#💬│general
1 messages · Page 293 of 1
how do i fix it
It really depends on the benchmark, theres a reason to why he quoted that tweet
Yes, we are all aware of this, unfortunately
https://discord.com/channels/1047197230748151888/1403076559207796927
https://discord.com/channels/1047197230748151888/1403268394475716679
Do you have a pro subscription
yes
.
so now what will the ydo?
also is there any possibiity to add something like canvas or code preview?
Fine tuning a berd bot might actually be possible
I need to format the messages though
iam waiting for 20B-oss-Berd
Yoo
780 items, and the message.jsons include some useless stuff
nothing not a quick python script cant fix
true, I need ti find a way to like filter out discord links though
Uncle is patrolling the hood.
regex and ai
hmm
@dry kestrel I'm planning on running GPT-OSS-120B locally
Have u tried any model ?
the 20B one also
I've gotten both the 120b and 20b variants working
what do u think about it
and how much resources do it take u
around 120gb of RAM
and like 90% of my GPU
holy
I'm on mac though so its a bit different
u got a vps or it's ur main computer ?
main computer
This might end in a disaster
128
It depends on your VRAm though
on my gpu ?
I have to load the entire model into RAM due to how apple silicon works
Yeah
If you use LMStudio it should tell you which models you can load
@dry kestrel How Can I Know Proper Perplexity Token Limits?
can u send me the link
and it'll stop you if the model is too large instead of the computer crashing
i see
Berd?
You can tell the model to "repeat after me" then give it a like a long sample text.
There isn't any official documentation on the max output limits
DeepSeek
s l o w
Or i can just tell it to keep writing Until it Hits Maximum Token Limit
Stating Tokens/Characters/Words in The End
Why You Guys Want to Run Models Locally
When You Have Subscriptions
Comet update when
NAWW
OpenRouter API response error: {
error: {
message: 'Rate limit exceeded: free-models-per-day. Add 10 credits to unlock 1000 free model requests per day',
code: 429,
metadata: { headers: [Object], provider_name: null }

@dry kestrel Would This Work?
openrouter has a limit of 40 messages/day
so much for the free
No, max output limit is like the max tokens per message.
depends
rip
Comet needs some more like idk
when i need a fast answer
To become my default browser
Ohhh
Then I Will Know The Limit Per Message Right
O
convince u
To use comet
Well I Trust You
This is probably the best rate limits you are going to get for a free API
Brave is pretty good
I'd just buy some credits for that 1000 requests a day
i said better
it really depends on ur use cases

alright
fuck capitalism
Ahhh So Subscriptions Aren't Always Enough
ayo chill
There is A Price for Everything in This World
compute costs money :/
That's The Law of The World
uhh
EXACTLY
We Should be Greatful We Can Use Many Models for Free
not like the companies run in a loss offering free models
if anything it boosts their sales
Of Course Because then You are The Product
The Ultimately Utilise The Free Data
Is Like Watching YouTube Ads
yez
@dry kestrel i can run 20B pretty smooth
i wonder for other model
10GB of ram for a Hi
lmao
wow
We Need More Efficiency
for us
it's a trade
we bought it
nice
so it's kinda fair getting the service
I mean for Free Users
Guys, someone sent me a bit ly link saying its free perplexity pro
And it worked lmao
I know I'm on the Perplexity server but... My one year subscription has just expiring and I'm wondering which AI has the best performance this time around. Should I consider resubscribing to Perplexity or should I go with ChatGPT?
I mostly use AI to search for answers to my questions
Perplexity better tbh + Search feature is too good
Always gives uptodate information
@snow remnant
If we're talking reliability then perplexity for sure
You can always relay on perplexity's anwser
Moral of the story
Hi
Hello
Looks like I can use gpt5 with Perplexity. Does open ai offer the ability to use different models if you get their pro plan?
nope
Perplexity it is then 
LMAO
wtf
I think I cleared up all the data, trying to run it in the google colab now
this could just be a total failure or it might just work
gpt 5 can't render latex right too 😔😔👉👈
that's sad
Windows users crying
Os creating a lot of unnecessary cpu load
Vencord 👀
I was running a local LLM model buddy...
vencord moment
ye why not ?
Does it run smoothly?
so r u testing or use it regularly and fine with spikes
Ye i'm fine with it
hey guys, how many requests are there for Chatgpt 5? I'm thinking of buying a team subscription or Perplexity Pro
K feel like the performance gap between Gemini flash 2.5 and 2.5 Pro is too much that I think Gemini Flash 3 wouldn't come close to Gemini 2.5 pro
I feel like depending on how gpt 5 mini turns out
It's all depends on how gpt 5 mini will crush flash 2.5
Hi guys, does anyone have invite for Peplexity comet? I really want to use it but paying 200 for it is a little too much right now
As a Pro user, you have a quota of 300 queries across all models (including deep research) within a rolling window. When you use queries, they slowly regenerate over time. If you consume queries faster than they regenerate and exhaust your 300-query quota, you'll experience heavy rate limits. Labs has a limit of 50 queries a month. Video generation is 5 videos a month with Veo3. GPT image 1 is 150 a month specifically, other image generation models are unlimited.
in all or in each
all
it should work, lora finetuning is very efficient
Nuts
600 per day
Not true - #💬│general message
I'm having a bunch of issues just getting it working
Starting to hate this
import unsloth first?
I... did that
It's going in circles atm
ib the code is generated by gpt
agi moment right there 
Weird, i see 600 here: https://www.perplexity.ai/p/api/v1/user/settings
In reality it is 300 - (I've tried this myself)
Oh, alright. Must have been hard to reach that limit
Was using the base models + testing some things out
dont read too much into this... pplx is now abandoning this rest api from my observation since this summer
Interesting, I noticed that it doesn't show the video gen limit
How long did it take you
for example pplx_alpha_limit is completelyt meaningless
Yeah how long did it take
I think like 2 hours ish?
I honestly don't remember
I gotta look through my own messages
hidden in the cloud. once you reached the limit the server will add max sub upsell in eventstream
this happens when you reach the limit #💬│general message
"upsell_information": {"name": "subscription_upgrade_video_generation", "upsell_type": "UPGRADE_TO_PRO", "app_location": "IN_THREAD_INPUT", "title": "You've reached your monthly limit", "description": "Create more videos with audio and enhanced quality", "cta": "PRO_UPGRADE", "icon_reference": "max", "button_color": "PRIMARY_COLOR", "button_text": "Upgrade to Max", "cta_information": {"upgrade_paywall_title": "Want to generate more videos?"}, "show_in_comet": true}
Yeah, I'd like to see this added so that the user has a general idea
Goes hard
Finally working
file upload limit #💬│general message
@short narwhal Why are back to matsku and not GPT-5 Nano
Nano died
Costed 10 feathers
Lol
and a bag of millet
Okay gpt 5 mini crushes flash 2.5
generational debt just to add a few gbs of extra ram in your mac 😭
Not that much
Hm
I think time for Google to ship flash 3 next month
and it's the 14 inch so a bit cheaper
does sonar model also subject to this daily limit?
previously the default model (mostly gpt 3.5) w/o pro search/copilot is truly unlimited
Nice
It does, at least when I tried in May
is it the m4 one or older ?
m4
a lot has changed in recent months...
I wish I did more screenrecordings :/
True
damn m4 is actually pretty darn impressive, but ngl ive been on m1 for a few years now and it still performs through everything like a beast
Yeah the apple silicon leap was insane
Whe are you gonna buy one?
actually much bigger of a deal for me than AI ngl, because getting such a performance in such a sleek and lightweight ... fanless?! machine is insaneee
yep
Hmm... M4 is in the performance of the m1 ultra
I'm a bit annoyed that they still haven't fixed how the keyboard gets inprinted onto the screen
I can’t figure out how to edit my server name anymore
THIS....!!!!!
Too broke rn
But I do so much more in Comet, maybe it doesn't apply they're?
...
you know when i got the laptop and i used it for sometime ... i thought i broke something or did something to the screen and it got me so worried
arch linux btw
Get an air!
I want a top of the line one with atleast 32gb ram but it's so expensive
60hz
i love flash, it is the best model for repetitive & non creative task
(I keep mine at 60hz most of the times)
and dirt cheap
ahem... 😒 you got comet (please share invite code if you got pleaseee) 😭
get a good windows laptop and install linux
I'm surviving just fine on it with my air
I have 2 laptops and a legion go already
I wanna try mac
They've already gone long ago I'm afraid, sorry..
Peasant behaviour
or just get a laptop without windows and install linux
oh we in the rich people territory? 
erm airs are fine for light work
Smort
install linux eitherway
Do people still use Windows?
WHY
Yes peasant work
battery life mostly
yes
Hasn't Microsoft completely abandoned them?
unless I actually need the 120hz
This is crazy work
no worries, there is such a scarcity of invite links and i applied for access such a long time ago like as soon as it got released and still havent recieved anything
Yeah
Not quite. Never had an issue with my high refresh rate screen of my MacBook Pro
it took me a looong time to go full linux
Comet needs to fix its crappy low refresh rate animations, it's silly
mostly because i wasnt dedicated to it but windows is actually good. dont take the crap from people
80% of people use windows or something
I mean yea but mostly cuz of app support
windows is alright, if you are into gaming but that's really all the usecase there is for it. If you put that aside you can literally get a much better and lightweight setup with linux
Windows 12 have been delayed for two years now, so idk.
50-80% of that 80% don't even know what Linux is
windows uses ram, true. but the gc isnt bad or anything, ram unused is ram wasted, and when other applications need ram os uses ~1-1.5 gb of ram
Ive optimized my windows a lot, i dual boot
(i don't)
whats gc ?
garbage collection
Did you guys know I was in the room when Linux was presented for the first time
memory related stuff ignore it
I mean I have them on a VM on my Mac but I don't really have an app that specifically requires windows, maybe some old legacy ones
9% of people running unknown os
My people
windows respects your time (i.e after removing bloat)
Stone tablets
f nvidia
Why is the gang named after got models
what now
Can someone fill me in
i had to bang my head against the table several times when i started using a server linux, then went to arch, banged my head more
and why are you GPT-3.5 now
.

I realized nano is bit too fast for me
Use gentoo
Mat is 100 years old confirmed
Huhh
that's not the only issue with windows, the issue is that its a corporate junkyard inside there. You dont have control over your own machine as much as a user should have, and if a user is required to go through like 100 youtube tutorials and bunch of tech articles just to make things work then its really not worth it
Guess I'll be GPT 5- Thinking Pro
It's rlly ez to set up trust
Hmm
womp womp, dont you have to do 5x the work for linux? and for what windows gives it's good
wen matsku bot
you strike me as someone who hasnt experimented a lot
only love ubuntu / debian / arch
I'm convinced you are 90+
if you use kde plasma on linux, the ram consumption becomes almost similar to that of windows
Not really, just use Mint or something light and user friendly
Me too
you havent used it have you
more evidence
gpt-5-thinking longer-pro-max-ultra-mini
That's totally alright for you to have an opinion about me, but my opinion still stands
I have Mint, Ubuntu and retroarch as virtual machines on my Mac
uhhh
Let’s have a moment of silence for her
I'm going to run out of free compute
only 6 hours for 1000 epochs
the intention of my question didn't involve opinions of you, I simply stated that because kde uses almost as much resources, for what windows gives with easy tweaks (like tiny11 or atlasOS), it's a great deal
and mint is kde
free tier on t4 will disconnect if you were inactive
if your first instinct if to get aggressive instead of having a healthy argument then you're really not worth the time for this conversation my friend
As an arch/fedora/debian/ubuntu/mint/windows/popos/endeavour/rhel nonabooting user, I agree
I need to find a way to do this on my Mac
I should install gentoo-
Decabooting off a single drive, what can possibly go wrong
use bare metal
Also almost at the 100k message mark
who needs an os
Just use pen and paper
fr
punching holes in my card rn

Just hand build transistors for everything possible to solve or sum shi
Fr
tails os for paranoid
Just get a bunch of people
love staring at an empty black screen with at @
reminds me of that guy who build a computer inside minecraft lol, bro literally built the whole architecture from ground up
@dry kestrel show up in my model selector plz
Dw I’ll be there
Never!
||i could risking it by running this script on shared cluster without permission||
😂 i have seen this before somewhere i just dont remember but ... very faintly remember the context for this. Do you know where its from ?
please dont
Three body prblem
it is a100 & h100 cluster
ohhh

☹️
its right there
I think I could do it on my mac with mlx instead of unsloth?
Wow
For mobile or pc?
You need a paid account on ChatGPT
here vvv
on perplexity
dammit you were fast
It's legit the chipset thingy
for the first time
screenshot wars
My servers are dying
try the serverless servers
Fr
pen n paper?
lul serverless server
Makes sense ur in peak hours rn
Have you tried running in the cloud instead
I'm poor
i could build an AWS wrapper over this and give you with a monthly fees
Just shut them down
What service do you use for this tho
I have no idea what I'm doing
No preserve root
||free from research institute||
i should run it sometime

I agree
everytime i see someone say this joke i feel so cringed, its so overused now that its lost
because what is life? 
Last time I did that, holy shit did 7 oses not get nuked
its not a joke
run it
I'm going to see if I can buy some cheap compute
omg please... i love my french language pack files

😔
can you ask chatgpt to save the checkpoint every x minutes and load the previous checkpoint to continue the training?
W
but a good life lesson tho
Maybe? I don't know if thats possible
Yea about that uhhh I installed gentoo after that
💀
you didnt learn anything did you
lmfao
tried to use it for 5 mins
😭
And then switched back to ubuntu
installed or simulated
I wondered what % more the head banging would go up by
But then my touchscreen wasn't working
..okay finally
if i switch to gentoo
....
Installed
oh god
Then I tried arch again
no wait
s2 has enough raw power for any of this?
Display manager refused to work
you keep making this whole timeline worse for me
Modded
linux has the following system requirements :
- electricity
- device (optional)
idk
Tried to install mint, didn't even load shit
I think I can rent a 4090 24gb for like 7 USD a day
Electricity is optional
Am I being a shill if I say I prefer grok 4 compared to gpt-5? It just response more unfiltered and more engaging
both are optional
but at what speed 1k operations per second on samsung s2?
We dun care it's yo opinion
linux is omnipresent
U prefer it
Gng android is still technically using a Linux kernel
no, both has their own usecases tbh, if i were to speak from my experience
chatgpt : for learning about a topic, or getting explanations and summaries
claude: code generation
grok: if i want unfiltered and unbiased response
gemini: when i feel like im the worse it reminds me there's still something much worse than me out there
heavily modified in user space and kernel
but okay... idk if ya'll agree on this or not but out of all the AI web pages perplexity probably has the best design and interface
no one's even close, everything feels so well designed
perplexity doesnt lagg when the context gets huge
!!!! EXACTLY!!
the way chatgpt, grok, gemini and llama do
and the animations are so buttery smooth
yea
i can just watch it generate stuff whole day long its so mesmerizing
And we don't even talk about copilot when it generates a 9k line code for no reason
I thought of it as a glorified search engine but dang this turned out to be good
lmao
Perplexity is no doubt a great alternative for someone who wants quick and bite sized information in their day to day life
does perplexity give api tokens?
Liek I tell it to make a simple python print statement and it whips out a fucking 190 line code like wtf
Idts
you'll find yourself using ai for A-Z everything
heck you wont realize it
I ditched chatgpt I wasnt thinking straight
You get $5 worth of API credits a month as a pro user
because it takes away the pain of googling something, then not getting the response back that you expected... then writing "[question] reddit"
@dry kestrel When You Became GPT-5 Pro
thanks
10 minutes ago
Wait fr.?
Ahhh is Fresh Nice
5$ ..
gemini is so trash at every single thing like... idk what to even do with it i have free 1 year subscription of their best models and everything along with it
For Sonar Yes
Yep
Ask for Recepies
Figures
head to https://www.perplexity.ai/account/api/group and try it out
Sonar ain't allat
Free cloud storage is always nice
Icloud > google storage
Can't really provide other LLMs through an API, would go against like every providers ToS
there's like 2TB of storage for a year, there are other stuff as well but i didnt bother checking it all out
type stuff
Gemini got me
😂
like.. i dont wanna hate it. I geniunely want to have to say good things at least a few good things about it but there's just nothing i can come up with
i tried veo3, idk what to do with it other than if i want to be a catfish
does google give free api tokens?
there text to voice takes a whole lot of time to even generate a simple prompt
i dont mind an absolute trash ai i just need api tokens without a rate limit of 3 effing messages 😭
Yeah, but that comes at a cost of them training on your data (+ slow API/rate limited)
yeah im currently using its API for code completion
its not like im not trying to give it a chance
who cares about data, it's stolen eitherway, rate limit is my problem
rate limit?
i need it for a chatbot on my server
no no im on their pro plan idk how much limit it has i havent bothered checking
i got like 600 ppl so the rate limit is a big big issue
oh another rich man territory I've wandered upon
pardon me
I don't think any free API is gonna give you that
orewa kaizoku
yea i gotta spend money at this point
nah nah, just that 1 year free on student id
it was a fun side project only
I Got 1 Year Free by Airtel
on perplexity?
yeah that as well, for the Indian users
Obviously
Airtel is 360M Customers
yep
Is Obvious
I wonder
Openrouter!
companies want more users, but they dont understand the quality of users they are getting this way is shit
was it a collaboration, did perp pay airtel, airtel pay perp
don't they have 40 messages a day?
you just charge 10 credits
This is Jio Tactic
wrong territory to say than in 🏴☠️
and you get up to 1000 free requests a day
is that 10 buckaroonies
its 10$
Yes
I mean not bad
Thats really not too bad
kiss my ass openrouter
and they support a ton of payment methods
We Survive on 2-3$ A Day
I do know india
good things should never be free, if internet was not so cheap only people who actually have a good use for it would pay for it.
if you think of 2-3$ a day you're delusional beyond measure
crap ideology
90₹ is our meal cost of 2
i can give you a example to look back and do your case study on
cut the crap bro im not from russia or smth I know
thats b.s
it simply doesn't fit my view of the world, you may hold your opinion
look up for state of the internet before and after Indian users were given free internet access by Jio
Believe Whatever You Want Why Would I Care
1 gb per month
What are yalls default browser
Totally fine, im not offended or trying to offend. Just what i think
Brave for now
Same Happening With Perplexity
let's say you love pizza, and I too love pizzas, you're proposing that since pizza tastes so good, we shouldn't let anyone else have it, I'm saying since pizza tastes so good let's make everyone have it
not comet 
Comet comet comet
whatever supports extension on mobile
dreaded Microsoft edge beta
Kiwi?
I'd love comet but brave has goated adblockers
You get adblockers on Comet too
I don't like Firefox though
why would I care? 😭 bro are you >9
Firefox
no im sorry age doesnt matter, ive seen smarter >9
Good Grief
They have to add some fundamental features found in chrome in the Android version
surviving on 2-3$ a day? possible? yes. does it fit the context of what you're referring to? HELL NO.
It's missing some very basic like per website zoom
Don't Know Any Better Alternatives
Just Forget it
or... lets let pizza be something that is adequately priced in a way that people who actually do like pizza and appreciate pizza as a food would be pay for it to eat it, if you make pizza free or so cheap that people who dont want to eat pizza, who dont appreciate or like pizza will eat a slice and waste the rest just to talk shit about pizza.
the people who live on 2-3$ a day are not perplexity users
nor do they have the priviledge of entertainment, your entire logic falls apart
Yeah 👍 only 2 engines exist blink, gecko
Agreed
Business is Business
what’s happening here
Ahhh so There are No Other Options
Bananas
forgive me good sir, that ideology crumbles my thought and will
there should be careful use of the word affordable and cheap
free internet in india was released in 2016 by jio
calls dont use internet 
if these words aren't used in the correct places like then cheap means they consider it's a low quality
they do
call centers have a different kind of operation
ig they might, I don't know. sorry for pointing it out for no reason
Opportunity for making money
@languid isle You Can't Blame Entire Nation Because of Some People
friendly banter 
do you think a call center operates by paying the international call charges for every call they make ? or would it be cheaper to make use of VOIP
jus scam in your own country ez 💸

I am not blaming, im doing my masters and some of my great friends are from India. Please dont take me wrong here. Altho i apologise
you can and you should
Ahhh My Bad I Had A Pre Bias Wrong Assumption
nah nah its all cool, i couldve framed my words better im just really lazy and tired right now
Same Man
I Can Understand
because what is blaming other than words given weight by one's own mind that has societal impact respective to what role you have in this stupid hierarchy 
You Can Just Say "Don't Take it Offensively"
just tldr me this
😭
you can send a
and go to sleep
whats ur guys opinion on jailbreak prompts
😕
Yup Another Life Needs to Touch Grass
ok
Gray Area
lol
or you can have a conversation but that's probably too far fetched
Can someone tell me how to get comet?
I hate people who say "touch grass" as if they're not on a discord channel talking to random people online anyway
I can't see any potential

lmao
oldest scene in the book
You don't understand the real message

You'll need an invite from someone who already has it, or buy perplexity premium plans
It was meant to be a joke, I couldn't make a TL;DR shorter than what i'd written if I wanted to
I've the pro subscription btw. will it work?
i said that as a joke as well
it was intended towards me duh
Pro subscribers get priority on the waitlist.
why are you writing like that
Bad Habit
I dont take offense
I can just disappear off the face of the internet
But won't I need to join the waitlist? Or they already enroll the pro subscribers by default?
Great Idea!
no i meant, uppercase every first letter of the word
1/10
Sometimes even Pro users in a waitlist... Says how much there is demand for comet then
Yeah That's Bad Habit

🥀
You're the only one not doing it
huh
and me
..yeah
Yup Sarcasm
its okay

are you one phone by any chance ?
Obviously
You do need to join the waitlist
i cant imagine someone typing that way on a keyboard

What About Android Waitlist
Thats not a thing yet
Why tho 😭
@dry kestrel why's ur name gpt 5 turbo berd
He has Became 2 Ai Models
Certainly
they shouldve had 2 lines just like how disney does it, one for normal people for it and the other for VIP's
Why not
VIP Had More Data then Normal Users
Even Today VIP Have Netflix/Hotstar/Prime Benefits
Pro users already get priority on the waitlist, and the more you use perplexity, the higher the algorithm puts you at the faster you get access.
zzz
oh i didnt know that! hmm
ayy letme join the fun
perplexity servers are managing this workload? the users must've skyrocketed
but ive been a pro user for a long time now i shouldve gotten it :(
look at this banger
My kind of drink
Sake ?
Recency bias 
It's gin
how to get the pro tag... Out of curiosity i got the pro subscription and I don't know what to do with it
Not My Kind of Drink
A kind of alcohol?
idk how I got it I bought pro and landed here through some links
yes
It’s just cus it’s Japanese right matsku
Maybe
Lol
i bet it makes an UwU sound when you pour it
um
No
no tf
...
you're confusing it with Sake a bit
Need Manual Action
my fault
Why do people speak their mind so expressively
and sake is associated with
a lot of times
I didn't even realised I have been in this server for 2 years
why not
im working on that part
Actually Bottle Fault
Until I saw the join date
Like a few things should just not be said 😔
It’s okay you got this
Pre Psychological Programming
if i had not said it here now, i wouldve said it somewhere else
womp womp
it just had to be said
Flamboyantly
Oh god this is one of the very first servers i joined when I created my discord account
What is That
oh my god
What
!
wow


i just got a notification of app update on a website!?!?
Has Huang Invested in Perplexity?
check out e-girl servers, be sure to give authorisations to those applications so they can hack your account properly
Also ur credit card

Nvidia has
That's Just Huang
e-girls server
looks inside
it's all e-boys
How About B Girls and B Boys
disappointed
we dont talk about those
Thank You
you seem to know a lot about this particular matter
😎
What's grok 4 heavy
Now We Have Many Ai Models Here
A model
A very pricey model
Heavier One
I thought only the vannila and the mini existed
@short narwhal hello peasant
BMI 40 or greater
???
No it’s like a bigger one or something
3.5 Turbo
I join
So a bigger model...
No, it runs 3 instances of Grok 4 with higher access to tool calls
It then picks the best response
It uses the same base model as Grok 4
It's kind of interesting, actually... it spins up several instances to answer the question, then has a separate 'judge' process which reviews and merges them... kind of like automated n-shot
They kinda close sourced grok from the third model
Can't I make my own grok 4 heavy then
You can
So yeah it isn’t more parameter count
matter of fact people has already done that
The parameters are the same
Nice
Lovely thank you
is there any way to use web integrations to bypass api keys and use desktop chatgpt 
Lmao
If i ask Grok 4 same thing three times now thats Grok 4 Heavy
Isn’t there a super heavy or something
Nah
Nope
how can it be super heavy ???
I think meta will close source it's models when they make a very powerful model on par with the flagship model that's there currently
wtf does that even mean
@dry kestrel what do u rec on LM studio
heavier

Which model should i use beside gpt oss 20B
Hi
Gemma
Is really good
also this
what is your use case ?
You Should Use
A
Good Model
Funny both are made by openai and Google rivals of each other
qwen 3 is also really good
do you just have gpt 2 for the memories?
Gemma vs gpt oss
Hmm coding and studying
yeah lol
Then Why Perplexity Doesn't Have it?
GPT-2 is a very good model
Quen 3 32b 👍
claude and chatgpt free versions should do the job i thnk
Minimax M1?
Claude’s free rate limit is so bad
I want to mess with the local LM
@short narwhal so is bard
I got chatgpt and claude also
ah are you studying machine learning or something ?
I just want to mess around with the model
google bard?
its fair then
true
Yes 😎
The Qwen 3 235B is the largest Quen model
Ikr, forever in my heart
Shu Qwen not Gwen
And the coder varient is in the top 15
I think berd is a derivative form of bard

Were y'all even using ai when gpt2 was out
Which Year
i dont mind actually, if that's what it takes to keep things working for them on their end. And if i want more i would rather pay because the quality of code output is worth it
nah
Idk i wasn't around
Wait this is good but how much VRAM you got
For home use, besides those mentioned (especially oss and qwen3) - I think arcee's models tend to be good (they tend to be spins of other models that are a bit older, like qwen2.5, but are good for their weight).
Ai only broke out w gpt3
Rtx 3050
I think gpt 2 was used by AI Dungeon at some point?
32GB of ram
Oh so not a lot to deal with hm
Shall models would run
ye kinda
which vram model
but it suits my need
Under 8b
Is Minimax m1 good
When open ai was a non profit
At most
what was the first prompt ?
It's so insane to think about now
Why Openai is for Profit
It was like that when established
I’m guessing it’s 8GB?
They should name it Closed AI
It was created for making Accessible ai
Exactly
Priorities changed because of money
Sammy B became greedy
SAM!!!
Do y'all remember when they kicked Sam out
He has a regera 
Was a crazy week
imma let this download
To hire all employees
then go to sleep
True Atrioc Viewers enjoyed that week
gn guys
classic move to lure all the backstabbers
Basically saved sam
Adios
Exactly
Insane Big A reference
Microsoft saved sam in exchange for his soul
Glizzy Moment
Coffee cow
With $10b investment it would be as part of Microsoft someday
What if OpenAi Waa Closed
It is about to be
And Everything Happened in Microsoft
and now this would be Google vs Microsoft battle
Sam would have majority stake along w Microsoft
With Sam
I think Google flash 2.5 could make apps
The others were vc types
Holy
Who tried kicking him
Then Elon Musk Would Take Over OpenAi
It's much better when I tested it after using flash 2.0
Lmao
If Sam Was Moved to Microsoft
I remember Character AI now
Google got it's employee back
Lol.
I had thought is so good
Microsoft had no ai product
He has xAI

