💬│general

[tight prism]

We don't know if Gemini Ultra will be expensive or not since now they have an unlimited token option

[warm cave]

How do you do that?

[agile jay]

If it's pretty unstable, I would assume they are using http requests.

[agile jay]

F12

Then network tab.

And refresh the page.

[warm cave]

Let me grab my laptop

[agile jay]

What do you mean by unlimited token option?

That just sounds like RAG to me.

And not actual context.

[tepid pilot]

It’s websockets

But this is cooked

[tight prism]

For Gemini Ultra but still don't know yet how many tokens they will set and they said if the price is reasonable they will add an upload video option

[tepid pilot]

I didn’t even know that they added a limit

And now I’m alr out

Wtf

[agile jay]

Yep

[tepid pilot]

Genuinely stupid

[agile jay]

I doubt they will. They only allow 4 files currently. Why would they allow videos which are far larger?

[tight prism]

For Gemini ultra

[warm cave]

what dose it look like to you?

[tight prism]

?

[warm cave]

oh, network page, ill try again

[tight prism]

Not files but for video and only one choice for video like picture

[agile jay]

What is the average size of a video?

[tight prism]

Again we have to wait for Gemini ultra next month and see @halcyon coral said so

[agile jay]

And based on the Gemini tease ages ago. It's just a vision model which skips over frames of the video.

[agile jay]

Icelavaman is a discord mod, I doubt they know what the devs are thinking..

[tight prism]

I check google ai studio Gemini pro 1.5 , it can upload video option

[tight prism]

Wait and see and hope Gemini ultra like they promise

[agile jay]

Yep, you upload a video and it uses ffmpeg to extra extract frames every x steps.

[tight prism]

Gemini ultra analyse video

And give review

[tepid pilot]

https://x.com/AravSrinivas/status/1769475725965566167
https://x.com/perplexity_ai/status/1742915781690798290

$73.6 mil run out real quick huh

Aravind Srinivas (@AravSrinivas) on X

We have made the number of daily queries on Claude 3 Opus (the best LLM in the market today) for Perplexity Pro users, unlimited! Enjoy!

Twitter · Mar 17, 2024, 9:27 PM

Perplexity (@perplexity_ai) on X

We are happy to announce that we've raised $73.6 million in Series B funding led by IVP with participation from NVIDIA, NEA, Bessemer, Elad Gil, Jeff Bezos, Nat Friedman, Databricks, Tobi Lutke, Guillermo Rauch, Naval Ravikant, Balaji Srinivasan.

Twitter · Jan 4, 2024, 2:27 PM

[livid mantle]

OH NO NO MORE UNLIMITED

[tight prism]

It’s for Gemini ultra

[tepid pilot]

I get that, just no warning

[tight prism]

I know😔

[agile jay]

I doubt it's for Gemini

Makes no sense.

[livid mantle]

ITS REAL

AAAAAAA

[tight prism]

That’s in January and they could have many projects to deal with

[warm cave]

https://discord.com/channels/1047197230748151888/1230342648036982795

[minor shard]

Gemini Ultra isn't even out?

[agile jay]

They are paying on each request, so why would it make a difference?

[tight prism]

Gemini ultra could be expensive

[tight prism]

Next month

[agile jay]

And?

[tepid pilot]

L

[tight prism]

They are preparing for Gemini ultra released

[minor shard]

No one knows how expensive it will be or if it will even be better than the best models atm. Ppl jumping to conclusions too fast

[tight prism]

It is for Gemini ultra api released next month

[tepid pilot]

I mean, if they aren't gonna do smthn abt it soon imma want a prorated refund on my yearlong sub

[agile jay]

Yep, it should have nothing to do with Gemini.

[tight prism]

They limited opus for Gemini ultra

[livid mantle]

MAYBE GPT-4-Turbo-2024-04-09 SILENT UPDATE SOON IN PERPLEXITY? AND UNLIMITED?

GOODBYE UNLIMITED OPUS

[tight prism]

It should

[tepid pilot]

imo gpt4t and opus no comparison

[agile jay]

Nope, why would you do it a long time before Gemini even comes out...

[tepid pilot]

I can't go back

[agile jay]

Same, guess it's time to jump ship

[tepid pilot]

so at this point you.com gives you half the queries to opus for free that perplexity gives you for $20

[tight prism]

Again they are saving money for gemini ultra

[tepid pilot]

again, idgaf

[tight prism]

Otherwise why would they limited opus?

[minor shard]

Plz stap spamming this lol

[warm cave]

Everyone loves Claude, gemini is a mixed bag

[terse hound]

I can understand a limit, but 10?? wtf lmao That's like less than claude's own website gives you for the subscription

[tight prism]

They limited for Gemini ultra because for the Preparation next month released

[tepid pilot]

so at this point you.com gives you half the queries to opus for free that perplexity gives you for $20
with half the funding too

[tight prism]

Gemini ultra is for creative option

[warm cave]

5 a day, then 600 a day, then 10 a day, that type of progression is that, its really bad planning or its a way to get users and cut the benefits

[agile jay]

Yep, they got +80M in the last funding.

No way it's run out in a month or two.

[tight prism]

They previously limited to 5

[agile jay]

Especially with them capping the context...

[tepid pilot]

yea capping the context too

[warm cave]

5 then 600? and then 10, what is this

[tepid pilot]

was willing to live w it

[agile jay]

Pretty sure you.com didn't even cap the context...

[tight prism]

A lot of people using the $10 off refer code

[warm cave]

they don’t, i confirmed with the discord and tested it myself, gemini and claude

[tepid pilot]

still costs money bruh

[tight prism]

They have many businesses deal

[tepid pilot]

interface is dogshit but I'd rather have to deal with a poopy interface than not have access at all

[warm cave]

same

[severe ingot]

I'd be miffed but I got 6 months for free so im pretty happy chilling with GPT-4 😏

[tepid pilot]

aight bro

[warm cave]

i am a student so its 10$ a month until i graduate

[livid mantle]

🙏

[tight prism]

The refer code is for everyone not for student

[severe ingot]

[warm cave]

yes, true, just saying i got a good deal too

[livid mantle]

POE OPUS vs YOU OPUS?

[warm cave]

You.com bc you get 500 a day

[livid mantle]

okay

[tepid pilot]

screw poe.com

[livid mantle]

we are back to $200/5 opus in perplexity.

[tepid pilot]

quora is poop

[tight prism]

For now let's wait for Gemini ultra release next month before we can complain about their funds

[livid mantle]

i will look at Mr. Arvin's twitter if he said anything.

[warm cave]

You may be happy, and thats great, i just dont realy use gemini 1.5 even tho i have you.com

I just want a reasonable amout of opus

[tight prism]

I'm not happy about the limit for opus but I really want the released for Gemini ultra for creative reason

[livid mantle]

So what I am hoping for, silent update GPT-4-TURBO-2024-04-09 UNLIMITED in perplexity.

If not, I will go back to Gemini Advanced 1.5 Ultra when that gets released.

[tight prism]

Gemini ultra is best creative option compare with opus

[tight prism]

Gemini pro 1.5 api just only released

[livid mantle]

1,000,000 million context window gemini 1.5 pro so u can imagine, the advanced one will be 100,000,000 trillion context window

[warm cave]

guys, if you want you can upvote this: https://discord.com/channels/1047197230748151888/1230342648036982795

[formal parcel]

💯

[formal parcel]

What? I still have 600 I think

[tight prism]

We don't know the tokens size perplexity will Set yet

[warm cave]

are you on web

[livid mantle]

it will be 32k

[warm cave]

i have only 10 a day

[formal parcel]

They already released a paper on AI models having unlimited context length

[livid mantle]

you use paper?

[warm cave]

can you link it? I remember the 10M testing

[tepid pilot]

bruh gemini 1000000 token context is dogshit

[tight prism]

They said depend the price plus they already limited opus now

[tepid pilot]

not real

plus perplexity will never unlock the full context

[formal parcel]

https://arxiv.org/abs/2404.07143

[livid mantle]

on the official opus its 50 messages $20 a month. in perplexity we are back to $200/6 messages

WHY

anyone here using ChatGPT Plus? is GPT 4 unlimited?

[tight prism]

It won't be 10 Million token but we have to wait and see how much they set the tokens

[chilly forge]

Nope where its mentioned about 50 messages ?

On claude Pro

[formal parcel]

I can’t see it on my end? Anyone else confirm about Opus?

[agile jay]

I get the name of the site now.

[tight prism]

A lot of people using refer $10 a month now

[formal parcel]

Yes GPt-4 is unlimited

[chilly forge]

Oh yes

If your conversations are relatively short, you can expect to send at least 45 messages every 5 hours, often more depending on message length, conversation length, and Claude’s current capacity. We will provide a warning when you have 7 messages remaining. Your message limit will reset every 5 hours.

[agile jay]

Their choices really are perplexing

[tepid pilot]

at this point imma run command R+ selfhost

[livid mantle]

Source

[chilly forge]

Yes Captain

[tepid pilot]

W prompt engineering

[chilly forge]

Anyways I was going to cancel Claude Pro but seems like I would need to continue it

[formal parcel]

Any Poe users out there? How is it?

[tepid pilot]

lowk 50 is not enough tho

that's why I liked perplexity

[tight prism]

This is so sudden

[chilly forge]

Opus is hungry asf

[tight prism]

With the limit opus issue

[chilly forge]

I still remember just pasting 2 to 3 images it literally gave up in like 10 messages

[agile jay]

It is 2.5 times more hungry than GPT4Turbo

[warm cave]

Guys, is there an option that you guys like more than you.com, its teh best option i have found so far

[tight prism]

I hope Gemini ultra they promise we get soon

[agile jay]

10 is not 600/2.5

[chilly forge]

Yeah man its vision capabilities are also too much hungry

[tight prism]

I want Gemini ultra

[formal parcel]

Gemini 2 is in training

[chilly forge]

Can you stop spamming it

[warm cave]

240 is enough

[agile jay]

Yep, even 200

[chilly forge]

Anyways the main problem is the lack of communication by the team

[agile jay]

I only use around 25/day on average anyway

[tight prism]

It's not spamming perplexity said they will released once the api released

[agile jay]

Yep, they just ghost

[tepid pilot]

I think I use around 100 queries per day

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[tight prism]

Better than 5

[terse hound]

Get users to subscribe for 600 messages, then change it once they acquire funding/user base. classic rug pull

[tepid pilot]

I know mf

[agile jay]

Yep, I would assume 250/day covers everyone who use it normally.

[formal parcel]

100 is a lot for most people

[warm cave]

100 might be a little too low

[tight prism]

So many ai will be released soon

[chilly forge]

Tbh with you, it was never like this (atleast in their initial 6 months) because I am from very beginning.

Companies do change once they gain a lot of users sadly

[agile jay]

Yep, but you might have one day when you are working all day

[tepid pilot]

100 for opus specifically

[formal parcel]

Give us lower limit but WE WANT OPUS

[tepid pilot]

exactly

[formal parcel]

💯

[tight prism]

Like the Gemini ultra or the gpt-5 or even llama-3

[formal parcel]

LLaMa 3 will be crazy tho

[tight prism]

So many released soon

[tight prism]

I know right?

[tepid pilot]

who cares? I paid for right now

[warm cave]

what makes you think that them cuting usage will make them give unlimited usage for new models, i think its more an indication that expensive models will only have a handfull of usage a day

[tepid pilot]

agreed

[chilly forge]

This should start blowing up on twitter maybe then we could get a word from @blissful raptor or anyone else from the team

[tight prism]

Opus is very expensive

[tight prism]

Opus very expensive

[warm cave]

I know

[tepid pilot]

not my problem

false advertising

[agile jay]

Yep, also it's only 2.5x more expensive.

And only the output.

[tight prism]

Plus everyone using refer code $10, so the $20 is not a lot

[tepid pilot]

I am not

[warm cave]

but just because they cut opus dose not mean that they will be giving unlimited for gemini

[tight prism]

You don't know the refer code

[tepid pilot]

@tight prism been yapping absolutely nothing for too long

[tepid pilot]

I didn't when I subscribed

[chilly forge]

Exactly lmao !

[tepid pilot]

now I do

[chilly forge]

They have true 200k context

Perplexity just RAGs it

[agile jay]

Yep, since it's techically 32K Opus, then you.com actually gives you more context for free...

[warm cave]

if you.com is too buggy i will just move to anthropic and use bing for web search

[chilly forge]

Blocked him so that I cant see the useless messages

[tight prism]

I dm you the refer code

[warm cave]

a lot more

[tepid pilot]

bro I'm not boutta subscribe again lmao

[thorny canopy]

so they just added the limit suddenly? wtf

[tepid pilot]

I literally just emailed to cancel and for a refund

[warm cave]

I gave it 200 pages the other day, flawless, it was awesome

[tight prism]

Why me?

[livid mantle]

Guys keep in a look out for silent update everyday, always ask gpt 4 turbo in perplexity what is your cut off date?. If it responds April 2024, it's already the new powerful turbo that is better than opus.

it should be unlimited becacuse this GPt 4 turbo we have in perplexity has always been unlimited.

It's the same so it shouldnt be limited to 5 like opus.

• gpt-4-turbo-2024-04-09 $10.00 / 1M tokens $30.00 / 1M tokens

IN FACT ITS CHEAPER THAN THE CURRENT ONE:

• gpt-4-32k $60.00 / 1M tokens $120.00 / 1M tokens

[minor shard]

I really just want to know why it is so hard have a area that has "Release notes" so customers are not left in the Dark. Its very non-transparent and bad for business. Any other company releases notes about any updates they have for products they sell.

[tepid pilot]

new gpt4 is still 💩

[tight prism]

You don't know the refer code for $10

[tepid pilot]

absolute garbage lazy cra p

[formal parcel]

UPVOTE:

https://discord.com/channels/1047197230748151888/1230342648036982795

[tepid pilot]

what channel is this in?

[minor shard]

Pro feedback

[tepid pilot]

oh, I don't have acces to that idk why

[formal parcel]

@signal hamlet Care to explain the new Opus situation here?

[agile jay]

Leave and rejoin using the link at the bottom of settings

[tight prism]

Probably they are afraid we get upset over it

[warm cave]

Its in pro feedback

[minor shard]

I mean that is the definiton of shady lol

[thorny canopy]

limits and no temp change, wtf perplexity.

[warm cave]

@signal hamlet Nice time to be offine

[tepid pilot]

oh bet I see it now

[tight prism]

The power of cancel culture they fear

[thorny canopy]

someone better be messing up on their end

[minor shard]

We will prolly find out tomorrow AM PST

[agile jay]

Why do you think they removed the message counter a few days ago?

I knew something like this was gonna happen

[tepid pilot]

need to take to twitter

[tepid pilot]

was thinkng the same thing

[agile jay]

Just didn't think it was gonna be only 10

[tight prism]

Better than 5

[thorny canopy]

saw that, but didnt assume theyd pull a poe on us. Oop

[warm cave]

Yeah, and we complained non stop when it was 5, bc 5 is an insult, why even pay for 5 a day, you can do almost nothing with that

[minor shard]

I mean they are hiding the number in the settings where they also hide the model selector lmao. So now you even have to go into settings just to see how many Opus messages you have left lol

[agile jay]

You gives 5 for free, with 200K context...

[tight prism]

👍

[warm cave]

how funny h a ha ha

[tight prism]

We don't know they will still keep the 32k now they limited to 10

They are last minute change

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[minor shard]

Honestly sometimes I am wondering if they are trying to get people to move to You.com

[tight prism]

So the context could change d

[agile jay]

Even the bot says it's 600...

[warm cave]

the bot is on our team

[minor shard]

lol

[tight prism]

It only happen today last minute

[minor shard]

next thing they are gonna shut down the discord so you have to complain in twitter/reddit

[tight prism]

Last minute they limit today

[warm cave]

no one is notified, its prob just Alex in there changing the number

[tight prism]

They won't shut down

[agile jay]

Also the message count is supposed to show up in the chat when it's less than 100

why not for opus?

[minor shard]

I've been super appreciative of Perplexity but changing this without any warning/communication is dirty.

[agile jay]

Yep, same with a lot of stuff

Like talking to a wall

[tight prism]

I know but not the first time

[warm cave]

Yeah, give us an announcement, instead of trying to hide it

[tepid pilot]

Perplexity Community 🧱 @signal hamlet @blissful raptor

[tight prism]

They fear the cancel culture or they are probably take action and then inform

[agile jay]

Aravind just came online on mobile...

[warm cave]

bruh, that is not how you avoid cancel culture. you show respect and are transparent

[tepid pilot]

Mr. CEO

[livid mantle]

When I wake up tomorrow, I will wake up with many questions answered in this chat.

[tepid pilot]

care to comment?

@blissful raptor

[tight prism]

I said or

[agile jay]

Yep, the paid users simply want transparency about what's happening.

[tepid pilot]

plus dispute if they don't refund me for the year

[agile jay]

Especially if you reply on it for work etc

[tight prism]

Budget reason

[tepid pilot]

don't care how many times I gotta say it?

[minor shard]

u keep saying this but haven't shown any evidence of this lol

[tepid pilot]

bros j lurking @blissful raptor

[tight prism]

Why else they limited opus then?

[tepid pilot]

idk idc

[agile jay]

Probably reading the chat log

[minor shard]

we won't ever know because they are not being transparent.

[tight prism]

@halcyon coral said will released Gemini ultra soon and the google give hint it is next month released

[livid mantle]

I will say nothing because I know Mr. Alex Romanov mentioned that they are going to add

GPT-4-TURBO-2024-04-09 UNLIMITED

So I will just wait.

[tight prism]

They have a bad habit of taking action after that inform us, multiple times

[warm cave]

I loved perplexity, but at a certain point it started to fell they where hiding and being very carful with how they say stuff to be deceptive, with like context length, hiding how many uses you have left, etc. I would not be unhappy about only having 32k, but they definitely deceptively made it sound like we had 100k, and then with removing the counter and not saying anything, yep they are definitely wanting to take usage without us knowing

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[tight prism]

You care that is why you are complaining here

[minor shard]

They have released other models without limiting others before. Like I said, we won't know the answer.. so I wouldn't worry about what the reason is. Just that they did it without any communication/discussion with the people paying for it.

[livid mantle]

i dont want to go back to copilot pro gpt-4 turbo please

[tight prism]

I agree

[livid mantle]

I agree

[agile jay]

Well now he's gone...

[warm cave]

what?

[agile jay]

The ceo came online when he was @ ed and then after staying online for a few minutes, probably to read what's happening, he went offline again

[tight prism]

Who?

[warm cave]

oh, haha

[agile jay]

If we see a tweet, I guess we know why...

[tight prism]

How you know?

[agile jay]

You can literally see who's online...

[warm cave]

it shows who is active in discord

[tight prism]

I guess the ceo didn't want to face the wrath of user

[agile jay]

And that's why you announce stuff before you do them... Especially if they are bad changes.

[warm cave]

@grand mauve @neat elk @toxic moss Mind chating about Opus?

[minor shard]

Isn't Denis a co-founder

[vital night]

lmfaooo they started banning ppl from the server

[tight prism]

Really?

[minor shard]

who are u

[warm cave]

really, i guess i might be gone soon lol

[vital night]

someone who j got banned

[minor shard]

who?

[tight prism]

https://x.com/perplexity_ai/status/1780768372268933629?s=46&t=Un1yLqIRg3sDiqpmnWHBfg

Perplexity (@perplexity_ai) on X

🚨 Update: We've added mixtral-8x22b-instruct to Perplexity Labs and our API!

Twitter · Apr 18, 2024, 1:20 AM

[vital night]

did nothing wrong imo

but ig that's j how they operate when they get criticism

[tight prism]

So cool the mistral said is better than any model

[minor shard]

what was ur username?

[warm cave]

what? who said its better

[tight prism]

The new released api

[fresh eagle]

30 uses?

[warm cave]

yeah, but 8x22b is good, but not that good, its like a geter gpt 3.5

[vital night]

@tepid pilot

[minor shard]

also available via API

[minor shard]

lol it may have been because you Kept @'ing the staff

[agile jay]

Denis is messaging in pro lounge

[warm cave]

I have not been banned, yet. and i was @'ing

[vital night]

ig they j like u better

[agile jay]

It's taking a while, so probably a wall of text

[tight prism]

Ohh

[tight prism]

Yea

[vital night]

better but still not good enough

[warm cave]

200 is the min, come on

[minor shard]

There is a bigger problem they are working out

[vital night]

lmaoo

what's the country tho

[minor shard]

Check #👑│pro-lounge

[vital night]

@sly minnow

[tight prism]

Maybe we complain a lot that’s why

Budget reason

[vital night]

blocked

[tight prism]

After we rant at them

[minor shard]

It was not a budget reason

[tight prism]

What I do???

[tight prism]

Why they limited opus then?

[vital night]

@sly minnow what's the country ur talking abt?

[minor shard]

Malicious actors abusing it.

[vital night]

were they like scraping with websockets? thirsty asf

[tight prism]

Again they have no time and last minute do it for they fear our wrath

[tight prism]

I hope it was worth the rant once we see any improve next month

[thorny canopy]

100-200 a day seems fair

[warm cave]

100 is a bit low

[cloud hearth]

Would be super quick and easy to put out a message before making the change

[warm cave]

lets just use the math, 250

[agile jay]

I would think it would be easy to flag abusers. I imagine they have a faster than normal messaging rate, and that they probably have prompt templates, if they're using a bot.

[thorny canopy]

Id prefer 200 or 250

[warm cave]

Im sure you are right

I am using it to study for classes and as a tutor, 100 a day is not enough, 200 to 250 is enough

[tight prism]

10 to 30

Better than 10 at least

[agile jay]

Yep, really the users shouldn't have to be bargaining for a deal they already paid for.

Pretty sure that's illegal in the EU where I'm from.

[tight prism]

I guess after the ceo saw our complain they change it to 30

[vital night]

#💭│ideas message

[tight prism]

30 is better than 10

[vital night]

p sure it's illegal everywhere lmao

should sue

[livid mantle]

im so sorry dont be sad pls

[vital night]

emotional damage for restricting my right to free speech

[agile jay]

But not better than 600...

[livid mantle]

wait for "gpt-4-turbo-2024-04-09" silent updated unlimited

[thorny canopy]

they should offer refunds for changing things of course.

[minor shard]

Don't need Opus anymore 😄 lol jk

[livid mantle]

Hmm

that is so sad, im so sorry. tomorrow make sure you call your bank and your general attorney

[minor shard]

Most of GPT requests are the new GPT4 turbo in perplexity but is it still as lazy?

[vital night]

nahh I been saying, even if it's better on chatbot arena, imho it's still bad, and I think a lot of power users share the sentiment

chatgpt is still lazy, that's what I'm going off of

[livid mantle]

We will see, I am tempted to just pay the api in the playground of openai

[agile jay]

Yep, but it also makes it harder to recommend to other people.

[livid mantle]

i will be patient

[minor shard]

same

[agile jay]

Imagine a friend or family member recommending it to you, and then a few days later they get limited messages...

[livid mantle]

everyone please hang up your phone with your bank, wait for announcement

[tight prism]

I think they decide the 30 based on this

[minor shard]

LOL

[vital night]

hang up phone lmao? just click a button and u get ur money back

[tight prism]

Last minute

[thorny canopy]

im literally yelling at the bankteller with a gun rn, wdym wait?

[pine valve]

I only bought the pro version after seeing the opus countless times

[vital night]

instructions unclear I hung the bank manager

[livid mantle]

everyone make sure you save this picture, incase he deletes it in his twitter. otherwise we dont have evidence

[minor shard]

Idk how much legal weight that holds lol

[tight prism]

I think we should wait for next month see they improve anything before we cancel culture them

[agile jay]

Unlike most other sites, perplexity gets most of their traffic from people recommending it, so it does matter.

[livid mantle]

yes join jawnze and I, the patient waiting club something unlimited is coming

[tight prism]

It's the cancel culture they hold

[warm cave]

I literally just got friend to join about a week ago and my brother yesterday, i showed them how good opus is and that you get 600 a day, now i feel like i made bad recommendations

[agile jay]

Unlimited Claude Haiku 200K...

[minor shard]

I hanged up on the bank like u said to do 😄

[tight prism]

So that's why you are mad

[livid mantle]

that is so heart breaking

[warm cave]

Im mad for more than 1 reason, thanks

[tight prism]

It’s last minute

[sleek harness]

Enjoyed that for less than a month. Perplexity is only as good as its best model, if we have limited access to the best model while paying $20 a month, we might as well not have a pro subscription that has nothing pro about it. AI market is super competitive, can’t have more than one subscription. They either beat everyone else or get beat.

[tight prism]

We got 30 what more you want plus many people are using the refer $10

[minor shard]

lol Daily you are great ❤️

[warm cave]

realy i use perplexity for opus, web search is realy good from you.com, bing copilot and many other options

[agile jay]

Yep, their tweet literally says Opus though...

[vital night]

they said unlimted opus

[livid mantle]

we are same, we all use perplexity opus unlimited 24/7

[tight prism]

I agree you said we all know it's the only model better than anything else plus most people use the refer code for $10 a month

[vital night]

honestly, give me opus unlimited, 200k context, perplexity search capabilities, ability to add my own APIs (like custom GPTs) I'll pay $60 a month

maybe even 100

[minor shard]

im sure a lot of people would pay 30$ or maybe even 40$

[tight prism]

Nobody are rich as you

[smoky spear]

At least 200 times, otherwise this matter would be like a joke or a scam

[vital night]

I'll pay 1/4 of the value of the time it saves me

[agile jay]

Most people have been using it for more than a month...

[warm cave]

I would be happy with the option to use 200k context gut i only get 30 or so questions but if i have it set to 32k then i get 250 uses. or something like that. so then i ca chose

[vital night]

prolly a lot more than 100 lmao

[minor shard]

thats not always a static number unfortunately lol

[agile jay]

I already pay for the other models

At most, I'll just not pay for perplexity too

[minor shard]

would be cool to have a option to switch between the two

[tight prism]

I'm using it but I paid for $10 and still reasonable because less people paid for $20

[minor shard]

Curious, which ones do you pay for currently. I respect you a lot as a individual.

[warm cave]

we get that you got 10$

[tight prism]

Nobody are rich as them

[livid mantle]

I agree

[tight prism]

Because they don't search online like perplexity

[timid jungle]

3/17 PPLX announces unlimited Opus.
4/17 pplx quietly changes to 30/day Opus.

[agile jay]

Perplexity,
GPT Plus,
Claude Pro,
Gemini Advanced,
Suno,
Copilot Pro,
And maybe one or two I'm forgetting about

[minor shard]

Perplexity is also getting ready for people to start recieving their Rabbit R1s

[agile jay]

It was 10...

[livid mantle]

Grok-1.5

[agile jay]

Only got increased just before Denis posted his message

[tight prism]

That guy is rich

[minor shard]

Nice, why do you pay for perplx if you have access to most of their models?

[agile jay]

perplexity search is just better

[minor shard]

true

[livid mantle]

did you know that you can search with perplexity without pro

[warm cave]

If you had to pick 3, what would your top 3 be?

[agile jay]

I do a lot of stuff which requires upto date knowledge, which the others aren't that good at.

[tight prism]

I'm in a foreign country the price is high when exchange currency

[agile jay]

Perplexity, GPT Plus and Claude Pro

[warm cave]

Im geting claude pro again

I have GPT

[agile jay]

I also have most of them paid for my employees too.

[tight prism]

Why do you have so many? Are you doing a review or something? You are rich as f

[agile jay]

Because I own a few businesses.

[minor shard]

Code is a very respectable person. Really good person to get info/opinion from.

[tight prism]

I want it released in my country but sadly they don't

[agile jay]

Also paying a bit for AI tools to boost productivity is a small price

[tight prism]

Better than 10

[agile jay]

We also use the API's for internal stuff.

But that is a lot more expensive.

[tight prism]

That's is awfully suspicious, why do they own so many

[minor shard]

lol what

oh

[agile jay]

I am pretty sure I've shown screenshots of me testing stuff for other people, lol

[agile jay]

I have money...

[tight prism]

So you are reviewer then?

[agile jay]

To me a few grand a month is not much

[minor shard]

I mean if u have the money, why wouldn't you pay for access directly to the vendor who is selling the product rather than through someone else

[agile jay]

For API usage?

We do

Yep, almost over 100K a month

[tight prism]

Again nobody is rich as you

Do you have any idea have much is $10 exchange currency it cost???

[minor shard]

Yeah true, API is expensive

[agile jay]

very large context is painful

We parse large documents all the time, which is why we needed to pay for it.

And RAG doesn't cut it.

[tight prism]

$10 is like $50 here when change currency

[agile jay]

Yep, frontier models AI are really expensive for a lot of countries.

Where the average wage is around 3-4K a year

[minor shard]

Im suprised ur not trying to use Copilot/Azure

[tight prism]

When they are rich like that guy and again nobody is like him a reviewer that owns many to review and gets paid

[agile jay]

We do, and also Gemini for google sheets etc

I pretty much am covering the cost myself, since I want to see how my employees will use it given the chance.

[tight prism]

People keep saying is $10 I got but they don't know to exchange currency is $50 is here

[agile jay]

For the tech illiterate ones, the GPT Plus, voice chat is a lot easier to use.

[warm cave]

a lot of people have like 6 streaming services, if people value AI then 30 is not alot

[agile jay]

Yep, it's more related to utility.

[livid mantle]

im going to the bed and cry, goodnight everyone.

[agile jay]

I only use around 15 prompts a day on perplexity, and that saves me more than an hour a day.

Which costs more than 20/month for me

[tight prism]

I work to get a least pro for $10 even though is expensive in my country but is better than $20

[minor shard]

Sleep tight, don't let the Opus bugs bite.

[agile jay]

Night

[minor shard]

lol im not sleeping yet

[agile jay]

Oh, it was a reply

[livid mantle]

nomore unlimited opus @south kindle

[tight prism]

30 is good enough

[minor shard]

Shhhh don't get him worked up

Hes already prolly stressin in OW2

[agile jay]

Not really, if you are using it for work.

I would imagine if you are working, and you are writing a prompt every 15mins, then that would be 8 x 4 = 32 already.

And most people are probably using it more frequently than 15 mins.

[south kindle]

😦

[tight prism]

Opus is expensive plus they probably limit to 30 because the rant and this

[tight prism]

Again nobody is like you a reviewer

I know people use opus but 30 is already good as enough

I care because I want Claude 3 opus and Gemini ultra I been telling everyone I want both plus nobody can beat perplexity when it comes to search source

[minor shard]

@livid mantle look what you did now

[cloud hearth]

Main problem is transparency is lacking here

[warm cave]

<\Code> uses says the best 3 are chatGPT, Claude Pro and perplexity

[tight prism]

Unfortunately nobody can Beat perplexity

[warm cave]

Maybe go to Claude pro, you will get usage limits, but you will be able to talk to it more than 30 a day

[calm swan]

yo, is cody gpt-4-turbo already the latest yet?

[tight prism]

We can't do anything but rant here at least

[serene sand]

#💗│sharing message

[tight prism]

They don't have search source the problem

[minor shard]

Always love ur images @serene sand. I feel a mix of calm and anxiety for some reason with this picture lol

[agile jay]

There is always something you can do, lol

[tight prism]

Everyone what we can do is wait for next month and see any improvement and if not Twitter is the next best thing to start cancel culture there

[tight prism]

Nobody is a reviewer with many follower and rich like you

[serene sand]

https://www.youtube.com/watch?v=j2d6T5G2rrY

[warm cave]

yeah, but when you need sources ask ms copilot and past it into Claude Pro, works well, did that before

[tight prism]

What he can do is cancel culture but again we have to be patient for next month, too soon to complain will be a problem for the upcoming project they have

[tight prism]

The price is not cheap and is not as cheap as Perplexity and they have referred code to use

[agile jay]

There are more options than just complaining. The AI sphere is pretty competitive, so making your users not trust you is a pretty bad idea.

Yep, but that's something you should do before putting the limit.

Not after people realise what you've done...

[tight prism]

Next month is the new released ai model

[agile jay]

Yep, but noone really cares about Gemini 1.5

[tight prism]

Perplexity did this many times whether is good or bad

[agile jay]

Especially since it's likely the context will be capped to 32K

[tight prism]

Is the Gemini ultra is way better than that

[tight prism]

Again we don't know it yet because the last minute limit

All we can do is wait and see any upcoming improvement next month before we start the cancel culture otherwise it will ruin their upcoming project

the perplexity is better than any competitor we know

[minor shard]

Haha yesssss!

[agile jay]

They should just not accept virtual cards then/ Pretty sure stripe gives you decent control on that stuff.

[rough flower]

Oh no. Now Opus daily limited to 30 messages is very low. At least we need 100 or 150 for a day. Please reconsider this

[modest lily]

That is sad, I don't know if reporting to Apple (those who paid using apple) will help to get refund. I regret this.

[tame current]

😦

[torn ravine]

Wow that's a big change. Didn't notice this. Now I get PTSD every time I ask a question. 30 is not enough. Actually, the cap is the End of Knowledge🥹

[tame current]

same 😭

knew there was something wrong they hid the 600 prompts left message

cursedsweat

[hallow linden]

phind pro seems to have 500+ opus uses a day. Might consider jumping once my subscription runs out

[tame current]

you.com too

[torn ravine]

Should have thought about it

[tame current]

this was yesterday 💀 did no one know about this or were they just lying

[torn ravine]

I subscribed to You.com days ago. It's buggy, but usable enough to assist daily work

[unborn minnow]

Perplexity is giving dumb answers. There is no alternative right?

[tame current]

this is such bad management

bruh

[torn ravine]

I think we're the bad guys now

[hallow linden]

ok, so in terms of actually searching for stuff, I actually just use sonnet as my default. The speed is great and it's good enough for anything i need to just look up. If I need something serious for programming or reasoning, I'll end up using GPT-4. That being said, I think opus is the best writer out of all of the models that currently exist. It will be a massive bummer if I have to work around this limit.

[tame current]

what does good guys mean, so only certain users got the cap?

Bruh

[serene sand]

👍

[tame current]

lmao

all of this could have been avoided if they just didnt limit everyone and took the abuse for a week more until they find a way to fix it, but hey im not running a multi million dollar company

Peepo_Bruh

[warm cave]

or prepared in advances, or just gave us notice of tha changes before they came out of nowhere

[tame current]

tbf I don't know what I would have done in this situation

they already use cloudflare

and if that didn't prevent it

[tame current]

yep, honestly idek if i use 30 prompts a day, probably not, maybe once or twice a week at best but this is just bad communication and loss of trust

limiting the number of queries was probably easiest way

[tame current]

true

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[warm cave]

Im just saying, put up a announcement saying we will be temporally limiting opus messages while we deal with abuse of the platform, your regular usage limit will be restored as soon as we can.

[tame current]

💯

[torn ravine]

Waiting for an announcement

[tame current]

there was that time they disabled the ability to remove sources, I think

i just joined this month so idk, cancelled all my subs and came to perplexity and even brought 2 people with me🫠

[warm cave]

Yeah but why update us after the issue, should take 10 min for them to approve an announcement updating users on why they lost most of there usage

Alex could let us know

[tame current]

yep if they gave us a timeline and an announcement i literally wouldnt care

[tame current]

cus i get stuff happens

[serene sand]

instead co-founder let you know...

[halcyon coral]

Please calm down everyone.
30 uses is still enough for most users (and opus is not the only model...) and I would like to point out that it is only currently limited to 30 😉

[tame current]

we have been edged 😂

[warm cave]

Yeah I know, but that was still way later, haha

[warm cave]

Fr man

[tame current]

and in #👑│pro-lounge ...i would have missed that if it wasnt for people complaining in general

[slow violet]

This is not enough if you need to work/study closely with him.

[halcyon coral]

You can still use GPT-4

[tame current]

• an announcement, in the proper channel
• timeline
• reason(which you already gave)
  and no one would have cared

Bruh

[slow violet]

His speech is not as rich as Claude Opus

[tame current]

his

LOL

[half venture]

30 per day ?

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[tame current]

lmfao

[half venture]

Thats even less than claude own official website

Atleast match that

70 per day

[tame current]

how much is that

[halcyon coral]

You still get virtually unlimited usage, only Claude Opus is currently limited

[half venture]

70

[half venture]

Make it 70 atleast

[tame current]

i almost exclusively use opus, gpt4 is not at all useful for programming its very lazy, 70-100 would be ideal until you figure out whats wrong

[halcyon coral]

It’s not possible to give a timeline for fixing abuse…

[half venture]

It will be restored right

To unlimited

[warm cave]

We were all worked up bc changes had been made in the past without explanation. Also, the context window stuff seems deceptive; there is no clear answer, but we kinda get it now, we get 32k plus rag. And then we though we had just lost Claude 3 forever, and many of us are already in year long subscriptions. Feels like crap. All avoided with a little notification, I know sometimes you guys can't get an answer out right away, but I feels like some things are kept secret from us, I really did not expect anyone to say anything at all. Keeping us updated on most thing would help, bc when something like this happens we will be more likely to think, no worries, we will find out soon. I hope you get what I mean with this

[tame current]

and because how convenient pplx is, many probably told others to switch, ik i did, now this is just bad

[slow violet]

Do you have any information when this limit will be lifted? (English is not my native language, I apologize if I make grammatical mistakes)

[warm cave]

He said no timeline

[halcyon coral]

#💬│general message

[warm cave]

Ok, actually time to go, good night 😴

[real temple]

I have a code for 200 dollar credit by Rabbit team, i am unable to use it, @tranquil orchid please help

[tame current]

❤️ thanks again

[halcyon coral]

Please contact [e-mail]

[real temple]

Thank you

[slow violet]

In the Android app, still shows unlimited)

[gilded coyote]

I would wait, all I want is 100 per day

[lilac stump]

on the "welcome to pro" email i got 2 hours before they put the cap in, it says unlimtied too 🙂

only slightly illegal where i'm from 🙂

[ocean tree]

Hello do you have test code ?

[torn sierra]

You advertise that Claude 3 Opus is unlimited, on X for example, with not a word, of course, about the context window because information about that can only be found on Discord. And now you implement a 30 uses per day limit. Is this temporary, so it will be increased to 50 later? I didn’t choose Perplexity to use other models, but specifically for the unlimited use of Claude 3 Opus as advertised. I also convinced several people to start using Claude 3 Opus with you.

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[tiny plaza]

I switched over from Phind because Opus had occasional downtime from rate limit issues TwT

I mean I get why, but I feel sick knowing I just wasted my money -_- 30 uses per day is not a lot x.x

[halcyon coral]

Please check my messages in #👑│pro-lounge

[torn ravine]

Subpoena warning

[tiny plaza]

I hope you guys manage to sort it soon, I'm so impressed by Perplexity 🥹 especially the writing mode

[torn ravine]

Limiting context length and now usage is kind of not decent...

[slow violet]

What did you say there? My access to this channel has been closed because my subscription ends today.

[halcyon coral]

You have access now, you can see the channel 🙂

[slow violet]

Thanks

[halcyon coral]

Are you also a pro subscriber?

[fierce ridge]

I'm a product manager and I need to tweak the requirements document frequently to get it to a level I'm happy with. I think 100 times a day is an appropriate number.

[swift charm]

hey just curious, what's the logic behind the Discover section? Is it user generated content or like some trending news summarized by AI?

[halcyon coral]

You can check the messages, the first message is here: #👑│pro-lounge message

[serene sand]

[tame current]

[serene sand]

https://tenor.com/view/hydroq-pro-hydroqpro-hydroq-op-gif-20865313

[sterile cosmos]

can i refund????

[torn ravine]

Anyone reaching 30 limit? What happens next?
If the limit is so low, at least there should be a model switcher beside the send button so we can control which questions use Opus and which not with granularity

Got it

[tame current]

https://tenor.com/view/donald-duck-counting-money-cash-rich-gif-15705037143588792180

[tame current]

he didn't say they'll keep it unlimited

¯_(ツ)_/¯

@mental stumpwhy u using an alt 😂

[crystal elbow]

I still dont get it, where do we see how many pro searches we got left? It's simply gone?

[tame current]

no, thats still for <100 messages, to see opus count go to the settings where u change models

opus is capped to 30

[tame current]

https://tenor.com/view/checkmate-check-mate-got-em-gif-6408903245572515122

[crystal elbow]

Jesus why the heck do they just change this while promising 600 a day. Imagine subscribing for a year.

And then we'll probably hear "YoUu AgGreEdD TOo thiSSs WheN PAyINGg"

gg

[tame current]

because some users found a way to min max opus usage, they have made the very smart decision to cap to 30 for everyone

[crystal elbow]

Uh... what?! What are you on about? I don't care what other users do, it's perplexity's responsibility to deal with that while maintaining the promise they make. I get it if its temporary, but promising 600 a day and a week later you drop it to 30 is a bad move.

And granted, 600 a day is a lot, I get that. But going from 600 to 30 is like insane.

[tame current]

i was being sarcastic, i agree with you lol

[tame current]

yep, people suggested 70-100 for starters

or have a way to carry over unused prompts from today to tmrw until they find a way to fix it

[crystal elbow]

Yeah exactly, I rarely use it in the weekends, but during the week it's different. Would be nice if it'd get carried over.

[tame current]

I also just subbed a week ago and also got 2 of my friends to sign up 🫠

[crystal elbow]

I just was not expecting it to just drop without any notification, perplexity seems like more "professional".

[tame current]

maybe they should have consulted perplexity before making the decision 😂

Homework 💀

[tame current]

dont prompt shame

😔

Smirk

Cat

Cat_Shakes_Head_No_Inv

[swift charm]

bruv

[glad pilot]

anyone know any replacement for perplexity except you.com?

[tired jungle]

Hello , can I have a code promo to test perpelexicity pro plz ?

[torn ravine]

Surprise on one month-versary

[merry juniper]

What kind of metadata does Perplexity pull from for the sources? I noticed that a lot of danish media sites have wrong logos and generic titles.

[glad pilot]

Just scroll over the channel, I would say I am shocked

Just renewed the plan and now they told me I only got 30 opus per day

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[glad pilot]

wth

[tame current]

sadly there isn't

not with web search anyway

[glad pilot]

If I can back to the past I would slap on yesterday myself and yell:" don't believe the xhit, go and get a Claude pro."

😂

[tiny plaza]

Phind, but it's more optimized for coding. You can still use it for general purposes. That being said, they're having rate limit issues with Opus (from Anthropic's end) so there's occasional downtime

I actually moved from there to here lol

but well that turned out well -.-

[fast meteor]

my 2cent: excite sharing + retention/DAU

[torn sierra]

According to the information they provide here on Discord, they are experiencing some issues with misuse. Generally, I'm not satisfied with the way they operate. Responses that there are other models and that 30 is still a large number seem out of place here. Above all, I'm irritated by the way information is presented and the style of communication.

[tiny plaza]

Yeah it comes across as "You should be grateful we're giving you 30 messages a day"

[fallow escarp]

with perplexity pro, if I select gpt-4 or opus, can i turn off search?

[tiny plaza]

It's definitely frustrating dealing with everyone coming at them about it, but a little bit of communication goes a long way

I just hope they sort it out soon x.x writing mode is really amazing

[glad pilot]

The only reason I came to perplexity is that Claude pro only allow like 70 message per day(actually per 8 hrs), while perplexity got 600. Now they told me we only got 30

So who is the clown

Apparently me

[fallow escarp]

?

no, i mean i want to use opus but don't want it to search the web

[torn ravine]

30 limit is acceptable if they use Opus-200k. If not, why not go with Claude Pro

[fallow escarp]

didn't aravind say they were offering unlimited opus

[glad pilot]

should be bug

For me it is still 30

Ok that might be another reason

Though for me it is mostly the limit thing

[chilly forge]

||“Marketing gimmicks”||

[fallow escarp]

yeah, i got confused. thanks!

i was about to buy pro but now i see this

if no unlimited then what do

[glad pilot]

yeah don't be idiot

[stable radish]

I second that

[halcyon coral]

It’s only temporarily because of abuse

[fallow escarp]

buy what then

i want good big model

[glad pilot]

Claude 3 pro

[tame current]

you.com maybe, you can get your first month off, maybe use that until pplx figures out their stuf

[fast meteor]

opus is overrated anyways

[glad pilot]

You.com sometimes mixed their own model with opus

Which suvks

[sly kiln]

Claude 3 Opus on Pro seems to switch to Sonnet quite easily

Did they set a different cap for Opus?

[tame current]

you have run out of messages

yes 30 per day

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[sly kiln]

Yes but still have GPT 4 queries left

[tame current]

ooh

[sly kiln]

I have barely been using it, nowhere near close to 600

[stable radish]

What you mean 30 per day?

[glad pilot]

It just feels more ironic everyrime this bot shows up

[tame current]

they have recently changed claude 3 limit to 30 per day (previously 600)

[stable radish]

Where's the announcement?

[glad pilot]

Nope

[tame current]

some random message in #👑│pro-lounge

[glad pilot]

No announcement

[stable radish]

That's outrageous.

[sly kiln]

They should have made it clearer

There isn't even a counter anymore

For the number of Pro queries left

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[glad pilot]

Man

[buoyant kindle]

Wow. The spam and crying and whining here. Are you 12 years old? Do constructive criticism or go somewhere else.

[tame current]

we did, they just keep saying "trust me bro" instead of any actual responses

you have to ask a mod or sign up from the website using join button

[sly kiln]

It is constructive because both have easy fixes

[buoyant kindle]

Good. Then pick another service. There are plenty. I'm using You.com at the moment. When my subscriptions run out, I'll see where I stay. Unless ChatGPT improves.
Stop wasting energy on crying like entitled kids.

[stable radish]

Lol, I accept to be viewed as a whining 12 years old kid. I'm paying for a service.

[sly kiln]

Bring back the counter, make it clear that Opus has a separate limit

[tame current]

i am using it too, but i paid for an entire month here

[stable radish]

I expect at least communication when things like that happen.

What the hell.

[buoyant kindle]

I paid as well. But I'm sure we've all paid for things that turned out to be crap. Learn from it and go on.

Stop expecting. It's naive.

[tame current]

paid for a service you dont get

wah look at this entitled kid

[kindred phoenix]

hellow, why is my opus only 30 prompts?

[buoyant kindle]

Oh god...

[stable radish]

That's him, yeah.

[kindred phoenix]

how do that work?

[halcyon coral]

Hey everyone, please check this.

First message:

Extraordinary events require extraordinary actions, which is of course annoying but unavoidable.
If some users didn't abuse Perplexity, we wouldn't have this problem now!
You still have 600 uses per day at your disposal, only Opus is temporarily capped.
Please be patient 🙏🏻 and sorry for the inconvenience

Question:

What do you mean by abuse? I usually use Claude3-Opus only a few times a day, but when I was creating an app the other day I used it almost 200 times in one day.
Answer:
Of course, your method of use does not count as abuse.
Abuse is when, for example, you write code to automatically integrate the full 600/24h uses into your own product (API,...)

Question:

Why is it that not only suspicious users but all users must be limited to 30 uses?
Answer:
It's not that simple, there has to be a reliable system to stop the abuse and this system is currently being developed 😉

Additional information:
The limit is only temporarily.
It's not possible to give a timeline.
It will be announced as soon as there is news.

[tiny plaza]

don't shoot the messenger k cx

[tame current]

true i feel bad for the mods tbh

[kindred phoenix]

ok this make sense but how long will i need to wait?

[tiny plaza]

honestly though, if we weren't so blindsided by this i don't think it would be as disappointing

[fast meteor]

pls integrating maxmind asap

[chilly forge]

Tbh everyone needs to know the exact duration of this “temporary”

[halcyon coral]

This is of course the top priority, but it's not possible to give a timeline

[tame current]

• an announcement, in the proper channel
• timeline
• reason(which you already gave)
  and no one would have cared

[halcyon coral]

It's simply not possible

[glad pilot]

How? So they just share accounts?

And talking about a new system, it sounds like things can took for month

[fast meteor]

they built bots which automated the service and max-ed out the limit every single day

[winged moon]

Wait, they just added a daily limit for Claude 3 Opus. This sucks

[vapid onyx]

#💬│general message

[torn ravine]

Nice pun

[winged moon]

I only learned about it when I wanted to switch models to get a different outlook

[tranquil forge]

You good guys just keep blaming those bad guys! Well, our company is as pure as an angel.

Listen up, drop the gun! This is all your "600" guys fault. Now we have no choice but to rough up your poor "good" sister!

[wise edge]

why opus model has been limited to 5 times ?

[wraith crow]

#💬│general message

[tranquil forge]

The good folks will just have to suffer a bit. They'll understand. It's all for a brighter tomorrow! Ura!🥹

[fallow escarp]

probably pin this

[fallow escarp]

yep read this

[tranquil forge]

Its a good day to die!
When you know the reasons why!
Citizens, we fight for what is right!
A noble sacrifice!
When duty calls you pay the price!
For the PPLX Federation I will give my life!
Good guys,let's charge!You wanna live forever?

[buoyant kindle]

The moderators deserve a medal for their patience. 👏

[clever jetty]

Perhaps you could limit the rate instead of limiting the number of times, for example, single-threaded access once every ten seconds.

[winter geyser]

sup

[sonic musk]

Why Only Opus got limited?

[wraith crow]

The restrictions are temporary please check the pinned messages.

[lunar ivy]

My subscription just almost cost me fee hours later, gladly I'm saw this message,and cancel subscription before too late 👀

[crimson obsidian]

which one are you guys using ???

[sleek harness]

Guys, this really requires an announcement or the chaos won’t stop. Took me hours of confusion before finding that message, and the only reason it was found is because a mod was finally around. You guys don’t have to work overtime, and we don’t have to be angry and confused if the message was in announcements.

[torn ravine]

Maybe the team’s still sleeping and will wake up drowned by messages here

[upbeat tiger]

What the actual f, only 30 daily uses? this is REALLY NOT ENOUGH.

[torn ravine]

Here comes the competition

[upbeat tiger]

May I ask what service is providing that ?

[tame current]

eh I agree but they brought this upon themselves

[wraith crow]

please dont share referral links to other services

[upbeat tiger]

Thanks, I will check that Orfeo. I need unlimited uses so badly, or at least, just not 30.

[torn ravine]

I was just about to say, if inappropriate, I can delete it. For informational purposes only.

[fast meteor]

free opus here

[wraith crow]

at least hide the embed 😂

[tranquil forge]

I love this guy

[fast meteor]

I forgot! I messed up

[tranquil forge]

Rick Stley，very good song！

[scarlet pagoda]

you guys really put a dent in my productivity with the OPUS limit lol

would have been nice if it was at least 100 or 200 for the temporary cap or something

still appreciate the service, 30 better than nothing of course

[torn ravine]

Now it’s like power is out and everyone gets out chilling on the street

[placid coyote]

hi, went to discord to see what is going on with opus, since recently it went unlimited and it was quite good for programming tasks (chatGPT feels lazy again) and great at writing. now I feel a bit bad about working on my toy project, well more like letting opus write it (many dozens or low hundreds of messages every day for a few evenings) while in his funny persona 😅

[tame current]

dont give them ideas

[opaque stump]

What’s the process of canceling my annual subscription that I purchased last weekend?

I’d like to switch it to a monthly sub instead because of these changes

[torn ravine]

You can email them at [e-mail] and state your cause

[tame current]

contact support/your cc

[minor shard]

I think it may be [e-mail]

[torn ravine]

My bad

[foggy tapir]

well, they increased the cap by 5 ? 🤔

[opaque stump]

Done. This is quite a rug pull.

[tame current]

https://tenor.com/view/klatschen-clapping-gif-12186138

[upbeat tiger]

At least increase it to 200...

[opaque stump]

Last time I’m buying an annual subscription to anything LLM related!

[modern crown]

@opaque stump yep monthly is the best way to go

[minor shard]

AI space is definitely changing constantly

[stable radish]

I'm feeling the absence of Claude 3 Opus

[supple sleet]

I confirm. I thought Opus was unlimited?

[foggy tapir]

Once my monthly sub ends, I think I will buy from Telegram instead of the official channels.

[stable radish]

Look at pinned messages

[cosmic folio]

How does the Perplexity API represent that he and Openai's request address are not the same？

[pine socket]

Bot abuses, that cap 600 queries a day is obviously problematic, but the hemorrhage could have been stoped with a cap at 100 requests, rate limiting every 15s would have helped too. Cutting the issue by at least 6.
Clever rate limiting & bot detection is primordial.

Extraordinary events require extraordinary actions, which is of course annoying but unavoidable.
If some users didn't abuse Perplexity, we wouldn't have this problem now!
Bots on an "unlimited service" are common practice, nothing to be called extraordinary. It's the service's responsability to anticipate properly these kind of abuse, "punishing" the whole user-base because of bad priorisation is a costly move.

If we look closer, considering the cost of a model like Opus, 20$ will not cover a regular daily usage.
A practical solution to lower the number of Opus queries would be to allow greater control on the model selection, query by query (A feature that has been requested numerous times now).

[stable radish]

Lisan Al Gaib

[opaque stump]

I'm sure I'm not the only one asking for a refund but I am hoping they respond sooner than later. Or at least let us cancel on the website prorated

like, Sonar is fine but the real value prop for me was Opus so I might as well toss the money directly to Anthropic at this point. Although, they need to come out with an iOS app already.

[wide mesa]

yo, is it better to use Claude Sonnet or GPT4.0

[wicked parcel]

Does anyone know how to get around the context limit of about 30.000 tokens on upload of text files? Someone mentioned Haiku in PPLX labs but it crashes when I paste a text of about 100.000 words. Its stuck. Or any option in europe to use the full context window in claude opus? I only need opus but cannot subscribe there directly..

[stone valley]

100% that ! And I don't really get the concept of abuse in this case. Sure bots are something problematic, but even if everyone who has a sub do 600 queries with opus per day, I don't see the abuse here : it's litterally what we are paying for. Using the product we subscribed at his full capacity is not abuse

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[minor shard]

Yeah, not sure why the cap exists if they cannot afford it to be used “hypothetically”. It’s a shady practice unfortunately to get people to sign up I assume.

[wide mesa]

Actually where do you find Telegram places for that?

[tranquil forge]

No,our angel company can't afford it,just 100 pls sir🥹

[warm cave]

Through perplexity I think it is impossible, haiku works for me buts takes a while before it unfreezes, I don't know many options but you can look at you.com, but be warned it's buggy, idk about other platforms, I would look around and see if you find any good

[tranquil forge]

Don't make 600 times Lxxe with opus everyday,that hurts.🥹 You bad guys really bad bad

[wide mesa]

AI sites and being shady name a better duo

[wicked parcel]

You.com has a limit as well.. but its said to be a little larger.. but still no real solution.

[livid mantle]

we will not settle for

• "100 opus"
• "200 opus"
• "600 opus"

we will wait for OPUS UNLIMITED

[warm cave]

They say you get 500 a day and full 200k, but with web search always on it eats up some of that context. My main issue is it seems extremely bugy, I though it was a bit annoying, but now that lts if people are trying it, there is a lot of errors

[wicked parcel]

Heres the offial response from Perplexity: "Unfortunately there's nothing we can do about the context been capped at 32k, I understand if this situation is frustrating. The context window should still be able to be taken advantage of by breaking queries into smaller, subsequent follow up question as each follow up question is 32k context, you can continue to reference the uploaded .txt file as a source. Hope this helps." Does anyone know how to tell via a prompt that more text is coming and it should wait processing my request?

[stable radish]

Well, you just tell it to acknowledge previous set of instructions and then wait for the other set?

[fast meteor]

lesson learned: dont trust your users

[wicked parcel]

Ok now I am hitting the file upload limit. More than four files per chat or four text pasts is not possible

[fast meteor]

that limit is always there

[upbeat tiger]

and lesson learnt for us: don't trust any AI company with yearly subscriptions

[wicked parcel]

ahh it seems to work! I just told it: "save this text and wait for instructions"

[sly kiln]

Is the context window 32k for all the AI models?

[wicked parcel]

yes the maximum limit

[calm swan]

wait, it's out already?

[fast meteor]

dont trust any unicorn companies/startups

[foggy tapir]

Actually the guy who sold me linkedin premium, came with the offer of perplexity pro yearly subscription lol.

[austere kestrel]

that's an official response? I have hard time getting my head around this part:
"Unfortunately there's nothing we can do about the context been capped at 32k"

[livid mantle]

only thrust yourself

[austere kestrel]

Mistral, OpenAI, Anthropic - all impose that precisely same cap...

[uneven basin]

on arena always 3 opus request with no registration

[vapid onyx]

The official response regarding Opus can be found here: #💬│general message

[torn ravine]

Definitely should put it in announcement. It’s not a shame. Mid journey breaks down all the time. Keeping users in the loop is key

[cursive jacinth]

Hi sir. GPT 4 is so very very lazy sir. Feels like garbage sir. anyone still use GPT 4 sir?

[modern crown]

sir yes sir sir sir sir sir sir

[torn ravine]

Once I started using opus there’s no way back

[warm cave]

Same

[upbeat tiger]

They took GPT4 Turbo to remove OG GPT4, and the speed became absolutely trash despite being "turbo"
GPT4 turbo is the worst scam ever

[cursive jacinth]

i have been using opus more sir. Then i used gpt 4 sir. feels like it answers a garbage truck sir. it definitely does not answer your questions sir. Mistral answers better sir. What is happening to gpt 4 sir.

[upbeat tiger]

The issue is that gpt4 turbo is not gpt4, that's all.

[cursive jacinth]

you mean perplexity has fake gpt 4 turbo sir?

[livid mantle]

wait for silent update gpt 4 turbo

[cursive jacinth]

they mentioned 70 percent of perps people has the new gpt 4 turbo sir. how was it sir?

[frank nymph]

I’m a software engineer and Claude 3 Opus was serving me so well until they capped it. GPT4 Turbo is ridiculously slow. I hope the limit gets removed.

[stable radish]

Claude 3 Opus is the goat 🐐

[livid mantle]

good

[upbeat tiger]

No, gpt4 turbo is just not the OG gpt4, it's just a different thing, in a bad way.

[wicked parcel]

Is it possible that its not possible to switch the model anymore when a file is uploaded to the chat?

[cursive jacinth]

good sir? seems a below average expectation sir.

[livid mantle]

yes sir

[buoyant kindle]

Can someone ban the sir troll?

[agile jay]

He is just Indian.

[sleek harness]

How about temporarily having 300 opus responses per month instead of 30/day? Some days I use it once or twice others a hundred times, would be nice if I didn’t feel this limited.

[buoyant kindle]

No excuse for being an annoying troll. Just stop spamming "sir".
Although, in his case, it's so exaggerated that I don't believe he's Indian. It's more likely that he's mocking Indians.

[sonic musk]

300 / day is better

[sleek harness]

Yes for sure, but if they’re stuck due to API abuse, at least have a temporary monthly cap rather than daily

[agile jay]

Rip, first Claude was rate limited to 10/day, and now messages are limited in general to 1/minute ...

[wicked parcel]

right now perplexity is always going back to sonnet from opus. anyone else?

[wicked parcel]

I cannot switch back to opus anymore.. I dont get a message about my limit either.

[agile jay]

It's more that their team doesn't plan much. All they really had to do is limit new accounts temporarily and not all of their user base... Sadly they don't communicate with their users beforehand.

[onyx echo]

for file it is 32k

[warm cave]

I think it's pointless, I'm sure he is just the messenger, and can't say anything about timeline

[onyx echo]

If you chat with it for a long time it can be 200k

[warm cave]

I think a co-founder was the first to give us some info

[halcyon coral]

Nobody can give you a timeline, depends on when the abuse can be stopped

[warm cave]

I mean I'm sure there are people that know how bad the situation is, they have a better idea on the timeline, obviously not the exact date and time

[minor shard]

Llama 3 is out?

[warm cave]

What?

[minor shard]

its the azure marketplace idk- https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-8b-chat-offer?tab=overview

[warm cave]

I got to take a look 🙂 happy day

Zucky promise it would be open source, I wonder if it's out

[agile jay]

Yep, but the smaller models come out first.

[minor shard]

waiting to see it on HuggingFace

[warm cave]

Prob for the better, I would murder my computer loading the biggest one

[minor shard]

Same lol. Why use the lesser when u can murder ur PC with the best

[warm cave]

Same, hugging face, then ollama

Then when the big one is out we can start spamming perplexity to add it 😉

I think <\code> would explain it better

[agile jay]

Yep, I'd just like to self host a GPT4 level model on my server. For privacy related stuff.

Abuse is from accounts being used by bots for other services. Basically on another service someone will send a prompt, and then the prompt is sent to perplexity and then the output is returned to the user.

A normal user using 600 requests a day is technically not abuse. Just a power user.

[cursive jacinth]

i dont understand sir, why we are affected with this so called abuse by limiting our use with the current best LLM sir. It seems we are the one who is adjusting sir and we pay the same exact sir. I DONT UNDERSTAND sir.

[warm cave]

It's a bandaid, limit usage until they figure out how to solve it

[agile jay]

Yep, instead of only throttling users who have been using over 300 requests a day, which would most likely be abusers, they are throttling everyone...

[warm cave]

Yeah, or only new account who are using over 300

[cursive jacinth]

I still dont understand sir.

[agile jay]

Nope, fake news. It's simply from a few accounts abusing the limit.

[torn ravine]

I think good guys may be qualified for a fair amount of compensation if the issue persists for something more than a week

[thorny canopy]

the cap should be 200 at least. 30 is just insulting...

[agile jay]

Also the original limit was 10/day. You can thank the people complaining and Danis a cofounder, for the limit increasing to 30.

[warm cave]

They could have not limited us by narrowing down the restriction, particularly to new user with really high usage

[cursive jacinth]

why this abuse thing wasnt been an issue before sir, like when GPT 4 has its peak prime sir?

[torn ravine]

like an extension of subscription

[agile jay]

They most likely ignored it because GPT4Turbo is 2.5x cheaper than Opus.

[warm cave]

Claude isb lot more lucrative, cost per token, 2.5x gpt-4

[pine socket]

Thats pretty normal, no system is engineered (technically or financially) to be used at 100%, 100% of the time. Limits are placed to prevent "obviously damageable usage".
The bet is to have an average cost under the value of the subscription. Some users will cost more, others less.
The issue here is the size of the gap between limits, and the lack of informations ..

[tame current]

it was tbh, you can't even open perplexity if you have a VPN on

[agile jay]

Yep, the main problem has been with them not being transparent with the changes.

Literally if people didn't complain they would just pretend nothing was happening.

And they won't make an announcement, because they wanna save face...

[tame current]

abuse has been a problem not just for perplexity, but other AI sites too, even chatgpt had to implement some countermeasures (they require phone number for login)

[torn sierra]

The only appropriate solution is to simply revert to what was before. Let's hope that the restrictions will be lifted soon and the problem resolved. It's a bit odd, though, that the company didn't take the necessary steps earlier. I don't know the architecture, of course, and the type of this "abuse", so maybe my judgment is unfair.

[agile jay]

They require that from the start though. To me it seems like just bad planning.

[warm cave]

When they leave us in the dark about a bunch of small changes and then take a bit on a big change, it dose not look very good, hopefully they learned that overall transparency will build trust, so even in cases where they can't notify us right then, them users won't feel like they are stabbed in the back

[tame current]

I mean yeah, but it's for preventing abuse. I'm not saying you're wrong just that it was a problem even before only thing different now is that they took drastic measure

[agile jay]

If my company made money from margins of usage, I would have put a lot more effort on stuff like stopping abuse...

[warm cave]

And then they put a chat cool down to calm the fire lol

[austere kestrel]

"The issue here is the size of the gap between limits, and the lack of informations .."
I think also that the 600 limit was arbitrary - or to the extent it was deeply calculated, that was like 12 months ago (well before Opus). that said, given the policy has remained unchanged, and Opus has been available for months now, i don't really get why suddenly this has become an issue

[opaque stump]

anyone who emailed for a refund get a response yet

[warm cave]

Takes a few days, and I'm sure they are overloaded rn

[wicked parcel]

How can you see if you exceeded the limit?

[warm cave]

I think it's in the model selector tab for opus, but for the rest you can't until you reach 100 left, unless you use this: https://www.perplexity.ai/p/api/v1/user/settings

[agile jay]

I'm pretty sure most people would rather have max 250/day on all models, than 30/day on opus.

[lone citrus]

Noooo, Opus is limited now🥺

[agile jay]

@fast meteor you actually use more than 250/day?

[tame current]

tbf perplexity is a lot more generous then most of the other

[half venture]

They are but

[wicked parcel]

opus_limit:0 ok..

[lone citrus]

What happened, why there's no notification about it?

[agile jay]

The similar business are 500/day

[fast meteor]

for some active days, yes. the root cause is that opus is too expensive.

[tame current]

dunno what that is but you shouldn't talk about it here unless you want it to get patched

[wicked parcel]

allright, I will now make an additional you.com subscription.

[fast meteor]

and i prefer the gpt4. opus is not that logical and is over trained to perform the summarization

[tame current]

and a lot worse outputs in general (assuming you're talking about you.com)

@agile jay I think that's different

[agile jay]

Pretty sure it was already mentioned. People using fake cards for the 7 day trial when paying for 1 year sub.

[wicked parcel]

Another question: Can I use other models to upload text and then switch to opus promt and use the uploaded text?

[meager sparrow]

You.com is alright it is a somewhat powerful Ai-powered search engine.. It does not have as many LLMs as perplexity but you.com has a very nice sources display..

[agile jay]

The only difference is the UI...

[tame current]

you don't really think that right ? perplexity is leagues better than you.com (in terms of web search atleast)

[upbeat tiger]

IF ONLY! it's full of bugs and the system prompt is absolute trash

[lone citrus]

You.com is good, but pplx is better on chatting. They don't provide a writing mode or something like that.

[agile jay]

I know how they work under the hood. Literally perplexity is using an indexer and then placing the matches before your prompt to use as sources...

[fast meteor]

not necessarily. they built the automation bot to interact with perplexity website to max out the 600 limits every 24hours. they might just pay the 20usd/month subscription fee normally.

[tame current]

we don't know, but it doesn't stop anyone from paying and abusing

[warm cave]

Honestly AI web search is just good now days, free bing chat, You.com research and perplexity. I often use all 3 and paste it i to a chat to have a lot of variety in specific task. I think most will like perplexity, but if co text length is everything then they might prefer you.com

[agile jay]

The main benefit of using perplexity is the nice UI, and the number of sources. Plus the usage limit.

[austere kestrel]

"It does not have as many LLMs as perplexity"
erm..

[tame current]

and you.com does the same, just worse

and the reason I use perplexity is because it's much more accurate than all the of the others

[meager sparrow]

Nah I wouldn’t say that.. perplexity is not even that detailed or proficient in their Answers. You.com is a little bit more detailed, and has a good Source display. But perplexity does offer variety in terms of LLM’s and it does provide a veritable of focuses

[agile jay]

And at full context...

[tame current]

Have the mods said if it will ever go back too unlimited? I need Claude 3 opus for college 😦

[tame current]

yes, it's temporary

[warm cave]

Honestly if you.com was bug free I would be 100%! On board, 500 a day with full context I amazing. I would have perplexity and you.com

[meager sparrow]

And sometimes perplexity just gives 1-2 sentence answers to questions which is bad..

[austere kestrel]

as mentioned, I gave it a 300k token file with Gemini Pro 1.5 and a needle in a haystack test. It found all the needles. But it's Opus/Sonnet have been patchy (returning errors - not context issues per se)

[tame current]

That's a relief 🙂

[agile jay]

Yep, that is mostly because they are using http responses instead of web sockets.

Which is why perplexity is more stable than most other services.

[warm cave]

Yeah, if it was bug free it would be an amazing option, but with the bug I have mixed feelings

[lone citrus]

Well, I agree with you, but my personal experience is that you.com simply does not provide the same answers as Perplexity AI did. Claude in Perplexity AI just followed my prompts better than you.com does.
What really concerns me is that you.com does not have a good look on the phone, compared to Perplexity AI. I would like to say that the 30 Opus messages per day are acceptable.

[tranquil orchid]

Pro search is limited to 600 queries per day

Claude 3 and GPT-4 queries are counted as uses in the same context as Pro uses. Each use is given back to you 24 hours after you use it. You will be able to see the exact count when you hover over the Pro toggle if you get to less than 100 uses left.

[buoyant kindle]

I tried the same with a couple of pages in a txt file and it pooped itself.

[austere kestrel]

i agree with this for sure, it's way behind perplexity as an actual search tool

[warm cave]

For actual web search free ms copilot I good imo

[meager sparrow]

But that’s the problem because the Answers that perplexity provides lacks so much detail. That you need to make an entire Prompt that makes it more proficient. But many times it deviates from the prompt and it doesn’t work..

[agile jay]

Don't like their UI tbh

[tame current]

not really, the sources it considers are not good, they also nerfed it

[austere kestrel]

aha fair enough - I haven't tested extensively

@warm cave i always forget about ms copilot.. useful reminder

[warm cave]

Well yeah the UI sucks, and they murder gpt-4

[warm cave]

Yeah, it good at web search and is pretty smart, but put it side by side with chatGPT and it's not as smart, fails more often in math and other problems. But for web search I love it

[austere kestrel]

yeah it's specifically the web search part that I'm interested in (and always forget about).. will revist it :))

[wicked parcel]

Annying that the you.com free month code just ran out..

[lone citrus]

Wait, mine is 30 🥺

[tame current]

So which site currently provides the best version of gpt 4 besides api and chat gpt?

[wicked parcel]

Ah, now I can see my limit.

[warm cave]

Also you have to take into account that ChatGPT has tools like code execution for math and other problem solving along with web search. Which can increase it's ability compared to other platforms. But beside that perplexity is phasing into using the latest witch one are happy about and other hate lol