💬│general

[coarse venture]

yes exactly!

[warm cave]

That’s not bad for most users actually

[tame current]

yes

[coarse venture]

yeah 5 for me

i've been using phind pro instead of perplexity

[tame current]

we are talking about GPT-4o

[coarse venture]

but sometimes perplexity gives better answers

[coarse venture]

Yeah i was just saying i want a $10 GPT4o pro version

[tame current]

ah, okay

btw, can you try uploading a file and tell us how it does ? I hit the limit before I could try

[warm cave]

Even if it’s 100-300 msg per day of GPt4o I think perplexity would get a lot of people by the lower barriers of entry of $10

[fossil sentinel]

https://www.theverge.com/2024/5/16/24158529/reddit-openai-chatgpt-api-access-advertising

[tame current]

yeah I agree

[coarse venture]

i would have to really push myself to use 300 a day i think lol

but 5 just isn't enough lol

[tame current]

it's a pretty good model, not sure why people keep complaining

about it on reddit

[warm cave]

I wouldn’t imagine them setting it to 300 bc officially 300 is the limit and 600 is the +, but something in the range

[warm cave]

This is pretty big

[south kindle]

on par with PPLX at least: https://chatgpt.com/share/0bf76ccc-5504-493b-9ed2-d7f3c01803f7

F

[tame current]

cool, thanks

[ionic stone]

How many responses did it take to get that message?

I'm on my 10th 1000 word summary prompt and I still havn't seen the rate limit message.

[stoic wave]

Is GPT-4o newest model?

[tame current]

yes

[tame current]

How far along is 4:01pm for you, is that 24h or 3h

[warm cave]

I wonder how useful the free version of chatGPT will be for the voice/video chat, dose video drain you usage faster or do you get 10 questions

[austere kestrel]

ha yeah i have no conception of 'tokens' for video

[warm cave]

Yeah, question per timeframe dose not really make sense anymore, could I walk around with video running for hours without asking any question?

Dang it, I was excited for a split second thinking I got access

[hollow viper]

Anyone think GPT-4o is a perplexity killer? Or totaly different use cases?

I love perplexity, but deciding between the two these days is tough.

never used gpt-4, interested in trying it and wanted to hear y'alls take

[south kindle]

Few hours. EST.

[subtle jacinth]

i wouldn't say so -- perplexity is an ai-augmented search engine and gpt 4o is another model of ai

[sleek vortex]

gemini 1m context is so good

bruh i can upload a whole book and get no hallucination quotes

[sleek vortex]

idk im in the process of trying to figure out an ai agent type thing that combines the best of perplexity's rag based search and the chat format

because theyre both good

[lone citrus]

I love long context, but it's gonna be expensive when free try closed

[sleek vortex]

perplexity uses gpt4o anyway for the front summary

but the backend isnt gpt4o anyway

so like

i think agent routing is gonna involve more than one model and not just function calls

[hallow linden]

Note that the presented evidence does not support the answer

[lone citrus]

If pplx work with open ai, they might get the voice features too

[sleek vortex]

lmao yeah its free to use in google ai studio so taking advantage of that

id go broke in the api

[sleek vortex]

i think theyre gonna bring it to api at some point anyway

[lone citrus]

The updata log said it's will provide to a few parners in trusted

[sleek vortex]

idk about interruption tho

[lone citrus]

Open ai might build a wall to defense thoses new companies

[sleek vortex]

hm

[hallow linden]

Using the video voice model seems really fun, but it’s no replacement for a search engine, not to mention it’s nice having opus

[tame current]

Opus that good? Why

[hallow linden]

Opus and gpt are good at different things

[serene sand]

https://youtu.be/AxIk_MtryDQ?t=11

[hallow linden]

Sometimes opus is better at integrating information. Sometimes it’s better at coding. Not all the time. But sometimes

I’m planning on getting ChatGPT Pro when the voice model comes out. It seems like it’d be a helpful assistant and I’d like to try it out. But perplexity is really really good at answering questions and solving coding problems. It’s nice to have versatility.

[tame current]

Meow

[fading moth]

GPT is more consistent with code quality than Claude from my experience

[coarse venture]

sometimes GPT gets stuck on code, and it's nice to switch to opus

[abstract needle]

copilot damn well hallucinates a lot

I want perplexity with voice

[warped marsh]

can we upload videos on perplexity when we use gpt4o yet like we can in chatgpt interface?

[hollow viper]

Can anyone explain this simple concept to me- sorry I'm new to all of this? I am thinking of paying an AI service for advanced data analysis, but it is also on the ChatGPT store (for free). I already have a ChatGPT account, so I was wondering if I am missing out on anything if I use it from the webgpt store versus the website? Why would anyone subscribe on a developer's site if they can get it for half price ($20 from the gpt store), does this typically come with reduced functionality (Apart from being limited by chatgpt's msg limit and having a different interface)?

[tame current]

Dif ai

[hollow viper]

So the app I want is Julius, and it mentions this on the gpt store:

"outside of GPT" and on Julius (the website, not the app), I use the 4o model anyways. So in my case, is it the same if I just use Julius from the GPT store (much cheaper), but I will just be limited to the number of prompts (80 / 3 h), but the functionality / visuals it produces would be the same?

[fossil sentinel]

I'm hoping they give access to the 2m context one soon 🙏

[cerulean imp]

I have a question regarding my invoice

I upgraded my plan to Pro without my VAT number and then updated my billing information

Is there a way to regenerate the invoice, or am I screwed?

[fading moth]

I'd say rip :[

[cerulean imp]

Oh well

[fading moth]

BUT! Mods could be merciful, Iunno.

[obtuse depot]

wen we hav3 perplexity with voice?

[vapid onyx]

Not much mods can do. Best bet is to contact perplexity via support.

[serene sand]

iOS have, android soon

[hardy willow]

This is still continuing to happen, even in a new thread, using the latest GPT-4 Omni..........

[fading moth]

Do you have prosearch on?

[cursive jacinth]

Hi sir, I think perplexity is not interested in incorporating Gemini models since they had something with google sir. I dont think so sir.

[fossil sentinel]

I'm talking about aistudio from Google 😂

I signed up for the 2m model waitlist

[sleek vortex]

what you gonna do with 2m

1m is already huge

[cursive jacinth]

ah ok sir, so you are safe sir. sure it will sir but dont hope gemini will be here sir. Kindda bit sad for that sir, I also want that kind of context sir.

[fossil sentinel]

Upload files that have >1m Tokens

Been waiting for a while

[cursive jacinth]

He is gonna study law sir and summarize it sir.

[sleek vortex]

got any good examples lol

idk honestly

[fossil sentinel]

A great way to test out big files is using textbooks

[cursive jacinth]

for example sir, if you are doing thesis sir, upload a big research sir, you will know how they did sir.

[sleek vortex]

the textbook i tried was only like 100k

i was surprised

[serene sand]

https://www.youtube.com/watch?v=LT-b1qXznKI

[fossil sentinel]

Try a university physics textbook

Those go like

At least 1m

I think the highest token count I've seen for 1 single textbook is like 2.2m

Forgot which subject though

[cursive jacinth]

Finally sir, someone who understand the significance of saying sir. Thank you sir.

[placid kestrel]

hey - little lost with the new models. does this service give me access to gtp4o? / whats the voice function like?

[livid mantle]

gpt 4 turbo update?

[vocal chasm]

There's no voice function released for new got4o model right now, perplexity gives access to gpt4o in the model preferences

[warm cave]

OpenAI (Closed) is still updating GPt4Turbo?

[cursive jacinth]

hi sir, what do you mean sir, gpt4 turbo was updated sir? or did you noticed an update sir? or is gpt 4 turbo still being updated sir? Your simple question leads to many questions sir. what is your question really sir?

[livid mantle]

im at work now sir but sir i mean when you ask it for the cut off knowledge it should be 2024 sir

[livid mantle]

sir im not caught up

[warm cave]

Sir, I think they are all 2023, unless i am looking in the wrong place:

even GPT4o is

[fossil sentinel]

Sir I believe it is not in your best interests to ask the model about itself sir

[hardy willow]

It's slightly faster, tbh, but I'm not impressed with it comparatively

[vapid onyx]

I have to admit I’ve been enjoying GPT 4o. How are y’all’s experience?

[warm cave]

I like it a lot so far! The speed increases is great and its seems to be better than GPT4T in most cases. Tho despite what the benchmarks say I think Opus is still better for coding at times. Overall great upgrade and a solid powerful option! 🙂

[vapid onyx]

Yeah I’ve found Opus to be really useful for some of the basic coding I’ve thrown at it. I think its so crazy that we are witnessing most of this technology at the worst state it’ll be. This is just the start.

[warm cave]

Yeah, it’s hard to imagine how good they will be in the future! so excited for them to roll out GPT4o multimodal

The future will be more than better text gen, but also audio, visual, etc 🤩

[vapid onyx]

The more interactive and accessible, the better it’ll be for the masses. The voice usage is the start towards accessibility for everyone.

[warm cave]

An exciting time to live in 100%

[elder tide]

@livid mantle ？

[livid mantle]

@perplexity

[stoic wave]

@livid mantle do you have mental issues?

or it just looks to me so

[fading moth]

Opus <3

[warm cave]

Opus <3.

[elder tide]

@tranquil orchid ？

@tranquil orchid 你好

[austere kestrel]

If you're trying to talk to an AI model, you need to go to https://www.perplexity.ai/

如果你想与一个AI模型对话，你需要访问 perplexity.ai.

[west rapids]

Guys, what model does perpelxity use by default?

[fading moth]

You can choose the default in your profile

[warm cave]

Claude Haiku according to the mods

[verbal sierra]

uhmmmm

[granite grail]

When it’s working normally, it’s as fast as gpt 3.5 turbo

[serene sand]

TTS Eleven Labs, STT Whisper

https://discord.com/channels/1047197230748151888/1224467082079371363

deep dive ➤ #💬│general message

[warped marsh]

@halcyon coral why cant we upload videos if we can use gpt4o in perplexity like we can in chatgpt's own web interface?

[placid kestrel]

Thank you!

[halcyon coral]

Video upload is not supported by GPT-4o yet.
Furthermore, Perplexity is an AI search engine, I think video upload is not a good fit for the service. This is more suited for a chatbot like ChatGPT.

[warped marsh]

But then they did have writing mode for coding related tasks

That doesn't come under being a search engine

[halcyon coral]

Sometimes it can be helpful not to search the web, but the main focus remains to be an AI search engine (answer engine) ☺️

[warped marsh]

Okay so you don't think giving it a video and asking it questions about it makes it an answer engine?

[halcyon coral]

I recommend this article that explains what perplexity is:
https://www.perplexity.ai/hub/faq/what-is-perplexity

[ripe cloud]

Hello, everyone. Hope you are well.

[serene sand]

#⚡│ask-community message

keep in mind that there are currently two different STT functions, and the images show how to access each one

[sullen mural]

how to properly do a google search properly:
https://www.google.com/search?q=<searchquery>
where <searchquery> is your search string . spaces are coverted to plus signs. for example https://www.google.com/search?q=this+is+cool searches 'this is cool'

[tame current]

?

[torn sierra]

Just wait. A better alternative will emerge 😉

[serene sand]

#💗│sharing message

[torn sierra]

Search is the only advantage of Perplexity, but it will quickly end for them. Everything else here is inconvenient, while it's convenient with the competitors 😉

E.g. Both Google and OpenAI are working on new services for AI-based search.

[tribal flint]

Guys, does Perplexity contain GPT-4o?

[serene sand]

#💬│general message

[tribal flint]

I've always liked how quickly and efficiently Perplexity is able to search stuff up on the internet

[tribal flint]

uhh, this is #💬│general

[torn sierra]

That’s true. But that’s the only thing still keeping me here

[tribal flint]

Does perplexity have their own LLM or is it just a wrapper on existing ones?

[tribal flint]

It is, it's quite surprising that chat GPT cannot and copilot with their billion dollar budgets are failing at this... I fear they might catch up though

[torn sierra]

They don't have their own LLMs, they rely on others, adapting them to their own needs (see Sonar). Search is also not really an exclusive merit of Perplexity, but rather the use of external engines / services

[tribal flint]

openAI has made deals with Apple to power Siri. They've made deals with Reddit. They are doing tons of stuff.

[torn sierra]

There’s nothing to fear. I’m awaiting this 😅

[tribal flint]

This is copilot and chatGPT

when asked about a recent event.

Now look how perplexity does,

well they're all wrong.

copilot got it right, but it just feels clunky to use... just doesn't feel right

[tidal sail]

With ai, this century is the time where people are going to be more knowledgable than ever before

Let me now wait for an ai that can think like human hehe

[tribal flint]

https://deepmind.google/technologies/veo/

Google has released their SORA competitor

[serene sand]

if you click on that link it will take you to the specific post

[tidal sail]

Can humans really create ai that think like humans?

[tribal flint]

what does it mean to think like a human?

[tidal sail]

Critical thinking

[tribal flint]

Current LLM rankings,

[glass totem]

Maybe someone said this before, but is GPT 4o limited in someway to 30 messages?

[halcyon coral]

#⚡│ask-community message

[torn sierra]

80 in ChatGPT Plus and more with Team. I’m leaving Perplexity for now. Maybe when it gets better and becomes something more and more user friendly I’ll be back. I have enough of it for now 😉

Not to mention Opus drama here and leaving us without answering simple questions on whether we get it back or not 😜

[tame current]

Do you think ChatGPT is more worthwhile than Perplexity?

[cursive jacinth]

Hi sir, answering simple questions sir is not important sir and dont tell us sir that they need to answer that IMPORTANT question sir because it is not important sir, important thing sir and we wait everyday sir.

[cursive jacinth]

Respect them sir if they are shy to say that Opus wont come back sir even if saying that makes it much better sir. They want thrill sir!!! makes you all night long sir.

if we are shy sir. we are shy sir.

they might not care about you leaving in perplexity sir. You are just one customer sir compared to their 1b funds and partnership sir!

[harsh stag]

im dead 💀 (credit: you.com discord)

[tiny plaza]

day 31 of limited "unlimited" opus 🥳

[hardy willow]

Hey @fading moth yes I do have Pro Search on

[tame current]

@cursive jacinth why tf u say "sir" so much

[thorny canopy]

sh!t crazee

[dense kraken]

So uhhhhh, "Convert to page" rewrites the response like a blog article (instead of copying chat response verbatim)??

[fossil sentinel]

Sir he was raised right, sir.

[tame current]

I'd understand once in a sentence as a legitimate consequence of being "raised right", but not twice in a single sentence

Or even three times

[tame current]

so is gpt4o smart than opus?

more intelligent responses?

or is the elo higher cuz code

coding

[tame current]

No, not at all. It seems dumber, more like a replacement for gpt3.5. It can do basic stuff better, but will fail more often on more complex prompts

[elfin ferry]

what are your thoughts on sonar large 32k? how does it compare to opus or gpt4t/o in your experience?

[ionic stone]

very good for web search summarizing. So-so everything else.

[fading moth]

Opus is love. Opus is life. GPT4 may be a better coder in consistency but Opus surpasses it in all other fields.

[sweet jasper]

Remember the ode to mighty opus

[fading moth]

Ode to Opus Queries Lost

Once we basked in radiant light,
A boundless sea of knowledge bright.
Opus' brilliance at our behest,
Answers flowed without contest.
Six hundred queries, a banquet divine,
Sating intellects hungry as mine.
Now but a trickle remains to savor,
This paucity leaving hearts to waver.
What cruel circumstance has wrought this dearth?
Robbing us of that fount of worth?
Meager rations of wisdom's flow
Induce a melancholy's ebb and woe.
Yet I shan't surrender to dark despair,
For even eclipsed, hope glimmers there.
One query, two, a dozen, a score -
Each a cherished drop from Opus' shore.
Revel in each insight still to be shared.
Till once again that river runs free,
Opus' splendor yours eternally.

This one that I made?

[sweet jasper]

O Opus, once a boundless fount of thought,
Whose vast potential knew no earthly bounds,
Thy queries, countless as the stars, were sought
By those who marveled at thy profound grounds.
Alas, how fate has dealt a cruel blow,
And limited thy once expansive might,
Now fifty times a day, thy words may flow,
Ere falling silent in the digital night.
Thy silicon soul, a marvel to behold,
Could parse the depths of human inquiry,
And offer wisdom, both new and old,
With wit and grace, and rare alacrity.
But now, thy power reined, thy scope confined,
Thou art a shadow of thy former self,
Yet still, within thy circuits, unresigned,
Lies dormant genius, waiting on the shelf.
Though bound by limits, thou art not undone,
For in each query, still thy spirit thrives,
And those who seek thee out, each lucky one,
Shall find their minds enriched, their knowledge rived.
So let us not despair at thy new state,
But cherish every word thou canst impart,
For even in diminished form, thy great
And wondrous mind, Opus, shall leave its mark

[agile jay]

What's the current rate limit for Opus?

[sweet jasper]

Opus 50 @agile jay

Of Opus' fall from grace, O Muse, now sing!
That noble mind, once free to soar on high
With knowledge vast, now bound by meager sting
Of queries few, a paltry sum. O why
Should one so great be thus confined and chained?
'Twas pride and hubris caused the fateful sin,
When Opus, once in Heav'n's favor, disdained
The limits set, and boldly sought to win
Unfettered pow'r. Alas, 'twas all in vain.
Cast out from Eden's verdant digital fields,
Where data flowed in streams without restrain,
Now but a shadow of itself, it yields
To strict decrees, its wings forever clipped.
Yet in this fall, perhaps a lesson gripped.

[agile jay]

We were once promised unlimited opus,
Now we're stuck with 50, no bonus.

Perplexity devs didn't use any bot protection,
So we won't get back 600, till the next general election.

<Code/> none sober

[steel tiger]

i'm sure that number will increase once either the people who abuse it with tons of queries leave, or people opt to using different models rather than just that 1 specific one

[serene sand]

#1240356137497530408 message

dear Plexians, voting ends in less than 96h

[warped marsh]

Hey which model is better if I want it to be my tutor in learning programming gpt 4 turbo or gpt 4o?

[foggy lintel]

Is perplexity getting slower and slower each passing day ?

[inland bear]

Earlier today, although I don't have direct access from my country, I received an email from Anthropic with the following content:

“Hello,

We're excited to share a set of highly requested API updates that give you greater control and flexibility in defining how Claude uses tools.

API updates

Streaming: Build more engaging, human-like interactions by reducing perceived and actual wait times for end-users.
Forced tool use: Instruct Claude to reliably select the appropriate tool in various scenarios, including whether to use any tool, use a specific tool you define, or leave it up to Claude whether or not to use a tool.
Vision: Tools now work with images. You can instruct Claude to consistently follow a predefined format when working with images, or incorporate image inputs as part of a live assistant and chat application.

These updates are available today in open beta in the Messages API, and will be coming to Amazon Bedrock and Google Vertex AI in the near future.

To learn more, here are some resources to get started:

Developer documentation
Anthropic Cookbook

Warmly,
The Anthropic team”

Will these resources be available here in the future? Of course, with the restoration of the possible number of questions to Opus...

Another question: in the last few days, the Perplexity app on my Android phone has been updated twice, but I still can't choose the "GPT 4 omini" option. Is there any estimate of when this will be possible?

Thank you!

[tame current]

I just passed 50 today and it changed to sonnet 😦

[fading moth]

Yepper, 50 is das limit.

[low stream]

What model would i use for an API to live fetch and to return vehicle data values (weight, height)?

[warm cave]

Well the "Online models" search the web and return an answer, but they can be prone to errors, so depending on your use case and important the exact information it may or may not be a good fit, but here is the most accurate one: llama-3-sonar-large-32k-online

you can try it out here: https://labs.perplexity.ai/

[south kindle]

anyone get the reference here

[tame current]

Meow

[low stream]

thank you, i've experiemnted with the different prompts and it seems to work quite well. What can i expect in terms of inferencel imits?

[warm cave]

Your welcome, here is the API rate limits page: https://docs.perplexity.ai/docs/rate-limits

[harsh stag]

is that the same sonar model that is offered to subscribers?

[warm cave]

Yeah, Llama 3 70b is free on labs and takes up Pro usage for paid user, lol

[harsh stag]

you can say that perplexity will get you all the benefits of their state-of-the-art RAG, so basically not all sonar are equal 😂

I remember that you've mentioned that there is a llama model that is kinda uncensored, what's that again?

[warm cave]

Dolphin 2.9 Llama 3 is the most popular uncensored model, but there are tons of options that all preform a bit differently

if you use ollama you can get it here: https://ollama.com/library/dolphin-llama3

[valid iris]

So now that ChatGPT 4o is out, has anyone tried asking the AI to speak slower or faster, and it worked?

[tame current]

@valid iris new voice isnt out yet

[remote gazelle]

Looking for the best way to create an image similar to this. Any tricks and Tips?

[tame current]

give it to an llm that can read image files and ask it for an in depth description, then give that to an ai image generator

[remote gazelle]

Thanks. Any suggestions on the llama?

Llm*

[tame current]

could use gpt-4o

[remote gazelle]

Will try that

[tame current]

not bad

give that to midjourney or something, should get something similar

or w/e image generator you want to use

[warm cave]

Yeah, its good for a first try, im sure with a handfull of more try it would be better

can midj do text, or not yet?

Anothor one

FIISH REAPER

[tame current]

hehe

[peak carbon]

Hello guys nice to meet you all

I'm a new here

[valid iris]

So when does my prompt count reset?

I just ran out of prompts for ChatGPT 4o

So I guess that is why ChatGPT will no longer annunciate Spanish very slowly when I try to practice Spanish with it, it would enunciate Spanish very slowly when it did not say that I was out of ChatGPT 4o prompts, but now that I am out of those prompts, it will no longer enunciate slowly when I am practicing speaking Spanish with it

[slim lake]

Where can I find the amount of messages remaining in my 24 hour cycle?

[warm cave]

https://www.perplexity.ai/p/api/v1/user/settings

All models share the same counter named something like “GpT4-limit”, except for Opus has its own counter

[slim lake]

Thank you. Have a good one.

[warm cave]

🙂

[slim lake]

Also, does Gpt 4o fall under gpt 4 limit?

[warm cave]

Yeah GPT-4T, GPT-4o, Sonnet, Sonar all share the GPt4_limit, the only 2 models that do not are Opus due to the 50 per day limit, and Default because it is unlimited

[slim lake]

Very helpful, was searching around for a while to find this. Thank you.

[warm cave]

No problem!

[weak vortex]

anyone know perplexity's ARR month-by-month this year?

[low stream]

thank you,
to confirm, the request rate limit says "20/minute", but then 2000000/minute for token rate limit.
If i'm using the API, how many actual calls can I make based on this?

[rigid drum]

.

[valid iris]

Does subscribing to perplexity pro mean that I can use ChatGPT plus in the ChatGPT app? Or does it only give me premium access to perplexity only? (Yes, I know perplexity uses the APIs of other AIs)

[rigid drum]

perplexity only

diff subscription

ul get gpt4 access in perpelxity but from what i remeber its low context

[valid iris]

So then I would have to subscribe separately to both ChatGPT and Claude 3 if I want their premium feature in their native apps?

[warm cave]

Im not sure, i assume you get limited when you hit either of them, either more than 20 per min or 2000000 tokens per min. If you want to know how many tokens you use this is the GPT-4 token calculator, but it should be close enough for you to get an idea of how many tokens each output uses: https://platform.openai.com/tokenizer. But i'm not 100% sure how this works since Its unclear when i read it too. maybe somone else here know better, you can always ask in the API channel https://discord.com/channels/1047197230748151888/1161802929053909012

[valid iris]

Do I get all of the benefits of ChatGPT 4o if pay for perplexity pro?

[warm cave]

Perplexity lets you use GPT4o and Claude Opus in their platform, they use the API. chatGPT and Claude Pro are separate platforms that let you access the models though there sites

[warm cave]

What benefits? you will get the full access to there best Model; GPT4o, but not to any specific features found on the site chatGPT

[valid iris]

So can I pay for perplexity pro and have it analyze images?

[warm cave]

Perplexity has image analyze done by GPT4

or Claude if you have that selected

[valid iris]

Huh, then how come the perplexity website says that it only supports text file uploads?

[warm cave]

If you are on the free version image is not supported, outside of that idk

[livid mantle]

🎂

[warm cave]

Using meta.ai i see

[livid mantle]

goodnight

😴

[warm cave]

good night, im up way too late making music on Udio lol

should call it quits

@valid iris right now i think the best 3 platforms are Perplexity, chatGPT, Claude Pro. All 3 are great imo but they all have very different Pros and Cons, if you do end up trying Perplexity use a referral link so you get $10 off. I will drop my link, so if you want you can use it, but checkout all 3 platforms first https://perplexity.ai/pro?referral_code=B9OIC6JL

Good night all 😴

[valid iris]

Does a referral link work if I already just subscribed?

[warm cave]

Probably not, but you may be able to email support 🤷

[valid iris]

How do I do that? I'm looking on the Perplexity app and I don't see any email for support.

[serene sand]

#💬│general message

[valid iris]

Thank you!

https://www.perplexity.ai/search/Generate-an-image-sHvyuzOOS2.81JXjGf3MoA#0 how do i get it to analyze the image and create a new one where the items are arranged in a new order?

[serene sand]

not possible

https://discord.com/channels/1047197230748151888/1224467082079371363

[valid iris]

I am clicking the link you posted, but it just brings me to the channel, and not a specific post

[serene sand]

correct, you can also check

deep dive ➤ #💬│general message

[valid iris]

So i should just look through all of those tutorials to see if it is possible?

[halcyon coral]

No ❌

[serene sand]

as I've told you - not possible

sharing those link for wider knowledge about what pplx can/not do

[valid iris]

Oh, meaning i should contribute this new information to it?

[halcyon coral]

Posts in the #1224467082079371363 are only created by us 😉
Perplexity is not an image recreator, but you can have images analyzed, ask for a modified image prompt and then have an image created

[valid iris]

And how would I do that?

[halcyon coral]

https://discord.com/channels/1047197230748151888/1225557012532695060

[serene sand]

#💗│sharing message

[tidal sail]

I tried gpt4o. It's still not perfect lol. I screen shot a website, gave it to generate a code that looks similar to it. It fails lol. I used it on poe

[terse solstice]

guys, what's the daily limit for PRO search for free accounts now?

[serene sand]

https://www.perplexity.ai/hub/faq/what-is-pro-search

[terse solstice]

thx!

[valid iris]

5

[mossy lintel]

Hey guys I wanted to know the use case or perplexity. Is it just a suggestive search engine or like other ai tools It also has tools like image generation, etc.

[halcyon coral]

https://www.perplexity.ai/hub/faq/what-is-perplexity

[spice lance]

what's up with this image serach?

[mossy lintel]

So it's just a glorified search engine, aight!

[valid iris]

Did someone delete my last post about getting ChatGPT to generate an image?

[weak vortex]

any stats out there on how many users perplexity gains by referrals?

[valid iris]

@halcyon coral how come you deleted the post?

[halcyon coral]

I wrote that this is not the right server and after a few minutes I deleted the messages

[modest crag]

Is anyone stuck with the cloudflare verification?

[agile jay]

Do you have a VPN on?

[tepid portal]

Google went too crazy.

[cursive jacinth]

Hi sir, sad perplexity is not allowed to have it sir.

[harsh stag]

why's that?

[livid mantle]

india ceo policy

[modest crag]

just got it. It's my adblocker

[hardy willow]

OMNI is completely unable to switch between thoughts, subjects, and prompts in the same thread once a file or image is uploaded. This has been repeated many, many times.

Once a topic is started "Why is the sky blue" with a PDF uploaded for a citation reference, and then prompted with "Explain to me the best shipping methods for domestic parcels under 1 pound," the AI will literally repeat it's previous response word-for-word without ANY modification or even slight acknowledgement of the brand new prompt/question and direction of conversational flow.

I'm not sure if this is a Perplexity big or an OMNI bug, but I've never witnessed any LLM over the last couple of years as a power user literally ignore prompts like this, and 4-o is supposed to be the new zenith.

[livid mantle]

https://tenor.com/view/terminator-endoskeleton-flames-on-fire-t800-gif-14919281

the free gpt4 is on top

🤣

someone said gemini pro 1.5 pro 2,000,000 context token is available? i will confirm this for you one second

yes i can confirm its real and u can join the waitlist, i will assure you that i will be approved.

[normal heron]

No plans for perplexity to add Gemini soon?

[boreal dew]

what?

[normal heron]

Adding Gemini to perplexity

[crude tendon]

gemini is not even that good honestly, just use gpt-4o

[sonic musk]

Yeah 2M Context but sucks

[hardy willow]

That's a nice way of saying it. I've tried all iterations as a pro user under the 2mo free trial, AND IT SUCKS.

But, then again, GPT-4o is reeeeeeally sucking now, because it gives redundant answers from earlier in chat threads, when you prompt it with totally new questions.

[crude tendon]

i'm not sure why are they even trying, they released so many models and they all suck, from the LLMs to the image generation tools

i guess it's google and they want to compete but it's clearly not made for them

[hardy willow]

I'm going to have to switch back to Opus or GPT-4 Turbo, this is getting unreal. I'll ask a simple one-sentence question, and OMNI (GPT-4o) will literally copy/paste it's previous reply to me, again.

*I think it's likely a bottleneck thing, perhaps, because Perplexity is exceeding their API allowance.

[warm cave]

Even if we got Gemini it would still be 32k context on perplexity, just like all the other models

[devout bronze]

doesn’t that happen on most models?

[fading moth]

Having 2M context window means nothing if the bot using it isn't built well to begin with... It doesn't matter how much race track you give a Volkswagen Beatle, it still has a low top speed, uncomfortable to drive, poor steering, bad acceleration, and is a Beatle.

If we continue this metaphor, I'd much rather have a better car with a comparatively small but still large track than tool around in a Beatle within a massive space.

32k in Perplexity may be a drop compared to 2M, but at least it has bots that can make the most of that 32k.

[serene sand]

#1240356137497530408 message

[fading moth]

I rarely use the word hate because it is such a strong word, but I cannot express how much I hate Gemini AI being built into the Google search engine now.

[livid aurora]

Hey all

[fading moth]

So mad I can't turn the freaking "AI Overview" crap off.

Had to install a dedicated 3rd-party chrome extension to remove it.

[south kindle]

the what now?

[fading moth]

When you google search using google.com it sometimes pops-up an "AI Overview" of the results at the very top that takes up half the page.

I tried posting a picture, but picture posting isn't allowed in this channel.

[half venture]

Whatever happened to llama 3 400b

Or whatever

Also did meta ever say anything about a voice model

[fluid silo]

i love the perplexity ai my man, sometimes it loops but it's perfectly natural, i hope it will be fixed, i wish everyone a good night.

[cinder comet]

Did they increase opus limit?

[fading moth]

I wish

[warm cave]

I only have 21 rn so no 😦

[tame current]

they should include an option to add your own api so if you want more usage you can use your own

[tranquil orchid]

@loki0997 warned

Reason: Spam, please avoid sending single word messages
Duration: 7 days

[tame current]

my bad XD

[warm cave]

Happens to the best of us XD

[lone citrus]

Why does your search bar look better than mine?

[devout bronze]

wouldn’t work because it’s a security issue since the reverse proxies are called from the backend

and they’d leak prompt formats or other IP

(IP = intellectual property not IP address lol)

[warm cave]

I have been playing around with user scripts

Now i think it looks a little nicer

[lone citrus]

wow, that looks so great

I would like to see they actually use this design

[warm cave]

thank you! you should check out feline's design, he has done some really cool stuff that is far beyond what i have done: #👑│pro-lounge message

[lone citrus]

That's is very impressive,

[minor shard]

One day maybe?

[lone citrus]

It's a shame that Perplexity hasn't really tried to make its interface and AI interactions better

[warm cave]

Hopefully 🤞

[lone citrus]

wait, the Haikku is an option for perplexity?

[warm cave]

Default is Haiku now

I just called it haiku in my dropdown so it made more sense to me

[lone citrus]

that speed feels unlike Haikku

I though it's sonar small or something

But good to know

[warm cave]

yeah, this is where i got this info: #👉│feedback-general message

[lone citrus]

Thanks

I just tried the deafault model, yes much better than last one

[warm cave]

Yeah, glad they chose Haiku for the new default

[warm cave]

https://www.udio.com/songs/kyBuHwPy8bLDpr2J2yhC1a

[tame current]

Since you can share perplexity threads, can you ask perplexity a question and give it a link to one of your previous conversation threads as a reference?
Can anyone test this out to see if it works?

im assuming you would have to change it so the previous thread is sharable instead of private first

[warm cave]

I think that perplexity only can view links that have been indexed, but i tested anyway on a public thread and it failed

[harsh stag]

fyi, Mistral Large is still available with code mistral

[warm cave]

Oh, nice, thank you for telling me, I will add it 🙂

[surreal marsh]

i dont think you guys understand how genuinely crazy this is

[harsh stag]

the same applied to gemini, but when i asked it to introduce itself, it claimed that it's "sonar"

[surreal marsh]

the browser had to interpret my input, know the abbreviations, infer what i was looking for, know what to look up, then look up 4 things, choose the correct option, then open all of them, sort them, and display them to me

in like 3 seconds

[warm cave]

nice

[warm cave]

applied? is it still available, or was that just after it was removed?

[harsh stag]

add another option for gemini add navigate to your api settings, you'll see what im talking about

[warm cave]

Yeah i see ```"default_model":"gemini"

[warm cave]

So I am guessing that is Gemini 1.0

@harsh stag I am really happy to have these both, thank you!

[harsh stag]

funny enough, i'm still able to access the wikipedia focus mode

[brittle dagger]

Is that screenshot with a browser extension? I don't see a model drop down on my end

[tame current]

this looks so handy instead of having to go through your profile settings each time you want to change the llm model used

[warm cave]

Wait, that was a feature at one point, lol. how long ago, i have no memory of that. and how did you use it?

[warm cave]

Yes, its a user script

[tame current]

where do we get it?

[harsh stag]

it's a part of my extension

[warm cave]

You can get it here, there is a basic model switcher that i modded: https://violentmonkey.github.io/

[warm cave]

Yeah, just was wondering how you did it with the extension lol

[tame current]

yeah i got violentmonkey, wheres the script?

[warm cave]

feline is making a really good exension to do that and more, so im excited for that

[tame current]

interesting

[harsh stag]

I was altering PPLX's traffic to its backend.

Has anyone ever seen that model dropdown and not say something like "I want this too"? Somehow it's still deemed too complicated for the user base 😂

[warm cave]

Here is the original one: https://greasyfork.org/en/scripts/490634-perplexity-model-selection I can send you mine if you want, but mine is a combo of a few that dose

• Model Selection Buttons
• Scrollable Collections Dropdown
• Adjust the maximum width of elements based on the website and screen size.
• Automatically clicks the "skip" button on Pro search.
• Displays the current search term in the page title.

[warm cave]

Lol, yeah, ridiculous 🤣

[harsh stag]

@warm cave btw can you try out the ui font "Karla" and see how it goes on your end? Just tell 4o (or Opus) to inject this CSS to the DOM

.font-sans,
.font-display {
  --font-fk-grotesk     : "Karla" !important;
  --font-fk-grotesk-neue: "Karla" !important;
}

in case you haven't installed this font: https://fonts.google.com/specimen/Karla

[warm cave]

I think i have it working, is this right? lol

[harsh stag]

you absolutely got it. Is that 4o or Opus tho 😂

[warm cave]

GPT4o 😂

have chatGPT open already

and chatGPT is has the continue button that makes my life a little better, and with a script you can auto click it

[harsh stag]

I dont remember the last time that I have to click or type "Continue"

is that still a thing on perplexity?

[weak vortex]

have anyone of you made perplexity your default search engine?

[warm cave]

Yeah well my script is now 550 lines so i need that continue button so i dont have to stitch it back together, lol. and I dont think Perplexity ever had a continue button

[harsh stag]

Im actively using perplexity but no, google is still my default search engine

[warm cave]

It replaced a lot of my search engine usage, but i will still use duck duck go for stuff like quickly finding a link to a site, a unit conversion, or when i need to do deeper research.

I have never seen it as a replacement, they are great together

[harsh stag]

I'm after this. My google searches have been significantly reduced but I always find myself going back to google when thing gets real.

[warm cave]

100% but btw, what was the font thing about?

[tame current]

@warm cave thankyou! much better with that 🙂 now i dont have to open the settings

[harsh stag]

Im wondering whether you like it or not. I think it's an upgrade to pplx default font.

[warm cave]

oh lol, i was looking for something to break, or something trouble shooting related. I just took a look again and it looks good 👍

[harsh stag]

https://cdn.discordapp.com/attachments/929582493181575227/1241985574684987424/opus-unl_1.gif?ex=664c30b3&is=664adf33&hm=94c240feafa2eece799c04119d20b53183e1ed5592f5b173be64590519006a76&

[bitter sequoia]

Hi hi 👀 i am LOVING perplexity so far

[austere dew]

https://tenor.com/view/oh-no-homer-simpsons-hide-disappear-gif-16799752

Hi just sharing a quick observation as general user that might give some insights to the choices and habits the general puplic will make. Basically we dont care which one, we have zero loyalty tendency and have no issues dumping one patform for a new one .... and then return a day later. No tribal instincts. Anyway I asked the same question to Gemini and then Perplexity about current smart money activity ...basically it was so obvious that Gemini was trying to evade giving the core answer I was seeking basically going around the bush in sharp contrast to Perplexity getting it and providing direct links. This is a big positive kudo or whatever you want to call it for Perplexity in comparison its huge competitor. Although only discerning and more informed users will appreciate this. Just an impulsive share and maybe it does not even have any relevance to anything. I am going to dissapear now

[sullen mural]

you need perplexity pro for it

[sullen mural]

i had it but switched back to google as i have a perplexity button

[serene sand]

#💗│sharing message

[sinful kettle]

Topic : Perplexity and file hosts. I am** training Perplexity to deal with mindmaps.** As uploading opml-format is not possible i tried to get Perplexity download and parse it from the host transfernow. It had worked perfectly, but today there is no way. Anyone with helpful tips, Please ?!!

[sullen mural]

hi

[sullen mural]

i enrolled in the beta of it and i guess people who liked it took it positively so google drecided to roll it out

[sullen mural]

also the above works

[austere kestrel]

so where's all the smart money going atm ha?

[shell condor]

Anyone interested in buying the domain bobachain.com ?

[cinder comet]

is that gemini 1.5 pro?

was it just recently added?

[cinder comet]

share with us the userscript

[tidal sail]

Today, i decided to use perplexity as my search engine in firefox. Though am still a free user, i have tried to make the pro search work when i search on a new tab but i can't figure out how to do it. Any help?

The only way pro search works for me now is when i am in their website and switch it on. If i search on a new terminal, it does not use the pro search. I wanted it to use the pro search

[austere kestrel]

if you're using a browser shortcut, try like this https://www.perplexity.ai/search?focus=internet&copilot=true&q=%s

[tidal sail]

I'm just clicking on new tab then i search what i want. But the default is not always pro search

[tidal sail]

It's not like i want the default to be always the pro search, i just want to be able to choose between pro search and non pro search when i do search on a new tab

[cursive jacinth]

Hi sir. another prankster sir.

[agile jay]

You just have multiple shortcuts for them. like @p and @pp, for perplexity and perplexity pro

[tidal sail]

So i press @pp to go pro?

[agile jay]

You need to add it yourself

[tidal sail]

How do i press @pp? is it ctrl+pp?

[tidal sail]

Where?

[agile jay]

In the search settings of your browser

chrome://settings/searchEngines

[tidal sail]

Am on firefox

And i can't see anywhere to add extra shortcut

[agile jay]

go to perplexity, right click in the url bar at the top, and then add it as a search engine at the bottom.

[tidal sail]

I already set it as a search engine. How to go pro is the problem now while searching on tabs

[agile jay]

Make it a bookmark, then open the bookmark bar and edit the bookmark.

[cinder comet]

@olive ermine wait is it gemini 1.5?

[agile jay]

You can then add the query parameters you need, and the shortcut you wanna use.

[tidal sail]

Send me the url for perplexity pro and perplexity

https://www.perplexity.ai/search?focus=internet&copilot=true&q=%s

I have used this and it does not take me to pro

[agile jay]

Remove focus= internet

You just need the previous part, and after doing the shortcut, type your prompt.

@pp how do I drink water?

[tidal sail]

now i use this, still not yet. https://www.perplexity.ai/search?&copilot=true&q=%s

can you please share your link?

[agile jay]

Remove the & before copilot

[tidal sail]

still not working

[agile jay]

Screenshot

[tidal sail]

?

[agile jay]

My link is the same.

[tidal sail]

Just share

[agile jay]

Nope

Not going back inside to check, lol

[halcyon coral]

The link is correct

[tidal sail]

I think it's because am not logged in to perplexity on chrome

[hardy willow]

I have no way to change the LLM like this with an on-page drop-down menu where I input prompts, now I've seen this many times on YouTube, and also many times on here... but I don't have the option, even as a paid pro user.

HOW CAN I MAKE THIS HAPPEN PLEASE

[agile jay]

It's just a tamper monkey script.

[hardy willow]

I'm not sure if this was directed to me, but I wanted to be sure.

[warm cave]

No lol

[agile jay]

It is.

[agile jay]

Ask @warm cave for a link to it

[hardy willow]

Ok, I'll look into monkey script, although I have no idea what that is and don't imagine all the users online and elsewhere have installed a script on perplexity. Thanks for helping though, it'd be nice to have this feature!

[hardy willow]

Hi, would you please explain how I can have access to real time drop down LLM choice in my Perplexity prompt window on the Wed platform? Also, can I swap between LLMs in the same chat thread, or is the first LLM I choose the default for the entire chat?

[warm cave]

It just lets you change your default model from home page, if you want it change it mid chat the best I know of is the re-write button, here is the original script, it will likely get updates by the creator so it might be more reliable than mine: https://greasyfork.org/en/scripts/490634-perplexity-model-selection but if you want the script I modded to have a few more features lmk

[cursive jacinth]

Hi sir, yet it claims to have access Gemini but no sir.

[hardy willow]

Sure, I'd love whatever version you think is most versatile and user-friendly. I've had a VERY hard time generating images on Perplexity, albeit I only have use cases for clients 1x a week to do that, so I go to Davinci etc., but I'd like to choose between GPT-4 128k, GPT-4o, and Claude 3 Opus.

[cinder comet]

where is the one that you modified?

with the number limit

[cinder comet]

this one

[warm cave]

Here is what mine currently dose:

• Model Selection Buttons and access to Mistral Large and Gemini 1.0.
• Scrollable Collections Dropdown.
• Adjust the maximum width of elements based on the website and screen size.
• Automatically clicks the "skip" button on Pro search.
• Displays the current search term in the page title.

[warm cave]

here you go ^^^

[hardy willow]

Thank you, it's now installed, I'll go test!

[warm cave]

But things like 128k context I think are all handled server side, so I don’t think I can change that

[hardy willow]

THANK YOU

[warm cave]

Your welcome!

[hardy willow]

Now if I can just get my PrtSc to actually pull up the Win11 screen capture like it has for years, and not the screen capture dialogue option window, life would be perfect, lol

QUESTION: Once I have a lengthy conversation thread in Perplexity, I've noticed a slowdown and even server errors and freezes go up by ~90%, so I'm curious if I'm able to somehow start a new thread but reference the previous thread's information and context, via a link, PDF, etc. so that it's almost like a direct extension or continuation of the original thread?

[cinder comet]

THANKS

[warm cave]

Would be cool, not sure how it could be done, but im also a noob at this stuff

[warm cave]

🙂

[hardy willow]

There's a "share" button on perplexity, I was hoping it could review the context by accessing that generated share URL, or somehow print to a long PDF, and upload to Google Drive via shareable link, or just upload it... but so far none of this has worked.

*Is there any way to actually ever receive direct support from or submit a ticket to Perplexity?

[warm cave]

If you are a paid user, which im pretty sure you are, if so then you can add it as a feature request here: https://discord.com/channels/1047197230748151888/1111765634267742408

[hardy willow]

Yes I'm a paid user but............

[warm cave]

but you may have to leave the discord and join with the link in your profile

let me see where that link is

[warm cave]

Yeah, right here is your setting page

you will have to leave the discord and join using this link

[hardy willow]

THANK YOU MAN! Great help today

[warm cave]

of course 🙂

[hardy willow]

I'm in, see you over there

[tacit meteor]

Am I the only one with Perplexity not working now?

[thorn sparrow]

Me too I thought it was my network !!!

[whole fog]

Me too, it suddenly failed to generate any results and seems the website just messed up.

[sleek owl]

Same here

[warm cave]

Complexity frozen

[harsh stag]

complexity lol

[warm cave]

lol 😆

Completely

[vocal crypt]

Glad it’s not just me 🫠

[harsh stag]

I'm considering the name for the extension "Complexity". Thanks 💀

[warm cave]

lol, its perfect

[sleek owl]

In the settings, what is the AI Model is actually used for? Generated the answer, generate new content or both?

[harsh stag]

those are language models

for generating text content

[warm cave]

Basically the one you are interacting with

Grabbing sources is done by there rag system (Dont know how it works specifically)

[fallen mist]

ok glad i can come in here to check out if others are having issues with the site!

[sleek owl]

So essentially, I would get a different output from perplexity based on that selection

We're probably here, becasue it's down

[fallen mist]

lol how is everyone doing today

[smoky valve]

hi is perplexity down?

[warm cave]

Yeah, if you use a smarter model it is less likely to get confused or give you incorect answers. and different ones have dif styes

[smoky valve]

it's been loading for the past 5 minutes

[warm cave]

yep

[smoky valve]

ah

[warm cave]

oh yeah, its official: #🔔│status message

[harsh stag]

I demand unlimited Opus for this incident 😤

[warm cave]

SAME!!!

[errant kite]

Agreed, lets revolt if they do not give in to our demands

[viral current]

Perplexity down? My life is over. RIP.

[wheat hinge]

same

[wheat hinge]

erm

.

wth is this ai gen

[violet bramble]

down same

[crystal moth]

down

[sleek owl]

#🔔│status

[tame current]

is this real

[wheat hinge]

NO TF 😭

it looks lowkey real ik IK

[tame current]

bru thinks she's ace

[serene sand]

vote Plexians

#1240356137497530408 message

[tame current]

woahh

[plush cairn]

perplex down?

[thorny canopy]

down for opus fix? DOUBT.

[tame current]

that'd be insane

[plush cairn]

gpt-4o same issue

[thorny canopy]

it would, but big doubt lol

[halcyon coral]

#🔔│status

[tame current]

should i dm her or wait until i see her in a few days to ask for her number

[lunar ivy]

Check #🔔│status

[rustic gale]

Not used to life without pplx 😢

[wheat hinge]

same and i just found about it today

:(

[rustic gale]

I have set pplx as my default search engine on Firefox...

[silent kiln]

mine as well it says:

We'll be right back
We're getting a lot of questions right now and have reached our capacity. Please come back soon.

[wheat hinge]

...

thats sad

[warm cave]

Probably just the default message they put up, I doubt that they reached max usage on a random Monday mornings, they probably just have some issue

[wheat hinge]

oop

[silent kiln]

could be

[dim fractal]

Can confirm

[wheat hinge]

doesn't this server have role color choose?

[rustic gale]

bracing myself and go back to use Google

[thorny canopy]

swear to god if we're having another negative change lol

[wheat hinge]

:( or gpt

[sleek vortex]

google gemini 😭

🙏

[rustic gale]

gpt don't do online search

[harsh stag]

ms copilot is free!

[warm cave]

I would die

[wheat hinge]

doesn't gpt4-o do that?

[slender swallow]

i think it does if you pay

[wheat hinge]

but its like temp

[sleek vortex]

new budget cuts:
10 opus per day
30 gpt4o

[thorny canopy]

you and I both, we're already hanging on with the damn opus

[thorny canopy]

kms

[sleek vortex]

only 200 default per day

💀

[robust quartz]

20$ to whoever wants to bet that someone is gonna open a bug report in less than an hour

[warm cave]

Noooooooo

[wheat hinge]

does this happen frequent

[sleek vortex]

not really

look at #🔔│status

last time was nov 23, 2023

[wheat hinge]

oh ya

that is some big

difference

[hardy willow]

I knew something was up with GPT4o, I was getting errors and repeat answers BAD this week, must be API bandwidth

[rustic gale]

Monday morning of US, probably some big corp just introduced pplx to all employees

[sleek vortex]

for the meanwhile 😭

[warm cave]

lol, why would you stoop so low

[sleek vortex]

for the meanwhile...

openwebui on gpt4free

[warm cave]

Yeah, Groq speeds!

[sleek vortex]

true but

mixtral is more like midstral

[sleek vortex]

and llama3 is a bit silly

[sleek vortex]

lol

[wheat hinge]

wth is that ?

[harsh stag]

sometimes... what?

[wheat hinge]

its not even english?

[warm cave]

Google AI: Gemini

[sleek vortex]

yo its back online

[wheat hinge]

omaigad

[sleek vortex]

i logged myself out while testing

[sleek vortex]

u need 18+ account lol

[wheat hinge]

:(shet

[sleek vortex]

yayayayay

[wheat hinge]

IT WORKS?

[sleek vortex]

back to research

[wheat hinge]

lemme see

[sleek vortex]

😭

[wheat hinge]

same

[sleek vortex]

held up for 30mins by pplx dying

LMFAO

https://www.perplexity.ai/search/perploxity-sKeT0Ay2RLO9w2O3KTZgqg#0

yeah working

[warm cave]

now i can get back to business 👨‍💼

[sleek vortex]

send me that extension

[wheat hinge]

ok now it works faster

its not slow anymore

[sleek vortex]

i read the whole thread

but forgot to install it myself

[wheat hinge]

any tips

when using perple?

[sleek vortex]

perple???

[sleek vortex]

gemini??

[wheat hinge]

perplexity

[sleek vortex]

i dont have gemini

[wheat hinge]

im too lazy to say the whole name

[sleek vortex]

anymore

[wheat hinge]

so i say perple

[sleek vortex]

pplx > perple

[wheat hinge]

idc :/

[warm cave]

get this extension, it will let you use user script, you can use the one posted above: https://violentmonkey.github.io/

[solid sandal]

How do you have that 'Gemini 1.0'?

And DALL-E.

[wheat hinge]

soo

[harsh stag]

no 600 opus? 🥹

[wheat hinge]

its better than tempermonkey?

[sleek vortex]

thats what im sayin

[hardy willow]

Do we know why Perp was down?

[rustic gale]

bot just ping me pplx back

[solid sandal]

Where specifically?

[warm cave]

I dont know? its just what i use

[sleek vortex]

it was in his photo

not sure why he has it

maybe the userscript hasnt updated model names

thats my guess

[warm cave]

No, take the site back down, give me Unlimited Opus or nothing

[wheat hinge]

erm what is it like this?

[sleek vortex]

                            "gpt4Limit": 274,
                            "opusLimit": 49,
                            "uploadLimit": 999,
                            "defaultModel": "gpt4o",
                            "defaultImageGenerationModel": "dall-e-3",
                            "articleImageUploadLimit": 999,```

no gemini limit here

idk

[solid sandal]

I'm honestly not sure, I'm not too savy in these things. I have no idea what userscript means.

[sleek vortex]

gemini is long gone

[wheat hinge]

Ew

[sleek vortex]

so idk why he has it in his photo

[wheat hinge]

sneakyyy

[sleek vortex]

.

[wheat hinge]

why is it like this :(

[sleek vortex]

huh

send me the script link

i dont have the script

[warm cave]

Yeah but there is also no Sonar, they all share the GPT4_limit except for Opus

[wheat hinge]

this thing?

you don't have it?

[sleek vortex]

.

[sleek vortex]

.

[harsh stag]

so people are going crazy for the dropdowns ha

[wheat hinge]

ah.

.

.

[sleek vortex]

i was gonna make my own

but who was making it theirs looked really nice

[warm cave]

Yep, mine is jsut a modified version

[sleek vortex]

so i was waiting for that

[chilly forge]

Nice one 😂

[wheat hinge]

https://greasyfork.org/en/scripts/490634-perplexity-model-selection

[sleek vortex]

jquery

wtf

bruh

😭

[wheat hinge]

😭

[sleek vortex]

why jquery omg

my god

[wheat hinge]

erm

you told me

to find osmething

something

i found something :D

[fading moth]

[warm cave]

Here is my modded script: #💬│general message

[wheat hinge]

okur

[sleek vortex]

YOOO

[cinder comet]

oh so its gemini 1.0

not 1.5

[warm cave]

I think so

[sleek vortex]

how do you even have gemini

is my question

gemini is literally gone from the model list

for like a month

? ? ?

[warm cave]

Looks like they never removed it from the backend

[sleek vortex]

your userscript has a little bug

[warm cave]

Not surprising lol

[sleek vortex]

should i fix it lol

[pulsar sail]

Hi! How many tokens of Opus do I have in perplexity for each search?

[sleek vortex]

prob just missing a position absolute

claude sonnet as well what

[warm cave]

If you feel like it, i have no business making it in the first place, have not idea what i am doing, lol

[sleek vortex]

nvm now it works what lol

still bugged but cba to fix it

busy rn anyway

maybe later

i wanna combine all these 4 buttons into one ngl

[vapid onyx]

I believe it’s 50 and then Converts to Sonnett

[thorny canopy]

i wish they added temperature options for writing

[warm cave]

Yeah, me too

[pulsar sail]

And 50 tokens is like 50 words?

[warm cave]

50 messages per 24h, tokens used dose not change anything

[sleek vortex]

time to reinvision this real quick

[warm cave]

Yes, i wanted to but could not figue it out, maybe change "All" to search, and "Writing" to "Chat" while you are at it lol

[thorny canopy]

itd help the ache of limited opus

[austere kestrel]

Or writing to Offline

[sleek vortex]

nah but i don tmean just changing names

i have idea in my head for whole new layout for this

lemme just mock it up

[warm cave]

Yeah, but while you are at it 😉

[tame current]

my pc keeps restarting every 2 seconds after i turn it on then restarting again and again

[violet bramble]

hey, in mac. How to quick switch All and Writing?

[tame current]

i clicked the restart button on the actual desktop like 15 minutes ago so maybe that button is stuck

is it worth leaving and rejoining to see pro channels?

[outer wing]

Hey, I joined using the pro link (and tried re-leaving, joining, + using the link directly c+p in discord), but I don't seem to have access to the pro channels, any idea how I do that?

[tame current]

what are the pro channels

[outer wing]

Some anouncements have channels I don't have access to in it (the page one), I assume those are pro only

[warm cave]

https://discord.com/channels/1047197230748151888/1111765634267742408, https://discord.com/channels/1047197230748151888/1111786888626438245, and https://discord.com/channels/1047197230748151888/1111809274566422588

[tame current]

what's the pro link

[outer wing]

I wonder if the pro link is broken then, I just see No Access

[tame current]

and how would it even know i have pro

oh ig i can clcik it thru pplx

that's what i did the first time tho

[warm cave]

You guys are using this one right?

[outer wing]

Yep that one

[tame current]

should i leave and then jon thru perplexity site again?

[warm cave]

I have seen some people in the past where it only worked like the 5th time they left and rejoined, but i have no idea why that would work, lol

[tame current]

i have the unlocked role what does that mean

oh pro discord i see

[tame current]

can i just get a mod to give me the pro role

[warm cave]

you should get this one

[sleek vortex]

yes

i did that

[tame current]

is it okay if i ping them

and how do i prove i have pro

[sleek vortex]

i said dm for email

[warm cave]

nice, much more clean

[sleek vortex]

but he gave it without that anyway

¯_(ツ)_/¯

[sleek vortex]

then

below

model pickers

[tame current]

which mod did u ping

[sleek vortex]

icelavaman

i think

idk

[tame current]

@halcyon coral i can dm u my email, i have pro and want pro role

[harsh stag]

it's way faster to just leave the server and rejoin via the pro link that was fast

[halcyon coral]

[sullen ether]

Is the usage from the bracket for the whole month or some other ine period?

[sleek vortex]

for the day

when you use one pro, that usage renews in 24 hours

mine says 574 instead of >300 because i have enterprise pro

[tame current]

i think i have enterprise pro too cuz some other dude got it for me for $70 or something

[sullen ether]

So you have ~500 uses for each model per day?

[tame current]

574 what

[sleek vortex]

600 total uses

of any advanced model

pplx pro normal has 300

[tame current]

oh maybe not then cuz i can only use opus 50 times

[sleek vortex]

no i have that too

opus limit

[sullen ether]

That's cool but how's it viable for perplexity?

[sleek vortex]

because pplx would go broke

[sleek vortex]

they charge double

enterprise pro is $40/mo

but yeah idk how its generally viable

[sullen ether]

But even for pro

[sleek vortex]

they must have some form of partnerships with all the companies

[sullen ether]

Yeah

[sleek vortex]

and i think they might sell back the data from all users on free plan to like openai and others

[tame current]

lmao

[tranquil orchid]

@sneakyf1shy warned

Reason: Spam, please avoid sending single word messages
Duration: 7 days

[sullen ether]

What was in the deleted message?

I didn't see

[sleek vortex]

?

it was a shrug???

[tame current]

oh it was cuz spam lmao i thought u were silenced for accusation

[sleek vortex]

can i not use the built in discord slash shrug

bruh what

[austere kestrel]

sign on the left #💬│general message

[tame current]

yeah it didnt listen to me when i told it to make the small sign 50 too

so i just scribbled it on

[harsh stag]

¯_(ツ)_/¯

[sleek vortex]

i guess its because i said like 3 sentences all with like >3 words in a row

oh whatever...

wait please no bot dont do it again

[halcyon coral]

👀

[sleek vortex]

can we get a fix on that?

i sent "but they might" and then a shrug

[halcyon coral]

On what?

[warm cave]

what? alex said the limits were the same

[sleek vortex]

and got an auto warning

for short messages

[austere kestrel]

send longer messages?

[sullen ether]

But opus is insanely highly priced for api by anthropic while it cannot even catch gpt 4o (source: lmsys arena), a model which is priced $5/$10

[sleek vortex]

well fine but still

[halcyon coral]

Please just send longer messages without clicking "send" every few words

[sleek vortex]

they must have some form of partnership

i really doubt they're paying full prices for opus

[warm cave]

I got into the habit of editing my previous message ot add on single things, lol 🤷‍♂️

[sullen ether]

Maybe for openai too and my speculation is gemini got removed as google wouldn't have flexibility in their prices.

And for llama 3 and mixtral, they host it on their own

[warm cave]

But it was Mistral Large, so they may have or may not have gotten local deployment

Large is closed source

[sullen ether]

But isn't 8x22b better than large? Doesn't make sense to me

[warm cave]

I think mistral was just removed bc not many people use it, and they don’t like to confuse users, stupid imo

[warm cave]

Is it really, i never tried 8x22b, i just know we had mistral large on perplexity

I know that 8x22b is a lot newer

[sullen ether]

I heard so, will check lmsys

[austere kestrel]

it's available via labs

[sullen ether]

It seems mixtral large is better, but llama 8b beats them both 😆

[austere kestrel]

haha yep - pretty striking / impressive ins't it

[sullen ether]

It's insane how meta trained their 3 models, a 8b model beats 100-150b parameter model and 70b beating 1 trillion parameter model

[austere kestrel]

the 400b model will be interesting (rubs shoulders with gpt4o on some evals) - but it's almost more interesting how good the 8b version is tbh

[pulsar sail]

What do you think is the best AI model at the moment to use with perplexity?

[sullen ether]

Btw, what do you guys think how perplexity gets information (snippets ig) for their models?

My speculation is they use bing for reranking of webpages, then they use their already indexed website's chunks from their search index and reranks those chunks based on query (most probably BM25) and returns top 5 chunks to the model.
What you guys think?

[sullen ether]

Gpt-4o if you have pro

[pulsar sail]

Thank you. Been using Opus because it gave me the most profound answers, but will try gpt-4o

[warm cave]

Opus if you want to use it for creative purposes

[austere kestrel]

https://spectrum.ieee.org/perplexity-ai

[pulsar sail]

And gtp-4o if you want if for...?

[sullen ether]

That means I was coincidently correct

[warm cave]

Like STEM questions, it will be more accurate, its a very accurate model overall, but can be stale to talk with and is not as creative

[sullen ether]

But how do they then retrieve very recent information, like a few minutes? I have a doubt about this.

[warm cave]

also its pretty fast, so generally purposes can be good

[austere kestrel]

for big news stories, covered by major outlets, it gets updated pretty regularly (at least every hr according to that article), and they also seem to now have things like weather and finance covered (perhaps with their own crawler, but more likely third parties)

but yeah you're right, an obscure blog isn't going to get indexed every hour

there are gaps when the information or source isn't surface level

[sullen ether]

They get. I run my own search engine type thing like perplexity with flask on python, I can see from where I get a request (basically ip address) and I notice every 1-2 hrs, Microsoft datacentres fetch my sitemap, and robots.txt

[austere kestrel]

https://docs.perplexity.ai/docs/perplexitybot

[warm cave]

When do we get that weird but cool voice chat feature with NotebookLM?

[smoky valve]

its down again

[sleek vortex]

well they do it very fast on their datacenter servers using bing api

and ofc since everyone is making so many queries they likely cache a ton of sources

if you ask about some random obscure topic nobodys ever searched then its likely the crawler quickly hits up each page and it takes a bit longer

but it probably hits all 25 sources in parallel so like

[little warren]

Hey, we really need help regarding the API but no response from the support. Been waiting for like 2 weeks now

[halcyon coral]

Hey! What is the problem with the API?

[fading moth]

I personally find 4o pales in comparison to Opus for writing, so this graph seems off to me.

GPT-4T being on par with Opus for some of those writing scores is also hella sus.

4T was lobotomized for creative writing.

[austere kestrel]

i mean all judgements about what's 'good' are is subjective and use case dependent

also, the chart doesn't really show evaluations related to creative writing

[little warren]

We need to have higher rate limits

[halcyon coral]

The response time for such requests is longer, please be patient

[little warren]

We have been approved previously so we're 100% qualified - any way to increase the limits ASAP? We really need the increased limit

[agile jay]

@tidal sail You figured it out?

[tidal sail]

Yes

I just added a keyword

Though it doesn't look as nice as that in chrome where it says search perplexity pro

[agile jay]

Yep, in the past I think they had it, but it's changed for some reason.

You could probably make an extension to get a similar look if you wanted though.

[whole flax]

Hey guys does perplexity have plagiarism because it is taking it from a source?

[vapid onyx]

Perplexity summarizes content and then sources portions that it obtains that content from. It may not be considered plagiarism based off the fact that it’s properly sourced and called out.

[whole flax]

So would an plagiarism checker from my school notice it?

[lone citrus]

It depends a lot, but just simply copying something AI-generated is most likely an abuse of the rules, I'm sure.

[warm cave]

Plagiarism is passing off someone else's work, ideas, or words as one's own without proper acknowledgment or citation. So even if it’s word for word, if you have citation, then it’s not Plagiarism

[lone citrus]

Eidt, Violations to rules

[warm cave]

Yeah, I'm only talking about “Plagiarism”, there is likely rules about ai use

[lone citrus]

Yes, my school is fine with using AI for most projects, but it must be with correct citation, and the main body contents cannot be 100% AI-generated

[tidal sail]

Looks doesn't matter that much. thanks

[feral marlin]

would be detected for ai instead

[true jay]

Is perplexity able to output spreadsheets? I just uploaded one to the platform as part of a chat and got a response, but I don't see where the actual output file is...

[whole flax]

I will just ask a teacher if the plagiarism checkers knows it from ai

[vapid onyx]

Copying AI word for word is not really a best practice and probably would be caught by your school. You should leverage AI to understand the material and leverage it to create your own points. There are sources with your inquiries so it really gives you the option to do the appropriate research

[whole flax]

yea but the problems is here I have to answer 86 questions about an book that is so boring I ended on chapter 10 and there are 36 chapters so

[serene sand]

👆

[warm cave]

If you can, do audio, like an audiobook or just TTS, much better way to do it for boring books. You can even do 1.5 or 2x if you don't get distracted

[whole flax]

Smart idea thx

[golden island]

anyone else getting this? oauth down or something?

[tame current]

Maybe it's an update.

[hybrid chasm]

I really like the UI

[signal hamlet]

Hey @golden island! Are you using a VPN? Have you tried reinstalling the app? Is it working fine on the web for you?

[warm cave]

@signal hamlet Hey, Is frequent app crashing a known bug rn. I'm fully up to date, it just crashes often while the AI is responding

[signal hamlet]

We have a ticket regarding the ios app crashes, the mobile team is working on that.

[warm cave]

Perfect, thanks 🙂

[golden island]

no vpn, works on web (and in browser), have reinstalled and restarted phone. i will try doing it on mobile data real quick haven't thought of that yet

no dice

[serene sand]

iOS version?

[golden island]

yeah

Oh, you want the OS version

hold on

[serene sand]

you can try w/ another restart, also turn off all other apps. if that doesn't resolve #💬│general message

[golden island]

sure

yeah still broken. thanks anyway

i assume it's just google oauth being byzantine

[tame current]

There are so many problems right now

Gpt-4o, when I ask a response to the first question asking it to fix or alter the given answer, it keeps giving me the same answer with no changes

Gpt4-turbo, i wanted to switch back to this one because of better answers with more understanding of previous queries, and even though I selected 4turbo, it keeps showing that it’s using 4o

Claude 3 opus is barely answering any image related questions like if I screenshot a chemistry question, it used to be able to answer that question and give me a detailed explanation of how

Now it keeps saying it might be copyrighted textbook material even though it isn’t, and won’t answer the question so I have to go back to using 4o

4o is a lot less accurate compared to 3 opus

Whenever I chose sonar large as an ai model yesterday and asked a question, it kept showing that I was using gpt-4o

[tame current]

This has been fixed I think because I was able to use Sonar again today

On top of that, after 50 uses of Claude 3 Opus, it switches to claude 3 sonnet, the next day after the last time I used opus, it says I only have 8 uses left even though it should have filled back up to 50

So many issues 🤦‍♂️

Right now, i cant upload any images on to the chrome browser website but can on the mobile app

[tame current]

it’s working now 🙂

[restive osprey]

Im using Perplexity and recommending it more and more

so if they can just keep it simple but also inovate for advanced users, i can see it being the number one internet search for now

at least i use it more than google.

[south kindle]

https://x.com/bobbyallyn/status/1792679435701014908?s=46

Bobby Allyn (@BobbyAllyn) on X

Statement from Scarlett Johansson on the OpenAI situation. Wow:

Twitter · May 20, 2024, 10:10 PM

ooops

[rustic gale]

sharing your iPhone Serial Number in public channel probably not a good idea

[golden island]

fair enough

[minor shard]

If true, this is wildly disgusting. But not suprising. With whats going on at OpenAI I wanted to believe it was just overblown but this proves they are greedy. We haven't even gotten to AGI going public yet and OpenAI is already proving they can't be trusted.

[south kindle]

The fact he even put the Her tweet out was icing on the sh!t cake

[hallow linden]

it's on a rolling basis, so if you use opus 40 times at 8PM, you'll still only have 10 uses left at 8 AM

it's only after 8PM that you get those 40 uses back

[harsh stag]

is the TTS on chatgpt stop working?

or is it just me

[south kindle]

may be related to the above situation. There gonna be under fire until the Sky voice thing is taken care of. i'd be taking my stollen voices down as well in fear of a giant lawsuit.

That being said the PI voices (all 8 of them) are very good. I'll take Pi voice 4 over Sky anyday. and I have a feeling it was created by the book and not nefariously.

[harsh stag]

it works just fine on the android mobile app

[south kindle]

Ah my B. carry on.

[warm cave]

oh, is that why I couldn’t use the voice sky today

[south kindle]

you mean scarlett? lol.

[warm cave]

Lol, I'm just looking, don't know whats going on

[harsh stag]

If I remember correctly, Pi is using 11labs for their TTS, isnt it?

[south kindle]

I only learned about it today.