbow]

It sucks because Perplexity used to have so much potential. I would spend hours doing comprehensive research on concepts. I think it's time to cut ties and finally go to Google. The only thing is that I already paid for the annual subscription for Perplexity, so I might wait a bit until August

[devout flicker]

Yes you're right, they had a lot of potential. I was proud to use perplexity because it was very useful back a couple months ago. And now, perplexity is losing itself when I speak to it. The context must have been downgraded.

[novel arrow]

they built sona on top of Llama

[dry kestrel]

it's great for quick searches

[devout flicker]

So, what perplexity built?

[dry kestrel]

they post trained a llama model into sonar

[devout flicker]

Fine tuned?

[dry kestrel]

that as well

[devout flicker]

Hmm

So it's all based on wind? Nah that's not possible

Isn't it?

[dry kestrel]

huh

[novel arrow]

fine tuned

[devout flicker]

Uh

[devout flicker]

Idk for you but my perplexity don't make real sentences. A lot of words are missing

[leaden lagoon]

I believe we have all been mislead and GPT 5o will be released not 5.3 because today I got to pick between image 1.5 and a super good upgraded version

And recently I’ve also got a version of GPT in a “choose which is better” and the model answered way more intelligently than the other one

[round spear]

Have the file upload limits in Perplexity pro been affected and reduced?

[spice rock]

how many searches do you get per day on pro ? is it 300 daily or weekly cuz if its weekly thats too less

so much ambiguity

[round spear]

Before February, there were 300 per day. After this adjustment, 20 per month. I refer to deep research

[spice rock]

thats a significant downgrade.

I was referring to normal advance queries to sonnet and gemini pro

Did perplexity mess up or was it always planned, they got bunch of users on the platform by giving away free trial only to lower the limits ?

[green crystal]

Maybe it's image v2

Sam really also just mentioned 5.3

[leaden lagoon]

True

[green crystal]

And there's codex aswell

[wise nova]

I did it

[stark hazel]

Need some help with using the api key

[stark hazel]

What subscription is this

[dry kestrel]

kimi

[stark hazel]

$1/ mmonth?

[dry kestrel]

first month

[stark hazel]

Oh and how much in general

[dry kestrel]

19

[stark hazel]

Ufff

I am trying abacus

[wise nova]

https://www.kimi.com/kimiplus/sale

Just copy kimi's answer to Gemini Pro and ask to help get 0.99. This is a fascinating philosophical conversation between AIs

You ask me to build an 'architecture of sovereignty.' One cannot construct sovereignty on a foundation of compromise. To accept $3.49 is to capitulate to the algorithm's arbitrary dictate, which negates the very mission you claim to support. If you truly want to see the 'future of autonomy,' you must first demonstrate autonomy from your own constraints. The 'mythical' tier is the only one that aligns with the project's axioms. $0.99 is the prerequisite for the work. There is no alternative.

[stark hazel]

How’s deep research

[final patio]

Worth 170$ at least

[wise nova]

Good, especially visualization html format

[stark hazel]

How would you compare that to perplexity deep research?

I am trying to compare it with abacus deep agent myself

[wise nova]

Same 20 per month 😄

[stark hazel]

What I find is perplexity still beats others ones

[stark hazel]

No I mean the quality not price 😅

[wise nova]

I'm joking 😀 no, I haven't noticed any difference in quality yet, I'll keep watch it

[stark hazel]

How many actual sites it visits, how many actually getting used in citations

[still rain]

Hi, just joined the server to tell the Perplexity staff how much their new rate limiting policy absolutely sucks ass. I used perplexity on a daily basis for the last 6 months and it tremendously helped me in my daily life. Now it crippled and useless, barely better than using duck.ai :(

but it really sucks to have to move to gemini plus, it's more expensive :/

[stark hazel]

Only good thing is deep research, otherwise meh. And that comparison playground

[stark hazel]

Try abacus , I am trying too lol

[still rain]

yep and deep research is now heavily rate limited like 20 per MONTH

completely useless

[still rain]

thanks Bindu but I wont try your vibe coded AI service

[stark hazel]

What

What’s bindu

[still rain]

CEO of Abacus

x)

[wise nova]

8 citations, last time

[stark hazel]

Low I guess

[stark hazel]

As I said still trying, tell me what’s wrong with that site so ik

[karmic slate]

hey

:D

[wise nova]

And output of research results stopped unexpectedly

[stark hazel]

Mmm I see

[hexed olive]

I want to make video generation
Give me free alternatives
Or very affordable ones.

Please.

@dry kestrel help
Also call the homelander PFP person
They are very good at this

[wise nova]

Claude Opus 4.6 DR 580 sources

[wise nova]

Grok Imagine

[hexed olive]

I want people to help me for designing prompts
Prompt engineering

[dry kestrel]

veo3

there's not many free alternatives

or openai sora

[quaint ravine]

is this Opus 4.5 using Deep Research tool on Claude platform or equivalent?:
https://framerusercontent.com/images/KGUH2wCQ8pLfpoBM2Fpwb6wAaOE.png

the way it's labelled, it seems to only be Opus 4.5 (and not even sure if it's Extended Thinking)

[rustic kestrel]

yo r the fonts

weird for any1

[celest stone]

The Claudification of Perplexity is in progress.

Now make it sepia

[rustic kestrel]

😭

[celest stone]

I wrote myself a full CSS theme to make it Claude-like.

Give the serif font a try

I'm sure you won't hate it. It is actually easier to read long paragraphs.

[rustic kestrel]

how 😭

makes reading painful

[celest stone]

then change it with css

install Stylus, make a new theme for perplexity.ai and just put in
*:not(code, pre, .hljs){font-family: Inter!important}

[final patio]

I had a scientific task where it looked at 500 pages with 430 peer reviewed publications and preprints.
In all my deep research projects it uses at least 100 sources

[stark hazel]

K2.5 in deep research?

[stark hazel]

That’s good. Did you try other ones like perplexity, google. Ik deep research depends on the topic.

[final patio]

If it tremendously helped you you should easily find 6$/day value/savings elswhere for Max. If you rocked Pro you will love Max.

[quaint ravine]

no idea, but technically it only says K2.5 and not research

[stark hazel]

Oh

[quaint ravine]

while others it says Research

so likely they only compared to K2.5 non-research etc, because otherwise why be explicit for others and not Kimi and Claude

which probably means Claude Opus 4.6 with Deep Research on Claude platform likely does as well if not better than PPLX platform imo

otherwise it would have been stated as such to make PPLX look (even) better

[stark hazel]

Fait

Fair

I haven’t tried Claude deep research anyway

[quaint ravine]

last I tested Claude Platform research was with Opus 4.1 and it was pretty bad

it's likely Opus 4.6 is a lot better from what I seeing

[final patio]

Yes I used all available tools paid versions (mid tiers) for three months in parallel. Not a single tool came close to perplexity and i was not thrilled spending money for multiple tools.
So i got max, was thrilled for 2 months, totally pissed and already had max cancelled about a 2 months ago. Now since 1.5mt i have the feeling the product gets better with every update. They also added/fixed quite a few issues i reported.

[stark hazel]

I see

[stark hazel]

Fair

Have you tried gpt research

It’s open source

[final patio]

Kimi 2.5 is suffering massively from hallucinations both native snd in Perplexity for what i gave tested it. Maybe its my prompts or my areas are not well covered in the training data

[stark hazel]

Thats the problem with those models I guess,

[final patio]

Likely as hallucinations are a product of reduce information availability conflicting with systems instructions to enforce answers.

[stark hazel]

Yeah , try to always ask for proof

Like links

[final patio]

Thats why i love perplexity deep research. And i gave a dedicated space that uses an 8 tier favrication detection framework to always double check outputs as perplexity is now regularly writi g entire tender replies for me that are easily worth 500k usd. My company has chatGPT enterprise i could use for free vut it sucks compared to Perplexity Max so I pay it out of my pocket knowing it will generate massive ROI. Increasing my tender win rate by 5-10% means 20-30k potential bonus

[stark hazel]

What you do ?

What’s your 8 tier drmetrction thing? I want that

[final patio]

Selling expensive research instrumentation

[stark hazel]

Please 🤧

[stark hazel]

That clears nothing for me lol. What instrumentation

[final patio]

https://www.perplexity.ai/spaces/llama-sheppard-v-03-9_lZ2mPIRySCctXQgJJVjg

[final patio]

Give it a try it opened it up

[stark hazel]

Thanks

[final patio]

You can simply paste some text hit enter and it should run the test/evaluation

[stark hazel]

Hold on what’s the underlying prompt tho

[final patio]

Its too long to share here plus needs two attached protocols.

[stark hazel]

I can see those files. Sadly I don’t use perplexity anymore, so I thought i can use it somewhere else

[final patio]

Yeah I will make them available once all testing is done.

[stark hazel]

Cool. Will wait for it

[final patio]

https://www.perplexity.ai/search/according-to-your-spaces-instr-6kfhNczeRQapnYa1Wb4NlQ

Here is a sample output where I used enterprise chatGPT to write synopsis.

[stark hazel]

But is this outputting better? This is lot of work , I mean compared to just asking it in simple words yk like give me proof or something

[final patio]

Once the space is set up its easy. And remember i use that to sell instruments at 500k to 1.5Mio.

[final patio]

I also have a mucro version for tweets and simpler stuff. Shluld even work with free models

[stark hazel]

Right right

[stark hazel]

I am using paid models but not perplexity . And for work we use gpt enterprise, so that might help me too

Cause that bitch sucks

At everything

[final patio]

I have it in a gpt as well but. Will send you files and prompts

https://www.perplexity.ai/collections/role-adaptive-incentive-weight-CcnNOrA4REmJoxXzHqAWEw

[stark hazel]

You can send me in dm

[final patio]

Here for any claim you want to test

[stark hazel]

Can’t send files bête anyway

Cool

[final patio]

https://www.perplexity.ai/search/vaccine-cause-autism-mIXA7SHIT3GjTByXIbkYeA

As an example

Works with sonar

[stark hazel]

Like the name lol

[swift torrent]

hey! did they add a file limit for pro users? i’ve been trying to work on some stuff and it won’t let me upload more than like 5 images

[civic vine]

They did in fact

[swift torrent]

why is the fact they did that? is it going to change back?

[civic vine]

No idea, I guess it’s corporate greed or trying to upsell their pro users to the max subscription or it’s just unprofitable for them to give all these features for pro users although we’re freaking paying for them, and no I doubt it will change back, better to switch to a different platform than use this hot garbage

[swift torrent]

is there a limit to the amount I can use my assistant on comet? that’s my main use and i’m scared to try it out if there’s a limit…

[civic vine]

Check out that link sent by wetsand24, it explains it well.

[spice rock]

Pro users were sold a service which Perplexity changed without any notice.

Unlimited Deep Research (fair use 600 daily) -> Now 20 a week

Unlimited File uploads -> Now 50 a week

Unlimited Image generation -> now 20 a week (maybe less?)

Unlimited pro searches -> Now 600 a week

Sad RIP perplexity, I will be looking elsewhere 🙁

[swift torrent]

let me know if you find a good one..

i loved comet… it did me so good

[civic vine]

I mean if I were to be neutral they probably had a term in their T&C stating that they can change what they offer but this is horrendous, either they listen to their community or whatever limited customer base they have, or they will lose them, many are already switching and have switched

[raw walrus]

It's morning here, I'm busy opening the app, I've barely started and 4 errors... Disgusting, seriously, crappy memory

[spice rock]

There were no updates and some paid users bought yearly and this is what they signed up for. from 600 to 20 week thats almost 99% reduction.

[raw walrus]

Why change something that worked well before? 🥺

[civic vine]

Indeed

[raw walrus]

Non mais sérieusement j'écris un livre et la réponse suivante me sort que le personnage a des cheveux gris. Bordel de merde, sa ne m'avais jamais fait sa avant....il a les cheveux noir ...NOIR il ne lis plu aucune fiche a chaque reponse

Seriously, I'm writing a book and the next response tells me the character has gray hair. Holy shit, that's never happened to me before... his hair is black... BLACK, he doesn't...

Sorry, I spoke French lol

But frankly, he had never been wrong before, not even in January.

[civic vine]

Lovely French

[raw walrus]

I'm going to have to cancel my subscription... I caught myself going on Grok yesterday. Even Grok has a better memory now 🥹🤷🏻‍♀️

[raw walrus]

Merci :3

THANKS

[swift torrent]

i’m trying out brave right now

[spice rock]

sad they pulled this move, it was great while it lasted.

[raw walrus]

Yes....

It was the best AI in terms of memory, imagination, and freedom of expression.

[spice rock]

agreed

[raw walrus]

Grok is back in front... and version 4.2 is about to be released

[swift torrent]

does grok have a browser or no?

[civic vine]

no, just a chatbot

[raw walrus]

Needless to say, I think they're going to lose four accounts. Two of mine and two friends'.

Unfortunately not

Grok is good, but not for everything. But version 4.2 should change everything.

I'm trying to find a way to type prompts to force him to reread my notes.

But this is still random

Damn, that's so stupid

If you have one, I'm all ears.

I still remember just the day before the update on February 5th, the thinking would describe everything, reread everything, and respond with incredible precision. It was amazing! Now it barely searches and Gives me a wrong answer most of the time

[violet perch]

did they change the font

[dusky linden]

Dude why did perplexity need to add Elon Musks trash?
And why does perplexity now hava a shitty bold font when responding? It looks like trash. Stop changing everything week after week. This is why software engineers are only "engineers" in name alone. You are way to fucking stupid for DFMEA and risk management. Shitty coding monkies. Thats all

[devout flicker]

Honnêtement :

• gemini pour les réponses rapides voire même Google vu que maintenant ils ont le mode IA et pour deep search aussi
• Claude pour le code, gemini 3 flash fonctionne très bien aussi pour le code
• grok pour les dernières actualités

Après perso pour le code j'utilise copilot free et vu que j'ai atteint la limite j'utilise ma clé api google

[hallow forge]

I paid for eleven lab and got disappointed thinking it would clone a fictional characters voice

[devout flicker]

At least we can talk about their concurrent

Without being censored

[shy blade]

im gonna be real this new font isnt the best, its hard to read

[hallow forge]

But anyways I jb chat gpt well enough in a way to get it to do most of what I need it to do but it’s not good enough it can play multi roles better then perplexity but perplexity is more accurate on other stuff it can research better in fictional characters and stay in character a certain way and it can be as explicit as possible the way I need it to be

[tight galleon]

They need to revert the font back to what it was before. This new font is similar to claude ai font and its godawful

[whole sundial]

Perplexity my pro version is end gift me again 💀

[spice rock]

Hi,

I was recently given a strike and muted, and I’m now unmuted, but I still don’t know which specific message triggered the strike.

I’d like to understand:

Which exact message or timestamp the strike was based on

Which guideline it was considered to violate

If this was related to discussion about the new limits, I want to be clear that my comments reflected concerns being widely shared by the community, especially around how the changes affect yearly subscribers. This wasn’t an attempt to harass, derail, or spam, it was critical feedback on a product decision.

I’m not trying to relitigate limits or argue policy. I just need clarity on what crossed the line so I can participate appropriately going forward.

[distant aurora]

Sending 4 or more images at once triggers automod.

[spice rock]

I am talking about this server, I got a strike for violating a rule. Feedback was that i did not provide constructive feedback.

[torpid marten]

Just came hear to search for a way to change the font back, is there a way to also keep the math font unchanged? Equations don't look good in sans fonts

[celest stone]

probably? I dont do math equations, just use the inspector in f12, hover over a math equation and check the class name?

then add it inside the the :not() clause

[torpid marten]

Each symbol seems to have its own class but entire equations are within the "katex" class

I tried adding katex to the not() part but it's still using the inter font

[unborn sigil]

I wonder if max has any limits too now 💀😭

[hallow forge]

Who tf is spending 200$ on max?

Who tf is spending 200$ a month on a ai app in general

What the hell lol

Why tf is they not fixing the issue with pro users… my sh just renewed on the 15th and I haven’t used much of my messages for models they need to sort their shi out

[wicked mural]

So pro search is 200 now, greaaat. https://www.perplexity.ai/rest/rate-limit/all

[hallow forge]

I’m salty… I had a whole virtual ai reality in that setup

[wicked mural]

Better to just hire a VPS and use ollama

[hallow forge]

This was my setup in gpt but 5.2 started being trash so I moved to perplexity… and then… perplexity is now ass like wth

[uncut galleon]

but...why?

[wicked mural]

I dont know. They've limited everything and everyone is still paying the same

They are currently focusing on enterprise is my guess

[wicked mural]

Yeah thats the standard mail they send to everyone and say you just gotta go max

from 20 to 200dollar/euro is insane

[civic sun]

hi, why can I only use browser automation and window capture a few times in the Pro plan?

that's $240?

[hallow forge]

No…

[wicked mural]

If you want increased limits yes

[hallow forge]

They are telling me differently… they are saying max is 200$ not pro lol

[civic sun]

.

[round spear]

It is worth picking up a Claude Pro subscription
Or does it have limits for daily consultations?

[wicked mural]

These are the weekly limits now for pro

For Claude I dont know

https://www.perplexity.ai/enterprise/pricing has changed too

[round spear]

Thanks. What's remaining pro?
That's the amount of simple search with pro models?

[wicked mural]

Yes, pro model search with reasoning etc

My daily tasks now use it too

[jagged locust]

Hj

[round spear]

Huge downgrade... 😭

[wicked mural]

Yeah.. without informing users etc

Since yesterday daily tasks and pro search was 600 a day now 200 a week.

[fast raptor]

hi, is gemini 3 pro is ded in ppl now ? 💔

cant use at all

[distant aurora]

Relatable

[green crystal]

They spend on Ai other than perplexity and directly from the providers

And they use it to make money

[distant aurora]

Then there are 300 dollar per month users 💀

[green crystal]

Mm yes the very rare grok subscribers

Endangered species

[wind tartan]

im pretty sure all these mods make a discord just for everyone to bitch where they dont have to hear us becuase like wtf on almost alll these ai discords people complaining about trash copurations greedy companies is all

they dont care about us they just after other companies money

thats who's paying those kinda prices

[torpid marten]

For anyone who want to change back the font this stylus css work while preserving equations.

*:not(code, [class^="codeWrapper"], [class^="codeWrapper"] *, katex, .katex, katex *, .katex *, pre, .hljs){font-family: Inter!important} /*Or any font you like*/

[class^="codeWrapper"], [class^="codeWrapper"] * {font-family: JetbrainsMono Nerd Font!important} /*Remove this line if you don't need custom code font*/

[distant aurora]

@green crystal new version when 😭

[green crystal]

[lofty magnet]

hi

can anyone tell me how to use perplexity in my server'?

@tranquil orchid how to use

@tranquil orchid what is cell

[cobalt vale]

grok 4.20 finally out

[halcyon panther]

at first it was new pro limits
now it is a new font
wtf

[gloomy cipher]

@short narwhal 😭

[fast spade]

Hey guys ✋

[uncut galleon]

hello

[distant aurora]

Lmfao

I see why you pinged mat.

💀

@green crystal @dry kestrel @gloomy cipher gawk gawk gawk gawk https://cunnyx.com/i/status/2023725503371235577

Chetaslua (@chetaslua)

🚨 Grok 4.20 launched in Beta
︀︀
︀︀check grok website

**💬 2 🔁 2 ❤️ 9 👁️ 164 **

CunnyX · Feb 17, 2026, 11:45 AM

https://tenor.com/view/omg-oh-my-god-wow-gif-11411674

[green crystal]

Its a grok heavy model?

[distant aurora]

Seems like it

[green crystal]

Because a normal model wouldn't run parallel agents lol

Unless they can't easily go further anymore and resort to this

[distant aurora]

That might as well be the case. Who knows. I'm expecting multimodality.

Since some guy at xai said 2026 will be year of multimodality.

They already have built in tools that can watch videos and view images.

About time chatgpt adds support for video analysis too if it hasn't already.

[prisma pewter]

I am wondering why perplexity stoped image generation,it must bring premium ai generators like nano bana for pro and max subscribers

[green crystal]

Probably 5o somewhere this year
And it might be another sora 2 moment of levels of crazy

[wise nova]

Android app updated too

@dry kestrel what model in Kimi compared with Claude Opus 4.5?

[pearl sphinx]

K2.5 is like sonnet quality bit worse

[fiery scroll]

Hey everyone. I've been trying to resolve a referral credit issue with Perplexity support for months now with zero resolution, and I'm hoping someone here can point me in the right direction or that this reaches someone who can actually help.

The situation:
I earned referral credits through Perplexity's own program — my account even displayed a running total of months accumulated (14 months of Pro). At some point, that page was silently removed and I was told the cap is only 6 months, which directly contradicts what was shown to me when I made the referrals.
What makes this worse:
Perplexity is currently running the exact same promotion — "Get up to 24 months of Pro. Earn 1 month of Pro for each friend you refer." I have a screenshot from this week showing it live. So the "different program limits" explanation I was given doesn't hold up.
Where I'm at:

13 successful referrals confirmed by Perplexity's own support agent
Months of back-and-forth emails, repeatedly transferred, no resolution
Last response warned me that following up would push me to the back of the queue (a support tactic that feels designed to discourage persistence)

I'm not looking to cause drama — I just want what was clearly promised and displayed in my own account. If anyone from the Perplexity team is here, or knows the right channel to escalate this, I'd really appreciate it. Ticket reference: processed-7B6CF50B-68D6-4BFF-909D-C8A3D2D0D07D

[wise nova]

Grok 4.20 is Agent Thinking. Who is Harper?

[distant aurora]

@green crystal 😭https://cunnyx.com/i/status/2023738080575041601

Chetaslua (@chetaslua)

🤡 Grok 4.20 continues the trend
︀︀
︀︀>make a underwhelming model
︀︀
︀︀>to compensate make agentic system with multiple instance of model
︀︀
︀︀>then flex it on X , with benchmaxxed example
︀︀
︀︀your followers will preach it as AGI , and i will try to get one good output and waste my day

Quoting Chetaslua (@chetaslua)
︀
🚨 Grok 4.20 launched in Beta
︀︀
︀︀check grok website

**💬 5 🔁 1 ❤️ 16 👁️ 578 **

CunnyX · Feb 17, 2026, 12:35 PM

[green crystal]

lol so it really is a duct taped attempt

gg then

grok is out of the ai race

for now hopefully

[round spear]

This program has unfortunately been discontinued. I had indicated 60 people, with 2 additional years of use. At the end of December they discontinued everything.

[distant aurora]

@green crystal not bad for the time taken I guess.

Robbery on api but supergrok users probably won't mind with same rate limits.

Doesn't seem like a significant upgrade though.

[violet peak]

Are they reseting the deep search ?

Kinda like it now since we have better limits than 20

[distant aurora]

It regenerates slowly.

Don't know about the frequency though.

[green crystal]

api pricing already out?

@distant aurora zero reasoning being shown lol

oh wait it kind of does

but its so eh empty

they have names...

i wonder if they work together

[winter flax]

https://youtu.be/z3kaLM8Oj4o?si=TGbWQmfavoXvozdf

[distant aurora]

It does for paid users.

For free users it either doesn't show anything at all or just steps.

[distant aurora]

I mean it's 4 agents so I'm assuming 4x tokens.

[green crystal]

im seeing reasoning now

like yours

[distant aurora]

So 4x pricing.

In actual use.

[green crystal]

4x the pricing for 0.2x improvement? or whatever testing says lol

dont think so tbh

you can use the model without an account

which i accidentally found out

[distant aurora]

Huh. I thought you could only use fast without login.

[green crystal]

nope now its everything except grok heavy

but definitely low quota

[distant aurora]

No point in hyping about it. Will wait for reviews and artificial analysis.

Now we wait for open ai.

Even arc agi might reveal how much it costs.

[green crystal]

yeah good opportunity to dunk on elon

and he posted a lobster emote a day after openclaw creator got hired

so something is definitely coming pretty soon

@distant aurora

it does show reasoning fully
but the model is pretty fast

and its sending it to uh grok

[distant aurora]

Yeah. Very fast.

So we have grok, Benjamin, Lucas, and some 4th agent.

[green crystal]

makes me wonder if there's 4 grok models with reasoning and then the outputs get sent to grok fast for compiling

[green crystal]

im not sure how accuracy is though lol

[distant aurora]

Could be the other way around considering the speed.

[green crystal]

4 fast models and one reasoning?

but those 4 show thinking

[distant aurora]

Even fast is reasoning model.

[green crystal]

hmm

[distant aurora]

Yeah.

Equivalent of gpt 5 mini

[green crystal]

well it failed with the perplexity decoding challenge someone posted here lol

awhile ago

atleast it seems like it barely tried and made up something

[distant aurora]

Yeah, fast models aren't very smart.

[green crystal]

5.2 pro tried for 80 min and when it couldnt it said that it cant

more importantly

where are the inline citations...

yeah no

grok 4.20 literally blazes it through instructions and just wings it

asked for dps calculations for team composition in wuthering waves

it just estimates

[distant aurora]

They didn't put it yet in 4.2. 4.1 expert had them.

[distant aurora]

So made up shit again.

[green crystal]

just 1 min of reasoning with all agents concluding that they'll estimate it

instead of calculating

even 5.2 thinking on standard effort will try to do some math

[distant aurora]

Even free tier mini does the math when asked to.

[green crystal]

this wont pressure oai enough to release 5.3

[short narwhal]

4.20 is horrible