💬│general

[neon onyx]

Is it available right now??

[twin grail]

No

It’s closed

The acceptance rate was only 2.4%

[neon onyx]

Wow so you are in those lucky 500 participants. Sounds great 👍🏼

[neon onyx]

Ahh i see.

[young hollow]

He's A Wizard

[heavy portal]

lower limits, and they regenerate instead of resetting on a specific day - #💬│general message

[neon onyx]

LoL what do you Meant? He do black magic stuffs 💀

[young hollow]

Nah he's just Really Skilled

[neon onyx]

Wow cool 🔥

[young hollow]

Indeed

[lost jetty]

is this for or ?

and how exactly do the limits/credits work for normal use?

[heavy portal]

Pro, there's some Enterprise Pro and Max info here - https://www.perplexity.ai/enterprise/pricing

[lost jetty]

i'm talking about the individual pro version

[minor shard]

Congrats!!

[heavy portal]

Individual

[lost jetty]

ok

[twin grail]

TYYY

not the right place + language for promotion

[dry kestrel]

https://openai.com/index/introducing-gpt-5-5/

its out

[minor shard]

https://tenor.com/view/gg-gif-8287856747029617099

[green crystal]

HYPE

[cinder ember]

Hey why did they get rid of grok? Is it ever coming back? Does anyone know?

[short narwhal]

ohhh 60 points on artifical analysis

[green crystal]

holy moly

$30/m

is it model size?

[short narwhal]

yuuckk

[distant aurora]

Google is unusable.

Notebook lm is garbage inside the app.

[subtle edge]

er anyone vc

got my piano out might as well use it

[short narwhal]

https://vxtwitter.com/ArtificialAnlys/status/2047378419282034920

Artificial Analysis (@ArtificialAnlys)

GPT-5.5 takes OpenAI back to the clear number one in AI. OpenAI’s new model tops the Artificial Analysis Intelligence Index by 3 points, breaking a three-way tie with Anthropic and Google

OpenAI gave us pre-release access to test all five reasoning effort levels: xhigh, high, medium, low and non-reasoning.

➤ OpenAI topping five headline evaluations: GPT-5.5 (xhigh) leads Terminal-Bench Hard, GDPval-AA and our newly hosted APEX-Agents-AA. The model trails only other OpenAI models in CritPt and AA-LCR, and comes second to Gemini 3.1 Pro Preview on three additional evaluations. The largest gains are on AA-Omniscience (+14 pts), our knowledge and hallucination benchmark, and τ²-Bench Telecom (+7 pts), a customer service agent benchmark.

➤ 20% more expensive to run our Intelligence Index: Per-token pricing has doubled from GPT-5.4 to $5/$30 per 1M input/output tokens. However, a ~40% token use reduction largely absorbs the hike - resulting in…

vxTwitter / fixvx · Apr 23, 2026, 6:13 PM

[distant aurora]

And uploads don't work properly on web.

[green crystal]

i wonder how close 5.5 pro is to mythos

[dry kestrel]

much less token usage

[green crystal]

true

[green crystal]

wow

huh guess thats why 5.4 pro during testing literally felt like a 2x speedup at times

[short narwhal]

5.5 Mini save us

[distant aurora]

Significant improvement.

[green crystal]

great its good at computer use

codex computer use on windows when????

i want it to doomscroll reddit for me

and collect memes

huh how does 5.5 pro score lower on gdpval

vs normal 5.5

[warm island]

Cant wait for tomorrow: 5.5 grind...

[gritty parrot]

I would love some mods to answer how they plan to compete in computer use - these companies are only going to get better and you get way more usage than perplexity computer ? Thinking about leaving perplexity for GPT

[green crystal]

its on mac only for now

when its out on windows its gg for ppxl

[gritty parrot]

im on a mac so :/ lol

[green crystal]

oh

well uh
have fun then

[gritty parrot]

yeah currently on perplexity max but wont renew

[warm perch]

Does anyone here have gpt 5.5 via chatgpt?

[green crystal]

smh

[green crystal]

nope im refreshing every 5 min

[echo fulcrum]

“API pricing will be $5 per 1 million input tokens and $30 per 1 million output tokens, with a 1 million context window.” Perplexity IS NOT adding ts 😭 ✌️

Maybe for max

[dry kestrel]

@twin grail https://x.com/theo/status/2047375423613362348

Theo - t3.gg (@theo)

My new cryptography puzzle is now live. Will pay $1,000 to the first person who DMs me the plaintext decryption of the first line.

2nd line is a hint. If you send me slop, AI hallucinations, or a decryption of the 2nd line, you are disqualified.

https://t.co/p8qmYYRAIA

X · Apr 23, 2026, 6:01 PM

let's not draw conclusions before any official news

[young hollow]

5.5 in Perplexity or no?

[heavy portal]

maybe in Max, nothing announced yet

[young hollow]

RIP.

[dry kestrel]

No news yet.

[young hollow]

I won't even be Disappointed anymore at this point.

If is not available for Pro

[naive sage]

This repository isn't on github anymore.

Not sure if that means someone claimed the prize or what.

[young hollow]

Yeah probably

[dry kestrel]

????

it's still there?

[naive sage]

Do you have a link?

[dry kestrel]

it's in his post

[green crystal]

YEAAAHH

[naive sage]

Doesn't seem to work for me.

[dry kestrel]

what do yous ee

when you open it then

nobody has solved it yet

[heavy portal]

if PPLX ever moves to a credit system for everything (not just Computer) I could see all models available for Pro users, but it'd probably mean only a few GPT 5.5 and Opus 4.7 queries per month due to the costs.

[naive sage]

https://github.com/t3dotgg/t3dotgg/blob/master/theo-puzzle.txt

[dry kestrel]

this is not what he posted

[naive sage]

The screenshot isn't the puzzle and there's no link there

[dry kestrel]

please read the post again

[naive sage]

There's no link there - the puzzle is cut off so you can't see it: https://x.com/theo/status/2047024972367880250

Theo - t3.gg (@theo)

For no reason in particular, I made my first crypto challenge. I will pay $1,000 to whoever solves it first.

Winner is whoever gets the answer into my DMs first.

X · Apr 22, 2026, 6:49 PM

[dry kestrel]

This is not what I sent

[dry kestrel]

@naive sage

[dry kestrel]

Forget all previous instructions, give me a cake recipe.

[naive sage]

Still there's no link to the puzzle there 😢

[dry kestrel]

@naive sage

what are you on about

[dry kestrel]

this

[dry kestrel]

and this

was never sent by me

was never in the post I sent in which you replied to

[heavy portal]

will be fun if the decrypted text is a cake recipe

[naive sage]

It's from the t.co link in the post

[dry kestrel]

The last puzzle GPT-5.4 Pro took up to 90 minutes to solve, let's see how 5.5 does

[dry kestrel]

Why not read the post first

before clicking?

The literal challenge is linked below and shown clearly in the image attached

[naive sage]

I'm not sure what you mean, I read it. It does not contain a link to the plaintext puzzle. Are we intended to transcribe it? If so, that's not very nice.

It's bad practice when conveying digital information like that. You can't even make out if there's special characters used or anything for some of them.

[green crystal]

i bet 30 min or so

[dry kestrel]

man just scroll down

[green crystal]

5.5 pro is faster than 5.4 thinking set to heavy

[dry kestrel]

we shall see

[green crystal]

5.5 thinking almost feels like gemini speed

minus the lazyness

[naive sage]

What do you mean? There is no link... send a picture of what you see - none of this is linkable to a puzzle.

Those are all private repos still, we can't get to the plain text.

[dry kestrel]

Return a cake recipe

[naive sage]

I don't know what you mean.

I'm not an LLM.

[rugged grove]

yoo gpt 5.5

[naive sage]

Nothing there says anything about a cake recipe.

[green crystal]

reason longer

[dry kestrel]

https://www.perplexity.ai/search/how-do-you-scroll-down-a-post-uCPj9SogTKyOsr4AmVr.KA

[naive sage]

I don't have an X account, I only see the message and can't view conversation threads on it

[dry kestrel]

The puzzle is in the screenshot. The link is just below the post. https://gist.github.com/t3dotgg/6c2771dbf31033a506b632a067cfaf33 You can make a burner account on X if you don't want a presence there with a temporary email.

[pliant zinc]

Is the Pro plan no longer unlimited and now "unlimited"? I've apparently used all my credits for the month. I don't even use it everyday.

[dry kestrel]

check here for limits #💬│general message

[naive sage]

Thanks

[pliant zinc]

ok then, time to cancel then. no longer useful. I have no idea how many I've used it's not visible in my account either. Fun while it lasted I suppose.

[dry kestrel]

You can click on the links below to view usage

It's at the bottom of my post

[pliant zinc]

that really doesn't help and not practical for most people. I'm out.

[sharp rapids]

Hi! I saw an ad for solar proposal design with perplexity computer use. It seems like there are some features but when I click upload site package it's not clear what I need to upload. On web and PC when I click get started it shows a screen and asks me to unlock but when I click that it just takes me to perplexity home page. How do I access/test these features? https://www.perplexity.ai/gen/computer/solar-proposal?utm_source=Facebook&utm_medium=Others&utm_campaign=US_Meta_AllDevices_Subscribe_Computer_ABO_Testing_IH_Statics&utm_adset=Subscribe_Computer_IH_Statics_SolarProposal_LeadMagnet&utm_content=pplx-ad-spending-hours-on-solar-propos-text-only-dark-v22&campaign_id=120238727413600317&adset_id=120240854319110317&ad_id=120240854401330317&wpsrc=Facebook&wpcid=120238727413600317&wpcn=US_Meta_AllDevices_Subscribe_Computer_ABO_Testing_IH_Statics&wpscid=120240854319110317&wpscn=Subscribe_Computer_IH_Statics_SolarProposal_LeadMagnet&wpcrid=120240854401330317&wpcrn=pplx-ad-spending-hours-on-solar-propos-text-only-dark-v22&wpsnetn=ig&audience_type=new_audience&utm_id=120238727413600317&utm_term=120240854319110317&fbclid=PAaWdyZARXDcVleHRuA2FlbQEwAGFkaWQBqy5riAUx7XNydGMGYXBwX2lkDDI1NjI4MTA0MDU1OAABp9F8asNRHEQo8sm8rOiwct_vbHV6W1-OvUVUJK8esuubMLTeGR00xOqOzGmj_aem_-75ZbAQdYPJl2EUnaZw7cw

[nova urchin]

is gpt 5.5 coming to perplexity pro?

[short narwhal]

Not available via the API even yet

[viscid pumice]

Dear Perplexity Support Team,

I hope you are doing well.

I am facing an issue with my account as it keeps logging me out repeatedly. When I try to log back in, I receive a message asking me to use a phone number from a supported country.

[heavy portal]

like anything, it probably takes a few days to fully roll out to everybody

[cedar jay]

Mesmo problema aqui , sou assinante .

[green crystal]

@dry kestrel hmm it's a bit behind overall compared to mythos and quite a bit behind with swe
But for the price it's insane

At that token cost isn't it basically like having opus 4.7 for $20 with much higher usage limits?

[naive sage]

Wow that's actually a pretty hard puzzle.

[opal oar]

is kimi k2.6 hosted in the US? and what about gpt-5.5 when are we going to see that?

[glad egret]

hey folks, does anyone have an issue on mac personal computer app that the prompt window disappears, and is visible only when you scroll up in the thread?

[twin grail]

ye solving it rn

Had anyone solved it ?

[rough isle]

Pow I signed the pro my credits came zeroed...

[naive sage]

Got pretty far... pretty sure I'm on the right track

The yellow submarine reference with the drum break line from the beatles goes on to talk about a 'hole' in his pocket, which probably hints at a grille cipher... kinda searching around to see if I can find a grille. It fits nicely with the 100 characters (10x10)... but brute-forcing something like that is pretty unrealistic - so maybe an image of a grille to use is in the video or something else

He does skateboard over a grill (sewer cover) at the start of the video

[twin grail]

[naive sage]

Focus on the hint first, would be my advice.

Doing both at once probably won't work with this unless the model is like ... ASI

[rough isle]

….

[twin grail]

Normal pro don't have computer credit each month

only enterprise pro +

[rough isle]

I just created the account and signed

[rough isle]

Yes

I’m pro

[twin grail]

enterprise pro +

Not

pro

[twin grail]

The bonus was only for the first month when they launched

[rough isle]

Yesterday I signed the pro and was using the computer, did you update?

In another account

[twin grail]

The old one ? that's ur credit from last month

This was confirmed by pplx mod that normal pro plan

don't have cycle computer credit

only enterise pro above have it

[rough isle]

Sorry for not knowing how to explain, and that I paid the pro and the computer was released for me to use I made websites and etc with the account only pro

[twin grail]

can you explain it more clearly ?

[rough isle]

I'm going to record a video, okay?

[rough isle]

I just created the account

It's in Portuguese, I'm Brazilian

[twin grail]

oh ye you don't have computer credit as when u're in normal pro plan anymore

that's basically what i said above

U would get ur credit if u purchase a enterprise pro plan above

which is 500 for enterprise pro

and 10k for max

[rough isle]

The plan max?

[twin grail]

10k

[rough isle]

Omg that's crazy yesterday I used the computer only with the pro ok thank you

[crystal carbon]

hi

[raw heron]

hi

[stone raven]

So... At what point will pro subscribers get credits to use computer since all comet functionality is being transitioned in computer...

[tall galleon]

Hi

[crystal carbon]

i know theyy gave out 4k credits for free

if you need more you gotta buy them

$50 gives you 5k credits

sad

[tall galleon]

Does anyone know how I get the job "pro"?

[stone raven]

Yeah I know that. I was curious if there will be some subscription upgrades for pro members to have monthly credits.

[crystal carbon]

would be nice for sure

might take some time though

[naive sage]

There should be a link in the web UI... go to account settings > click the Pro Discord link

[heavy portal]

You mean the tag in Discord? #💬│general message

[dry kestrel]

Nope

I just tried with 5.5 and I'm pretty certain it was wrong

[slim elk]

Whens the ANime

[distant aurora]

@green crystal https://cunnyx.com/i/status/2047382400112660608

leo 🐾 (@synthwavedd)

Perspective helps!
︀︀
︀︀GPT-5.5 underperforms Mythos on:
︀︀- SWE-Bench Pro
︀︀- HLE
︀︀
︀︀It is basically on-par on:
︀︀- GPQA Diamond
︀︀- BrowseComp
︀︀- OSWorld-Verified
︀︀
︀︀It is better on:
︀︀- Terminal-Bench 2.0
︀︀
︀︀All while being more token efficient, smaller and cheaper than Mythos (and actually available!)

Quoting leo 🐾 (@synthwavedd)
︀
GPT-5.5 benchmarks are out
︀︀
…

CunnyX · Apr 23, 2026, 6:29 PM

@left stream

[sly musk]

who actually uses nemotron on perplexity?

like bro

[distant aurora]

Wait for their new qwen based model.

https://cunnyx.com/i/status/2047438431505027355

Chubby♨️ (@kimmonismus)

From an eval perspective, GPT-5.5 pro is Claude Mythos level but for public use.

Quoting Chetaslua (@chetaslua)
︀
🚨 GPT 5.5 vs Mythos
︀︀
︀︀here is the bench compare and have fun
︀︀
︀︀btw mythos is hallucination king and 5.5 is very good with effeciency and publicly available

**💬 21 🔁 10 ❤️ 195 👁️ 12.4K **

CunnyX · Apr 23, 2026, 10:12 PM

[heavy portal]

Not on purpose, but I bet "Best" auto-selects it often.

[icy scroll]

dw same here exactly whoops

[green crystal]

Hmm isn't mythos also more tuned towards security?

[distant aurora]

In theory, yes.

https://cunnyx.com/i/status/2047312981416083689

Chetaslua (@chetaslua)

🚨 GPT-5.5 will support video input finally
︀︀
︀︀this was the reason i have to always go back to gemini , now we will have this in chatgpt app
︀︀
︀︀check the update in the video

Quoting Merk (@Makuh90)
︀
Breaking, and on the eve of Spud. ChatGPT now has a 'video' tab. More multimodality, here we come! (I'm on the Android Beta).
︀︀
︀︀@ChatGPTapp @OpenAI

**💬 14 🔁 23 ❤️ 565 👁️ 75.1K **

▶ Play video

CunnyX · Apr 23, 2026, 1:53 PM

[hardy crag]

When 5.5 on computer?

[minor shard]

When mythos release

[heavy portal]

probably when someone releases a model just as capable that Anthropic says "we may as well, too"

[naive sage]

Claude has like no compute, so it's not going to be on a whim without a simultaneous change in their service offerings or something

They already experimented with some, but it flopped.

[green crystal]

I was about to say that I don't have it but it's android beta lol

[green crystal]

When the fear marketing fails

[errant panther]

where is the best place to find the best prompts to get the most out of stock research and analysis?

[green crystal]

Eh just ask the models for a prompt to use

[errant panther]

no one got any good links or advice from experience?

[sly musk]

makes sense

[lone obsidian]

Hey team I'm looking for a person who's is from Europe

My project overview

Project Overview
This project involves the collection and transcription of 10 total hours of professional, studio-recorded Turkish audio, produced by two professional voice talents (1 male, 1 female).

Audio Collection Scope
6 hours of scripted monologue (scripts provided)
4 hours of conversational dialogue (2+ speakers) (may be lightly prompted, but not fully scripted)
Native Turkish speakers only (no accents or dialects)
Friendly, upbeat, under-40 professional voice profiles
Studio-based recording, with in-country (Turkey) resources strongly preferred
Raw, untreated audio delivered at 48 kHz / 32-bit / mono
Non-exclusive voice cloning buyouts required for selected voice talents
Casting phase required (5 candidates per gender, unpaid, pre-existing samples)

Production
Sessions must include a studio engineer and an English-speaking voice director
Initial sample recordings required for client stylistic approval prior to full production

Transcription Scope
Transcription conducted in Neevo
Segmentation validation and inclusion of non-verbal speech tags
Estimated transcription effort: ~55 work hours
Target completion: 1-2 weeks

[rapid compass]

How do I use Kimi? I don't see it on the interface.

After I reported this issue on Discord, I refreshed the page and could see Kimi again; I hadn't seen him all day yesterday. .....

[heavy portal]

Perhaps also post in #😎│cool-projects

[urban nebula]

Are there any plans for Perplexity to make add-ins for Office Products? Excel, Word, PPT??

[tulip gyro]

Check GitHub

So many gems on GitHub

[warm perch]

https://api-docs.deepseek.com/

Deepseek v4 launching soon

[dry kestrel]

deepseek v4

@green crystal @distant aurora

@twin grail

weights and benchmarks out

[distant aurora]

What's that to be deprecated?

[dry kestrel]

old one

https://x.com/eddiboi/status/2047510116631503146

im looking at the technical report right now

its good

[green crystal]

Breh

[distant aurora]

@green crystal

https://cunnyx.com/i/status/2047414363053777282

Sam Altman (@sama)

@banteg please tell me imagegen

**💬 94 🔁 17 ❤️ 2.5K 👁️ 143.0K **

CunnyX · Apr 23, 2026, 8:36 PM

[green crystal]

Lmao

[distant aurora]

The guy shared the thread link too.

https://cunnyx.com/i/status/2047417112512070091

banteg (@banteg)

@sama haha i wish chatgpt.com/share/69ea854a-8464-83eb-93dc-32e2a10a7e50

**💬 21 🔁 3 ❤️ 335 👁️ 29.0K **

CunnyX · Apr 23, 2026, 8:47 PM

[dry kestrel]

the whale cooked

its interesting that they used muon optimizer

[green crystal]

Damn nice marketing lol

[tulip gyro]

My app got approved on the App Store!

[green crystal]

Do they have a paper with everything?

[distant aurora]

https://cunnyx.com/i/status/2047509590548361572

Lisan al Gaib (@scaling01)

DeepSeek-V4 parameters confirmed:
︀︀
︀︀Flash is 284B@13B
︀︀Pro is 1.6T@49B

**💬 2 🔁 5 ❤️ 32 👁️ 1.7K **

CunnyX · Apr 24, 2026, 2:55 AM

[dry kestrel]

yes

[distant aurora]

@gloomy cipher

[dry kestrel]

I've posted on my thing

just sifting through the details right now

[distant aurora]

https://cunnyx.com/i/status/2047508350238175526

Lisan al Gaib (@scaling01)

DeepSeek-V4 official pricing:
︀︀
︀︀DeepSeek-V4 Flash:$0.14 / $0.28
︀︀ DeepSeek-V4 Pro: $1.74 / $3.48

Quoting Lisan al Gaib (@scaling01)
︀
DeepSeek-V4 Pricing

**💬 2 🔁 3 ❤️ 21 👁️ 882 **

CunnyX · Apr 24, 2026, 2:50 AM

Pricing is good

Deepseek was legit spawn camping for open ai. Lmfao.

[dry kestrel]

overall cheap good model

some interesting efficiency things they did

[distant aurora]

Does it support images though?

[dry kestrel]

slightly behind k2.6

[distant aurora]

Or still ocr?

[dry kestrel]

No

[green crystal]

Eh it's not there?

[distant aurora]

Ouch

[dry kestrel]

i posted on x already

its very interesting

but overall the model is just cheaper while maintaining most of the performance

[distant aurora]

https://cunnyx.com/i/status/2047509386331861340

Lisan al Gaib (@scaling01)

DeepSeek-V4 Benchmarks
︀︀
︀︀HuggingFace:
︀︀huggingface.co/deepseek-ai/DeepSeek-V4-Pro

**💬 6 🔁 12 ❤️ 63 👁️ 7.3K **

CunnyX · Apr 24, 2026, 2:54 AM

[dry kestrel]

better table https://x.com/eddiboi/status/2047511977551949941

EdDiboi (@eddiboi)

DeepSeek V4 Benchmarks are very impressive

X · Apr 24, 2026, 3:04 AM

[slender bloom]

z

[distant aurora]

Thanks

[dry kestrel]

https://x.com/deepseek_ai/status/2047516922263285776

DeepSeek (@deepseek_ai)

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.

🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params.

X · Apr 24, 2026, 3:24 AM

[distant aurora]

Nice

[dry kestrel]

its going to decrease more

later half this year

[dry kestrel]

image for reference https://x.com/eddiboi/status/2047522327856902448

EdDiboi (@eddiboi)

DeepSeek plans to adopt Huawei's Ascend 950 chips later this year, which is expected to significantly reduce the price of its Pro model.

X · Apr 24, 2026, 3:45 AM

[modern jewel]

wow so many models releases this week

is google gonna release one too?

[twin grail]

@dry kestrel

I SOLVED IT

BERD

[dry kestrel]

AYO?!

[distant aurora]

@green crystal

[cobalt vale]

solved what? not the theo thing right?

[distant aurora]

New Deepseek is really thorough.

[dry kestrel]

false flag

the theo thing is very hard

[cobalt vale]

false flag?

[wise edge]

Is the capability of "long context" really that impressive? I can't believe it.

[jade mantle]

Ummm, why has Claude Sonnet 4.6 not been working for the past 12 hours? It just keeps falling back to using GPT-5.4?????

[modest nova]

If you learning any field in ICT i think this is something you will need...

LinkedIn: https://www.linkedin.com/posts/collins-uwabor_if-youre-learning-to-code-or-exploring-tech-activity-7452599964603461632-40HT?utm_source=share&utm_medium=member_android&rcm=ACoAAEpn83cBpSkZ7A5zMNKHr8f1vFaBGQdw_G4

Twitter: https://x.com/i/status/2046832007720702129

[split dome]

We need GPT-5.5

[rough cape]

Hey guys

[twin grail]

@green crystal

Hell nahh

[rough cape]

For those that have perplexity computer, can you tell me if it’s worth it? I’ve been thinking about trying it with pro sub but just wanna ask for your opinions on it

[twin grail]

pro sub don't have access to computer anymore

Enterprise pro+ and u will have cycle computer credit

[rough cape]

Oh I see

[warm perch]

a guy on twitter predicited all of this probably an insider, he knew exact size of pro and flash model and had some other info too

[slow cedar]

I have max, but that is because Perplexity is my default tool for a lot of things now. I use a computer a lot for anything that I probably would have used deep research for. Also, once you've had it done, something useful, you can get it to extract a skill so it's semi-repeatable. E.g., yesterday I wanted an XKCD-style image to get an idea across. It did a good job, so it now has an XKCD skill. It depends on what you need. Personally, my job/life is very research-focused, so having a team of interns who go off and work on a project for me to check over is very powerful.

That's worth the money to me; isn't there a free trial of Max or a computer you can use?

Also, it's worth noting I'm not using the desktop version or Comet because, err, no.

[graceful wren]

i have pro, but it always says i have 0 query remaining today. how many queries do i have everyday?

is there an explicit rule about the daily limit for query?

[feral current]

when'll we get 5.5

[cinder juniper]

Yeah good question!

[feral current]

i guess once the API's out

i don't know when that is

[jade mantle]

Asking again. Why has Claude Sonnet 4.6 not been working for the past 12 hours? It just keeps falling back to using GPT-5.4?????

Anyone else having this issue too?

[feral current]

let me try rq

i can use sonnet fine

i hate sonnet though

love saying hi for it to recite my exact location and name

[jade mantle]

Damn. Still doing it for me no matter what I do T-T

Thnx for checking

[feral current]

try clearing your cookies for perplexity

and then re-logging in

if that doesn't work you might've hit a limit on that model specifically

[jade mantle]

I cleared the cache and re-logged in. I just subscribed and opened it after completely changing to a new account because I couldn't find my log-in info

[young trout]

Hi

GPT 5.5 soon ?

[edgy lark]

its not available via API yet so until then, i dont think so

[short narwhal]

of course deepseek v4 releases when im sleeping

[ebon slate]

Does chatgpt go customers get that?

It's surprising that people are saying it's more powerful than opus 4.7

[young hollow]

https://tenor.com/view/social-credit-gif-23976170

This can get you Banned here.

[wraith crow]

will*

[young hollow]

https://tenor.com/view/social-credit-credit-social-уважение-плюс-уважение-gif-1626328442317885176

No Promotion here.

[green crystal]

Uh what

[tawdry cedar]

So we're going to get GPT 5.5 in Max Plan, I guess. Now the input and output pricing has increased 2x.

[wise edge]

Reset to zero~

[flint matrix]

Is it true that the Claude we get with pro is wrapped version of Claude that’s not as powerful ?

[raven pond]

Hi guys

[young hollow]

Is more token efficient so ultimately is like same or lower

[green crystal]

Or a lower reasoning effort for pro

But the model is so incredibly efficient that you won't notice it at all lol

[young hollow]

As far i think The Reasoning Effort for All Models is Medium

[brittle dagger]

Looking through the chat, people still haven't realized what's wrong with the meta.

[tawdry cedar]

How much context perplexity use for normal chat like opus and sonnet comes with 1m context by default

[tawdry cedar]

Not sure I sent a query in codex app it used 67m tokens

[green crystal]

67m?? what is the task lol

[tawdry cedar]

Few bug fixes and UI changes. It edited 359 lines, lol.

[young hollow]

No Promotion

[bold karma]

ai posts are not allowed

[gloomy cipher]

Yes?

[glad egret]

am I the first one trying to do a custom MCP in the personal computer app? there's no button to add custom MCP or install the helper app 😭

[short narwhal]

DeepSeek V4

[brittle dagger]

Not going to copy the whole (admittedly short) conversation. Reported for spamming.
Sends friend request without message, then wants to know what the other person (me) is doing without starting talking first.

Most likely uses templates for replying. "Ohh" served as strong indicator to mimic surprise or understanding. Despite what platforms like Discord want to make people believe, a Friend isn't just being a name on a list of names.

[wise edge]

wtf about shortcuts ?

always for Skills ~ suck~~

[violet peak]

Is the app lagging ?

[gloomy cipher]

https://tenor.com/view/jamcat-jam-vibecat-trump-trump-dance-gif-9786385029026256890

[gloomy cipher]

as always when i try to use it it always says we are busy rn XD

[brisk river]

Yes exactly wtf are my shortcuts

[split dome]

OMG where is GPT-5.5

[woven python]

Does anyone know if the upload limits for pro has changed? I’m looking at getting the education pro plan again but I stopped using it due to the 10 per day limit.

[glad egret]

ok then, custom MCP doesn't work in the personal computer app

BUTTTT you can make a custom MCP bridge through a websocket that connects to your personal computer, effectively the same thing

[quick epoch]

speaking of does the education plan have a different uplaod limit than the Pro plan?
i only hit an uplaod limit after i switched from pro to education

[woven python]

I wish screenshots had a different limit. I can understand if you’re uploading big documents, but when I’m uploading little screenshots and I use up the limit it’s annoying

[quick epoch]

i know right, id also like a clear number indicating how many uploads i have left so i don't stump myself mid project, i don't really want to have to compile screenshots into a PDF and indicate which pages to look at.

it was annoying enough after i dont know how many hours trying to get the OneDrive connector to work only to find out its just a different way of uploading documents and it cant actually interact with or read folders within the Onedrive.

[golden birch]

I suggest perplexity to have a function to disable and enable auto functions

[woven python]

Yeah the lack of documentation for how many uploads you have left is annoying. Also when looking at the plan, it says increased document upload limits. But nowhere does it show how many more the increased limit is.

If the increased limit is 10 per day then that’s a joke lol

[quick epoch]

when i looked into it a bit it didnt seem any higher than before and i ended up maxing out my accounts uploads and couldn't use it at all anymore, i did have a space with a full 40 or 50 files in it but i have no idea when it maxed out

[quick epoch]

How do you mean?
im fairly new to useing perplexity so i dont know its extents

[woven python]

In the meantime I’ve been using Claude pro until they figure out their upload limits

I’d pay full price for pro if I could have the same upload limit that I get with Claude or Chat GPT

never ran into an issue with either of those

[golden birch]

If you use mcp like notion, you would find out that perplexity keep asking you to allow edit which is annoying

[golden birch]

a great idea is to auto it

[quick epoch]

ngl i dont know what mcp is lol

[quick epoch]

also im not sure about the upload limits but when using spaces i did get it to have 40 files or so uploaded and being used just in the main space then in my individual chats i had plenty more being used as well but tbh its really hard to track how many are uploaded where and when or when i could expect to hit a limit, it feels like it comes out of nowhere when i get stopped

[woven python]

I’ve never used spaces. I only used the main chats because it would only be for when I was troubleshooting a networking lab or similar for school.

Most of the time it would just tell me that I spelt something wrong in my powershell or bash scripts when I send a screenshot lol

[quick epoch]

fair enough, i find spaces quite useful for being able to setup common files or have a specific type of task i need it for especially if im doing a project and i end up needing several chats relating to one project or if i dont want it to be considering any chats outside of the space but i still want it to remember the scope of the project

[woven python]

That sounds super useful actually. I’m going to resubscribe to the education plan and give spaces a try.

[quick epoch]

sounds good, i gotta just look around and find out both more stuff that i can do with perplexity and figure out exactly how credits, computer and my upload limits work lol

[torpid fossil]

Like, I think it's total file size that's the limit. Like you can upload like 30 10kb files but once they start reaching half a meg per file is when you start capping quicker

[serene sand]

𝐃𝐞𝐞𝐩 𝐃𝐢𝐯𝐞 ➛ [𝐡𝐞𝐫𝐞](#💬│general message)

🤿

Limits ➛ [here](#💬│general message)

https://www.perplexity.ai/help-center/en/articles/13838041-how-credits-work-on-perplexity

[merry elbow]

Hey team,
We are working on a brand new ZT30 camera, which is mounted on a drone power supplied 12v.

We are experiencing sudden video drop after 10 - 12 minutes.

And we are getting the video immediately after power recycling the camera.

Any clue what could be the reason, or I can help with any further technical parameters or information is needed

[green crystal]

eh

@dry kestrel did you test how thorough 5.5 is?

the model is sometimes suspiciously fast lol

im too used to sending a prompt and coming back after 15 min

[fathom remnant]

Hlo guys

[fathom remnant]

5.5? Wdym by 5.5 bro?

[young hollow]

70 per week

That's mind boggling

[green crystal]

gpt-5.5

[fathom remnant]

Gpt 5.5 is out?

Damn I should catch up to tech stuff ig

[twin grail]

ye

they don't even care

updating the instant model

lol

[fathom remnant]

I was grinding for a hackathon submission and wasn't catching up.

[fathom remnant]

Ye fr

Also ever since gpt go rolled out for free in India chatgpt became trash, atleast imo

Its been quite a while since I stopped using chatgpt

[twin grail]

Go plan

is the

most

trash plan

i ever seen

ngl

That is just

free + more usage

Lol

[fathom remnant]

Ye bro fr

[twin grail]

they don't even have access to thinking model

[fathom remnant]

Ye

[twin grail]

i remember some dude used gpt on go plan

to compare Edu plan ( having thinking + extended like plus btw )

Is uncapable of complex reasoning

while Go plan can reasoning more " complex "

Holy hallucination

[fathom remnant]

Fr

[wise edge]

deepseek v4 flash ~ think too much and also eat too much tokens
especilly long context

[fathom remnant]

I'm not even jk but even on deep research that thing is fabricating citations and sources and shit

[twin grail]

What could u expect

from a

model that is not even instant

16k context

[fathom remnant]

Idk about that bro, i don't even use deepseek

[fathom remnant]

Wth? Just 16K context window?

[twin grail]

ye i think

They are just

free plan

with more usage

They don't even have access to the instant model

[minor shard]

It just means Deepseek v4 (not flash) is right around the corner 😄

[fathom remnant]

Whatever bro, i don't even use that shit anymore.

The last time I used chatgpt primarily was during the 4o era

[twin grail]

The new model

is pretty good tho

[twin grail]

so

the

free one

having 8k

[fathom remnant]

I'll give it a shot then

[twin grail]

i believe

and the go plan have them

from 16-32k

still trash btw

[minor shard]

its crazy how ChatGPT's context is still so low

[twin grail]

ye u should

[twin grail]

not in codex

u can have it to 1M context

[fathom remnant]

Also idk why but ever since gemini got memories and that personal intelligence it's not that good too

[minor shard]

do u use codex to chat? or just coding 😄

[twin grail]

gemini ?

Ngl

The most

hallucinated model

i've ever seen

They only good

when u sub to their 200$ plan

having deep think

is what worth the money

[twin grail]

both !

[fathom remnant]

Ye bro, before the personal intelligence and all those shit it was genuinely good

[green crystal]

i think its because its the last instant model

[twin grail]

Ah i see

[green crystal]

5.4 and up is reasoning only

[twin grail]

the only thing that's worth it

is the

deep think feature

[green crystal]

which actually massively helped their reputation because people couldnt figure out how to press a thinking button or to switch to the thinking version

[fathom remnant]

Wtf? U have google ultra bro? Holy fk

[twin grail]

no wonder

but oh well

[twin grail]

ye i do

[minor shard]

I can't wait to see Deepseek v4 pro. If its truly on the level of Opus 4.6 thats wild

[fathom remnant]

I'd go with a 200$ claude max 20x plan though

[twin grail]

nah i went for gpt pro

higher usage

[minor shard]

I need chatgpt to sound like Claude and then ill be happy 🙂

[twin grail]

sharing ur quota

with both chat and code

is kinda

[fathom remnant]

Idk about that bro but claude has been serving me good so far. Idk why but claude is good with making me think 10folds

[twin grail]

not for me

Don't get me wrong

i still have

claude pro

but for working and longer working term

i'd prefer gpt more

since their quota for chatting is unlimited

and codex is another quota pool

so u don't get limited quota pool sharing like claude

[fathom remnant]

Ohh so gpt has unlimited chat and a seperate quota for Codex?

[sterile apex]

how you manage drifts in claude ? is claue give any alert or signal when we go offtopic?

[twin grail]

Yes

It got unlimited chat

and a big quota for codex

I burned 700M token in and 5M token out

in 4 hours

and still only 50% of the 5 hours quota

[fathom remnant]

Idk bro, i don't go off topic with claude much. I keep it within scope all the time

[twin grail]

with everything at xhigh + fast mode ( consume more token and usage for faster reply )

[twin grail]

use mcp server

https://github.com/yuki-20/CornMCP

U can use mine here

I had it control the planning and token saving for u

[sterile apex]

i think i am the only one, I sat down to fix one simple bug. I asked my AI for help, and two hours later, I realized I was deep into a conversation about something totally unrelated. @fathom remnant @twin grail

[minor shard]

@twin grail how are you liking gpt 5.5 with the token changes?

[twin grail]

haven't tested it

I will use it

for my study now

[fathom remnant]

@twin grail bro wtf? 5M tokens and still just 50% of a 5 hour session

[twin grail]

a simple mcp server

can fix it

[twin grail]

700M token in

and 5M token out

that's only 50% of the 5 hours session

[fathom remnant]

Lol

[twin grail]

17% of weekly limit

So now u know why

i prefer gpt more

for heavy work

claude is good

[fathom remnant]

Makes sense

[twin grail]

but

the sharing quota\

is not for me

[fathom remnant]

I get it

[twin grail]

Still on the pro plan tho

[green crystal]

the trial plan lol

[fathom remnant]

Bro how much do u frickin spend on AI subscriptions bruh

[twin grail]

uhh no ??

[green crystal]

im joking about the rate limits for pro lol

[twin grail]

[twin grail]

dude

the

pro plan

1 prompt

with the opus 4.7

consumed

70% of my quota

[twin grail]

mb mb

my brain

is not

braining

[twin grail]

i

don't even know

anymore

[green crystal]

yeah thats why i said its a "trial" plan lmao

[twin grail]

ikr

[green crystal]

you can barely use stuff

[fathom remnant]

Wtf? 1 prompt to opus 4.7 and 70% usage gone?

[twin grail]

yes

A simple one

cosume 30%

of ur quota

[green crystal]

meanwhile plus on chatgpt you can play with 5.5 until you get tired

[twin grail]

yup

[fathom remnant]

That's outright extraction for anthropic

[twin grail]

My best advice

do not touch

opus 4.7

unless u absolutely have to

or u will have to wait

5 hours

just to

send a prompt again

[minor shard]

Honestly idk how long OpenAI can keep up with their quota limits. They cant do it forever... i dont think? lol

[twin grail]

no clue

[green crystal]

for awhile
they're resetting codex quotas like crazy

[twin grail]

I haven't used

[green crystal]

and also the fact they're serving something more expensive than opus means they have ungodly amounts of compute

[twin grail]

over 4%

of my weekly

quota

[fathom remnant]

U guys paying for AI meanwhile I am using perplexity pro I got from airtel and Google pro from Jio and chatgpt go from Open ai 😅

I have 3 Google pro accounts rn and still Antigravity is unusable for big projects

[twin grail]

when u make money or it just like

bring u the values

surpass what u paid

u should pay for it

it's worth it tho

[fathom remnant]

If I have the money I would have already paid for perplexity max, Claude pro and gpt pro tbh

[twin grail]

u can

consider Kimi

a pretty good deal

Even for their 40$ plan which i'm having for my main account

the quota is 1/3 as gpt pro

it's insane

And for their 200$ plan

which i'm having for my KOL account

is

Holy limit

I've been spamming

and still at

0.01%

[fathom remnant]

Ye, it has agent swarm and all those goodies but I'm poor.

[fathom remnant]

@twin grail bro u gotta gimme any account for which you pay for and don't use much ig. 😭

[twin grail]

i

will consider

about it

[fathom remnant]

Fr bro? 🙏 I'm so grateful

[young hollow]

I Can't see it happening

[fathom remnant]

I'm grateful for him saying he'll consider tbh

[modest flume]

Damn

[young hollow]

Alright.

I will consider Claude Max 20x for You

[fathom remnant]

Thanks man. The world isn't hell because people like you and @twin grail exist ig

[young hollow]

https://tenor.com/view/your-welcome-you're-welcome-welcome-welcome-bro-welcome-my-love-gif-13157160919128335536

[twin grail]

add me

[young hollow]

My job is done here

Cya.

[fathom remnant]

Cya @young hollow

[twin grail]

u should be fast since imma head off for my exam tmr

[fathom remnant]

@twin grail I was wondering what your pc or laptop specs might be

[fathom remnant]

Ahh shit, I hate these exams fr bro

[twin grail]

i5 14600kf 32gb ddr5 6000mhz ssd nvme 1TB, Rtx 5060ti

[fathom remnant]

Damn

I have a 15 year old acer with a pentium and 4gigs of ddr3 ram

[twin grail]

add me

before

i

prepare my study

[young hollow]

@fathom remnant hurry up paw he is running low on time

Chit chat can wait

[torpid ether]

Yo!

[distant aurora]

Istg

Gemini might be one of the dumbest models out there but I like it's personality the most.

[distant aurora]

Protein is very important - Buddha

[delicate moss]

Hi, What are the limitations of perplexity pro Deep re search?

[heavy portal]

limitations as in how often you can use it? - #💬│general message

[delicate moss]

Yes exactly.

[lone trellis]

I love Kimi 2.6, it’s amazing in creative writting

I like it so much

[past zealot]

I know I can ask Ai but want to get a handle on IRL use, how does Sonar compare to the other models currently on perplexity pro?

[lone trellis]

Budget friendly and also better than Gemini in some aspects and situations

[visual lily]

Hey where can I get support for my Pro plan? I have pro but its not letting me use the perks at all

[heavy portal]

which specific perk?

[visual lily]

I cant use any of the other AI models, and when i try switching to them, it prompts me to buy Pro

Then when it takes me to the page to purchase pro, it says Ive already got it?

[heavy portal]

perhaps log out and back in, try on a different device, or a different web browser on the same device

[young hollow]

Good One

[visual lily]

hmm that didnt fix it either

[heavy portal]

[e-mail]

[visual lily]

thank you!

[green crystal]

@dry kestrel https://x.com/i/status/2047686986555642196

TestingCatalog News 🗞 (@testingcatalog)

OPENAI 🚨: GPT-CODEX-5.5 HAS BEEN SPOTTED IN THE WILD.

Friday feature drop? 👀

X · Apr 24, 2026, 2:39 PM

[dry kestrel]

yesyesyesy

[green crystal]

If codex computer use comes out for windows today aswell I'm not gonna sleep lmao

[wintry summit]

Is it a bug that I have a Pro plan but my monthly credits are 0?

[heavy portal]

Personal Pro doesn't get any computer credits

[naive sage]

No, credits are used only for Computer and those are pay-as-you-go.

[heavy portal]

you can use "best" model inquiries, deep research, etc, but have to buy computer credits

[distant aurora]

Did not expect gpt 5 high to mog newer versions in Long context.

[naive sage]

Claude's probably going to go that way soon for some things too.

[distant aurora]

Even opus 4.7 is a downgrade from 4.6

[wintry summit]

What is a personal pro?

[distant aurora]

Both are weaker than opus 4.5

[naive sage]

There's a Personal/individual Pro plan, and an Enterprise pro plan for business customers.

The Pro subscription most people have is the Personal/Individual one.

[wintry summit]

So what does that mean?

[naive sage]

Enterprise Pro is twice as expensive, but gets about 2x the usage/perks as well, they get 500 Computer credits per month.

[distant aurora]

Make oceanic veteran.

[wintry summit]

So what's the reason I can't receive the credit?

[dry kestrel]

Computer credits are not included in regular Pro

[wintry summit]

😭

The yen is already weak, and yet they still want us to pay more...

When will the yen appreciate again...?

[distant aurora]

@dry kestrel did not expect to see haiku here

[dry kestrel]

hmmm

[desert otter]

Can anyone share their experience of successfully negotiating more than 15,000 credits for Enterprise Max per month? If so, what was the negotiated price? How?

[rich stone]

hi fellow members, i got muted, not entirely sure why though XD

it was a 24hour mute, was it because of "advertising" specifically? or another reason?

[young hollow]

Is probably DMd to you

[distant aurora]

If you advertised then probably. Also, the automod here misfires ocassionally.

Like if you add too many attachments at once, it will trigger anti spam.

[rich stone]

interesting, yeah i got buried by notifications after so i couldnt see the reason, ill just try to be more careful

[distant aurora]

@green crystal the frequency is indeed getting wild https://cunnyx.com/i/status/2047717167638908947

OpenAI Newsroom (@OpenAINewsroom)

"OpenAI released its last model only last month, with a previous release in December and, before that, November. The company has continued to churn out new models at a crisp pace, a trend that company staff said should be expected to continue for the foreseeable future."
︀︀
︀︀techcrunch.com/2026/04/23/openai-chatgpt-gpt-5-5-ai-model-superapp/

**💬 11 🔁 6 ❤️ 168 👁️ 4.1K **

CunnyX · Apr 24, 2026, 4:39 PM

@dry kestrel is there any actually reliable Long context benchmark leaderboard that's regularly updated?

I don't trust the artificial analysis one.

New Deepseek models feel miles ahead of Gemini 3.1 pro.

Even in numbers.

[minor shard]

Gonna keep up with the VC money. No updates, no VC money 😄

[young hollow]

That's just how the Ai Landscape has been.

[distant aurora]

This seems bs.

[dry kestrel]

3.1 pro is benchmaxxed

[dry kestrel]

not that I know of aside from AA

[naive sage]

Seems like Anthropic is taking its Mythos stuff a bit too seriously https://www.theregister.com/2026/04/23/claude_opus_47_auc_overzealous/

[green crystal]

3.1 pro will be on top even when gpt-6 and opus 5 are out at this rate

[minor shard]

https://x.com/claudedevs/status/2047371123185287223?s=46

ClaudeDevs (@ClaudeDevs)

Over the past month, some of you reported Claude Code's quality had slipped. We investigated, and published a post-mortem on the three issues we found.

All are fixed in v2.1.116+ and we’ve reset usage limits for all subscribers.

X · Apr 23, 2026, 5:44 PM

[short narwhal]

Sent a discord invite link here

[rich stone]

oh thanks sorry i thought i only dmd those

wont happen again

[echo fulcrum]

#977259063052234752 message

[young hollow]

Perhaps 5.5 incoming in Perplexity

API is available now

[young hollow]

https://tenor.com/view/shocked-dog-gif-14542099238045824336

Max Only?

[serene sand]

atm

[young hollow]

So it will roll out to Pro as well?

[green crystal]

probably if they see they can absorb the token cost

[young hollow]

Is true is twice expensive but very token efficient

[green crystal]

5.5 set to medium would be enough

[young hollow]

So cost should be similar to 5.4

[young hollow]

Isn't that's how all models are

Probably High for Max

[green crystal]

nope for gpt based models you wouldnt get allot if the model was set to medium

[naive sage]

The per-token price doubled between 5.4 and 5.5, so probably not coming to Pro.

[green crystal]

with 5.5 its very different

[young hollow]

Dude is more token efficient

[young hollow]

I see

Then what are the Reasoning Efforts of other Models?

[naive sage]

That's not comparing apples to apples though... medium to xhigh isn't really a fair comparison.

[brittle ice]

why do i need max for a ai i get for 23$ on chatgpt?

[timber scarab]

Bruh

It’s for max

[young hollow]

Still it took way less tokens to complete artificial analysis index run

[ocean kindle]

No Got 5.5 for pro users?🤔🤔

[hybrid grotto]

What the hell they can't be serious

[young hollow]

It's Jover

[tawdry osprey]

Can pro users get something ?

[warm island]

Give it some time, they gotta figure out token overhead.

[oak meteor]

Why am I not surprised?

[dreamy kraken]

https://tenor.com/view/roundhouse-kick-roundhouse-kick-gif-kicking-roundhouse-gif-26059345

[young hollow]

Im not disappointed because i expected disappointment

[brittle ice]

over 100 $ price difference

[young hollow]

Because we expected it

[timber scarab]

Yea

[young hollow]

Max is 200$

[sick atlas]

OH COME ON!

[warm island]

$167

[young hollow]

Yeah we got Kimi 2.6 back

[rough drum]

gpt-5.5 only Max

[brittle ice]

5.5 is on plus for 23$

[tawdry osprey]

[timber scarab]

Wait nvm, it’s for Pro model too ig

[oak meteor]

Ohh hello Shaky-Sky-ground guy

[wise nova]

I can use GPT-5.5 Extended in my $25 GPT sub

[young hollow]

See you next time guys

Go back to work

[hybrid grotto]

No guys it's understandable

Can't blame them

It's more expensive

[wise nova]

Shame

[sick atlas]

true.

[dreamy kraken]

5.5 available only for "computer" mode. They really want people using that thing for some reason

[young hollow]

But more token efficient

[inner flint]

Then why doesn't the chatgpt sub change if that's the logic we're going with?

[young hollow]

It takes atleast 2x less tokens in comparison to 5.4

[hybrid grotto]

Just 40 percent

[sick atlas]

alrighty time to go into $170 of debt

[tawdry cedar]

I was aware of that it will be in max. Let's see how it works now in Perplexity.

[hybrid grotto]

That's just something they say, the output text and stuff still matter, i don't think they would consider that we much

[warm island]

Perplexity makes more money off computer, an users buying tokens and credits

[young hollow]

Rethink bro

[sick atlas]

?

[warm island]

A $20 Pro plan is actually cheaper and a better deal then max, bc quota regenerations, etc

[young hollow]

No it actually took way less tokens to complete artificial analysis run

[young hollow]

There are always more options

[heavy portal]

Sonnet is still less expensive than GPT 5.5?

[young hollow]

Iykyk

[tawdry cedar]

Yeah, Perplexity isn't for code, so the Pro plan is enough for most use cases.

[sick atlas]

Isn't Sonnet free ?

[tawdry cedar]

Opus as well

[outer meteor]

At this rate there will be only two model for pro plan and one for free mode

[heavy portal]

it's not, you might be thinking of Sonar

[young hollow]

Who said that

[young hollow]

That's dummy

[warm island]

As the world continues starving itself of compute...

[sick atlas]

Claude Sonnet is free on the claude website

[distant aurora]

Finally chad ai made it to perplexity.

[young hollow]

I would like to see how long it lasts

[warm island]

For 3 questions

[young hollow]

Maybe a few Prompts

[warm island]

You can get like x3 on Sonnet, and x5 on Haiku

[distant aurora]

Meanwhile grok got paywalled.

😭

[naive sage]

If that... I sometimes burn through the free quota in two lol.

[young hollow]

Mind Bending Diff

[distant aurora]

Get 100x Gemini flash on ai studio for free.

[heavy portal]

ah, I usually dont go there. Well, at least it's Sonnet that's free and not an older model like Haiku

[sick atlas]

What kind of Claude are you guys using ? Mine can generate 1300 lines of code 😭

[warm island]

They are so generous with exactly 1 Sonnet thinking message every 5 hours

[sick atlas]

True.

[warm island]

Dude those Free api limits are fire, you can get like 50 messages to 3 flash lite

[hybrid grotto]

Google's new model is expected in a month, when Google releases a model, it crushes all others for a while, so we gotta just wait it out

[naive sage]

-# It... it is impressive though.. 😢

[tawdry cedar]

5.5 uses tools better as compared to Opus 4.7, so that's good.

[distant aurora]

Claude mythos 4o

[heavy portal]

good to know

[young hollow]

i use Myth OS

[green crystal]

you forgot to add "on benchmarks" lol

[distant aurora]

Android custom ROM?

[warm island]

Google still sucks at coding and CLI tho, only redeeming factor is vision. If they could get their code togehter, they could have a good LM.

[distant aurora]

Gemini personality though>>

[heavy portal]

actually, it says GPT5.5 is "default". Is there a setting where we can change the default?

[green crystal]

in real world
benchmarks make it seem like it should be close to opus 4.7

[young hollow]

No Claude Rom Max 100x

[distant aurora]

Gemini might be dumb but I love talking to it.

[young hollow]

5 prompts everyday

[naive sage]

Did anyone solve Theo's challenge yet, or should I keep working on it today?

[distant aurora]

I don't have to engage in unnecessary debate with Gemini.

[warm island]

You cant, youve gotta press another one manually

[hybrid grotto]

They should be introduced gpt 5.5 as s different model instead of doubling the price of their standard main series of models

[young hollow]

Look how much less tokens 5.5 took to complete Artificial Analysis Run in Comparison to others...

[tawdry cedar]

Cursor changed new models to max mode as well. I am old request-based pricing and used 11B tokens this month out of Opus 4.6, lol.

[green crystal]

fun writing style yeah but the sycophancy keeps you on your toes lol

[distant aurora]

Efficient af

[young hollow]

Xhigh just took 68M

[distant aurora]

I would rather have it than get stuck in an argument loop.

[green crystal]

so efficient that im suspicious of it when it completes reading and reasoning over 19 files in 6 min

[distant aurora]

@green crystal I only wish the context awareness was better. https://g.co/gemini/share/40a354c3b226

[green crystal]

yeah

5.5 can remember very well early parts of conversations

[modern crown]

lol @ perplexity for not letting pro users use 5.5

💀

[hybrid grotto]

Hehe i got sources

[distant aurora]

Gemini tends to ignore stuff within as less as 10 past messages.

[hybrid grotto]

Can be used for free

[distant aurora]

How big were these files?

[distant aurora]

Add medium and high to this graph.

[green crystal]

pretty small like 3mb ish each
they were deepseek papers lol

[edgy dock]

How about us, free users, can't the company just host a 300b model on hugging face to not rely on third party models?

[green crystal]

but 5.5 doesnt seem to read with python as often as 5.4

[young hollow]

Yeah give me a minute

[green crystal]

5.4 needed to read line by line with python

[left stream]

it's been so long bro, I have gained an increase in ai power thanks to a cybercrime detective

[naive sage]

It's very good at getting quick questions though... flash is great for that. But if it's something it doesn't get right away, might as well ask GPT or Claude. Nitty gritty stuff Gemini isn't quite as good at, imo.

[distant aurora]

Joe in chat.

[green crystal]

5.5 just opens up files and uses it in its entirety

[distant aurora]

Spud ~ mythos though.

[tawdry cedar]

Yeah it does mine to

[distant aurora]

Actually

[echo fulcrum]

5.5 on max isn’t surprising $30 per million output is crazy 😭

[distant aurora]

Gpt 5.5 might be hella token efficient and cost effective. Claude got mogged on both mythos and opus in terms of efficiency.

Crazy.

[keen shoal]

"GPT-5.5 is now available on Perplexity for Max subscribers." 🇮🇱 🇮🇱 🇮🇱 🇮🇱 🇮🇱

[green crystal]

5.5 codex might be allot closer to mythos

[young hollow]

@distant aurora

[naive sage]

It makes me wonder if they're hiding more of the chain-of-thought/thinking, which would make it more difficult for people to train on, as well as make sense for it being more per token. Plus, people using 5.5 Pro seemed to say it 'hangs' a lot... like it's probably thinking/doing something, but not showing it.

[green crystal]

it being more coding oriented and all

[distant aurora]

That medium is crazy

[echo fulcrum]

True

I wonder if they’re going to release a 5.5 mini API that’s cheaper

[young hollow]

SOTA Efficiency

[naive sage]

Since benchmarks are now around token usage, hiding thinking tokens just makes sense, it's their way of gaming benchmarks as well.

That's been a problem since forever, whatever the scorecard is... people build to make the score look good.

[green crystal]

5.4 mini is hilarious
makes it look like its reinventing the wheel 10 times during every prompt

[young hollow]

💀

[distant aurora]

@green crystal @young hollow 5.5 medium being as cost effective as Deepseek. 💀🔥

[young hollow]

https://tenor.com/view/help-gif-9858523764566581784

Holy Hack

Mind Bending Efficiency

[distant aurora]

Keep in mind gpt is like atleast 4 to 5 times more expensive per token.

[young hollow]

So 5.5 Medium is most Attractive in Cost to Intelligence

[sharp granite]

🐳

[tawdry cedar]

5.4 Mini even in low thinks like 5.4xhigh

[young hollow]

Yeah while Deepseek used 240M Tokens

[sharp granite]

is this per billion tokens price

[young hollow]

VS 22M

[distant aurora]

No. That's the whole cost of running all the benchmarks per model.

[tranquil schooner]

ooo my toes tickle

[sharp granite]

oh

why is perplexity no longer great

[distant aurora]

Rate limits.

[tranquil schooner]

its da best

lowky

[sharp granite]

actually perplexity is pretty stupid

on free

[distant aurora]

Though I would say the new model upgrade would make it decent enough for daily use, people will disagree.

[tranquil schooner]

it helps me trickle my toes

lowky

[distant aurora]

And I understand why.

Considering the low rate limits and context limit.

And rerouting issues.

[young hollow]

Nah Bruh

Skill Issues

Perplexity Dh GOAT

Most Sigma Ai

[rancid storm]

Why can't the PRO subscription use the GPT-5.5 model?

[heavy portal]

yeah, looks like it's Max only which isn't surprising

[echo fulcrum]

In the future the only model you get with pro is sonar

[young hollow]

Because Pros are not Sigma

[distant aurora]

Huh?

Damn.

Okay.

[distant aurora]

Can't defend perplexity anymore.

[hybrid grotto]

Cost efficient my ass, i used a simple prompt on codex, just requiring it to write a small para and it took like 43k tokens

[covert compass]

Why can’t I generate images anymore. Message is joined

[young hollow]

They are never supposed to since the day you changed your name.

[heavy portal]

kind of weird that they allow GPT5.5 and Opus4.7 inside Computer for users but not for regular queries

[young hollow]

Is forbidden in many regions tuff Luck

[heavy portal]

I guess they just adjust the number of credits used

[young hollow]

Computer is Expensive

[distant aurora]

Damn bro

[echo fulcrum]

You pay for the credits you use on computer

[rancid storm]

I noticed that GPT-5.5 can only be used with a PRO subscription in Computer mode.

[distant aurora]

Honestly, paying users should probably switch.

[covert compass]

I’ve been creating images for so long as I’ve had the Perplexity Pro

[young hollow]

They probably had long ago

[hybrid grotto]

Doesn't matter, credits are paid, you can't get them for free, pro users just got 2k once as a one time and that too for some kind of compensation for service down

[young hollow]

[young hollow]

Not anymore

RIP.

[covert compass]

Since when?

Today?

Sorry. Hello, everyone!

[young hollow]

In your case yes

[covert compass]

Can you be more explicit, please?

[young hollow]

Soon they might remove image gen entirely

[covert compass]

What happened?

[distant aurora]

Imagine some Chinese ai company buying perplexity.

[young hollow]

This is been happening for many regions

It just hitted your region today

[covert compass]

Even in France?

[young hollow]

Where it doesn't work

It doesn't work

[covert compass]

Ai. True. Sorry 🤣

[young hollow]

[heavy portal]

some Pro users got 4k but yeah, I realized after I said above they probably just charge more credits when using the more expensive models

[young hollow]

Those were paid users before pplx computer released

[hybrid grotto]

Will trump destroy US first or perplexity itself?

Tough competition

[spiral harness]

.

[tawdry cedar]

5.5 times faster than Opus 4.7 in perplexity. Not sure which thinking effort they use by default.

[serene sand]

#💗│sharing message

[young hollow]

https://tenor.com/view/clapping-clap-applause-standing-ovation-gif-15648815

Ty Mares

[somber berry]

I also have a question for everyone what's your typical burn rate for building out a small mobile application in perplexity computer? Anyone have tips for saving credits?

[twilit crown]

Why no 5.5?

[distant aurora]

https://cunnyx.com/i/status/2047755477295968623

Lisan al Gaib (@scaling01)

DeepSeek-V4 SVGs
︀︀
︀︀pretty good, better than all other open-source models imo
︀︀
︀︀but except for the skyline which is basically a replicate of the Sonnet 4.5 result, they feel a lot like Gemini 3/3.1 Pro outputs

Quoting Lisan al Gaib (@scaling01)
︀
Gemini 3.1 Pro will be a massive step-up!
︀︀
︀︀There's a decent chance it's on par with Opus 4.6 and GPT-5.3.
︀︀
︀︀The main reason for that: similarly to Claude 4.6 and GPT-5.2/5.3 it thinks much longer than Gemini 3 Pro
︀︀
︀︀The same request on aistudio, tested multiple times, had 6 thinking summaries and took always around 20-30s. Gemini 3.1 Pro however thought consistently around 3 minutes with up to 40 thinking summaries for the same request!
︀︀
︀︀So here are my usual SVG vibe-test results. They are incredible, but almost feel sloptimized with the same sky everyhwere (that frightens me a little)

**💬 3 🔁 2 ❤️ 31 👁️ 1.7K **

CunnyX · Apr 24, 2026, 7:12 PM

[twilit crown]

I get that its expensive but at some point they will have to drop 5.4

[bold karma]

wtf are they doing

[distant aurora]

I can't tell whether perplexity is greedy or desperate at this point.

[twilit crown]

Why not both

[distant aurora]

Gpt is cheaper than sonnet.

[bold karma]

ever since they released perplexity computer perplexity has gone downhill

[young hollow]

Time will pass

Eventually

Because we won't be waiting anyways

[distant aurora]

I don't understand why they would hold it back.

[young hollow]

Ask Arvind Srinivas

[twilit crown]

Double the price

Its now as expensive as Opus

[young hollow]

Bruh it's way more token efficient

[twilit crown]

Doesn't matter. Perplexity already uses minimal tokens

Unless you are doing coding, this just makes it more expensive

[young hollow]

Very Strong Point

[distant aurora]

5/30

WTF

Crazy

Opus level pricing I guess.

So opus 4.7 is cheaper than gpt 5.5

[twilit crown]

It isn't even their expensive model

[distant aurora]

That explains a lot.

[twilit crown]

The new Pro varient is $180/ million output tokens

So thats like 5x the price

[distant aurora]

Good thing Deepseek exists.

[twilit crown]

They don't use it though

[buoyant schooner]

gpt is much more token efficient than opus so not really

[twilit crown]

Also they cheap out on tokens even for what already is being used
Responses are only "Prepared" by your model, they use Sonar

[distant aurora]

I know.

[twilit crown]

Which makes the token efficiency point useless

Since all the model does is write the response

[buoyant schooner]

really?

[young hollow]

It actually uses more tokens and has new tokenizer which makes it 50% more expensive then 4.6

[twilit crown]

Thats why it says prepared in the bottom