#general
1 messages Β· Page 368 of 1
I just lost Agent mode? Lol
the meta joke of anthropic
π
Hello!
Welcome!
@rocky geyser hes pineapple approved now
Pineapple did you review my application for champion?
not Elara!
He reviews applications every once a week.
@echo aurora i would be happy to provide feedback on agent mode
damn I hadn't heard of Elara
π
apparently bugged according to twitter
cache got ruined
Yeah so,
next is probably feather, then watermelon
it's like stupider and tokens run out so fast
its being pedantic too
true
oh yea ampro
I js repaid
and renewed
my pro
π
nice
umm
umm
umm
adressme
This aint old
Real og knows about the first website
I forgot the name but it was like some sort of βresearch labβ
It had the github repository functionality ππ
Do you remember the old website name?
Basically what I got some days ago lol
lmarena?
Older
idk i used this back when it was created, but i dont remember much
it was a leaderboard
and a simple chat
back in 2024 ish
when first models dropped
There was a website where there was chat, image generation, github repo and leader boards
i used arena as free gpt demo
Then they added web lm arena website
nah image generation was added much later
first it was only text + leaderboard
Then they merged everything into lmarena
only battle mode
there wasnt even multimodal models back when arena was created
back when bard
why is there claude opus 4.7 in da leaderboard but we can't use it ?
Its crazy how many years has passed
gemini 3.5 flash on lmarena
π€«New Gemini 3.5 Flash Checkpoint?
An update has appeared on https://t.co/4ASKj19CPe under the technical name gemini-3.0-flash
Despite the naming, the initial test results look anomalous. The model is outperforming the current Gemini 3.1 Pro and even Claude 4.7 Opus.
Deepmind?
hmmmmmmmmm
HMMM
There were
It was chat.lmsys.org the domain
This was the real og
Crazy that a research lab turned a smoll project into a billion dollar company
e
Same
Ik it too
It was not named "research lab"
it was lmsys
π
the lmsys team (arena team) were the researchers
aka aangelo poulos
(Anastasios Angelopoulos)
he co-founded the company alongside UC Berkeley professor Ion Stoica and CTO Wei-Lin Chiang
so they were the research lab
7-day sales update π
This week came with a lot of testing, adjustments, and long nights, but the results made it worth it.
Ecommerce isnβt luck, itβs strategy, consistency, and understanding what customers actually want.
Appreciate everyone supporting the journey π€
you're talking about
<@&1349916362595635286>
@echo aurora A quick question, does AA talk here anymore?
and WL
Scaaam
hi Mr @wooden mulch congrats on the success of arena (lmsys) and im a big fan of your project
I used arena since it was created (literally day one, i remember the announcement), but i wasnt in this discord server π
same
I was in the old server tho
but diff account at that time
π
rip account (deleted)
rip
I remember
his account was deleted
Cuz he had a fight
with someone π
and got mass reported hard
dayum
lol
oof
π π
thats why this account was created in April 3 2026
not suspicious at all
this is my new account
hahaahahhaheshaaheahhaehdsjsiwjkesk
just ran out of servers
I have so many other accounts
200/200
old ones
π
u have 2017
?
Or 2020 atleast
if u do give me
or atleast 2022 π π
and I sponsor prohub
EVERYWHERE
discord? 2020 is my main
I had 2012 account
sold it for 300$ last year
it had good amount of robux too
soo
by good I mean REALLY good
Ye same
New account new life
or I think I sold it for thousand
We truly having the same experiences
yeah I sold the account for atleast 2000$
roblox account btw
it had like 300k robux with expensive stuff
lol
and old
truly tuff
I am
Since the beggining in 2024
But didnt join discord
back when tix
Not very often, they're super super busy
oh
mr wl last message here
AA last message here
they are kind
PINEAPPLE MISSILE INCOMING
@echo aurora
That one wasn't me. @quasi atlas gets that credit
ahhahah that's fine I also like kfc
Bro finally replied!!
Im crine
hello i am here to generate videos
Welcome! Would note the server bot that'd generate videos has been removed from the server. You can find Video Arena on our site here: https://arena.ai/video.
email signup worked
after you get the email, you must add arena.ai/ before nextjs-callback and 999 after signup_intent_id=
What is Boba by stealth
Was it not working for you?
It wasn't working for me when clicking on that verify button, but after I tweaked some stuff it worked
Were you running into some kind of error message?
modmail is back π
Yeah but likely going to change it for something else, it's not working properly atm.
Yo
Think we need a new bot

a watermelon and pineapple having a conversation... what a sight to behold XD
Hm
There's a banana.
π€
We have a hulkβ
We have a banana β
nice idea
I js need good hosting π
Banana mentioned???
Huggingface wifi so bad
Check the ui fr
π
im going to eat a pinaple
oh wait
ive a form to do ToT
It is in peak season
how can i do videos?
You can try the Video Arena in https://arena.ai/video. More information on how to use Video Arena can be found in this article.
Oh i will try. Thanks
They expensive bruh
It costs like 20$/mo for decent vps
OCR eats a notable amount of compute
Less than an llm but still
One video per day? OMG
I need like 2 vcpu
1-2gb ram
500mb disk is fine
Only if low usage
Only if noone requests it
If a paid user requests they shut you down
Lol
@echo aurora did they actually lower it down to one video per day from 2?
this just gets stranger and stranger π
No it should still be 2
You're only seeing one?
Didn't expect Gemini to be up there.
i've fired 3 app ideas at Agent Mode
a mood tracker
a budgeting app
and a mindmeister self hosted alternative
some promising results
Can u explain this like im 5
nvm im looking at the website on my computer
much more info
these are previously reported cves
and they see how well models can find / exploit them
right?
My issue is figuring out how to monetize creations...
π¨ Gemini 3.5 Flash on @arena β Minecraft in a single prompt.
οΈοΈ
οΈοΈBest implementation of minecraft ever seen by any ai model till date for one shot
οΈοΈ
οΈοΈWhat actually surprised me:
οΈοΈ>zero external assets (no .png textures, no models)
οΈοΈ>working block place + break with hit feedback
οΈοΈ>procedural textures drawn in code, not loaded
οΈοΈ
οΈοΈArena link in reply to verify by yourself
Quoting Chetaslua (@chetaslua)
οΈ
π¨ Gemini 3.5 Flash checkpoint notes:
οΈοΈ
οΈοΈThe old βcool Gemini UIβ vibe feels mostly gone. It is more GPTified now.
οΈοΈ
οΈοΈGood: fast, smart for a Flash model.
οΈοΈ
οΈοΈBad: it keeps filling websites with HUDs/panels, even when the prompt asks for restraint.
β¦
thats crazy bruh
very fun cursor
W ragebait
how do i know if its gemini 3.5?
Maybe Opus but I voted Gemini
Gemini has always been to lazy
Maybe they fixed it but idk...
everytime gemini releases a new model is crazy good tho
ive noticed it gets lazier over time but on release day those things are beasts
There is a reason Gemini 3.1 is never mentioned in coding
Gemini 3.5? the last model I saw was gemini 3.1 flash lite
It's benchmaxxed. Everyone knows that benchmarks dont represent model quality on software engineering.
whatever you say
Most people I come across keep telling me that Claude is better than gpt in actual use but gpt wins by pricing.
Wierd because I get the exact opposite, but from personnel experience both of them are equal somehwhat.
And yeah, codex has 1000x better quota.
Since 5.5 ?
Yes
And idk how 4.7 made it so high in benchmarks. I see a lot of rants on twitter. More than gpt.
Right?
Is Claude ai 4.7 will adds ?
what do you mean
claude opus 4.7 is one of the best
Hmmm π€ it already been remove from arena
it's very expensive so the smart people at Arena are looking for a more sustainable way to give it to users
im crine how did it find the user
wait why did the rate limits come back
or they were never gone
Idk after I started using after the limit update I didn't get any
Well, it's clearly Gemini and this is because of the versioning. GPT 5.6 and Opus 4.8 might be better than Gemini, but 4.6 and 5.5 probably are not >Gemini 3.5.
3.5 is going to be the best for a while after release then others will bechmaxx and pass it till next google release as always
Just like 3.1 Pro
Then google will release flash
which will be better than 3.1 pro
and flash-lite which will be as good as 3.0 or 3.1 pro
That's usually how it goes
Currently 3.1 flash-lite is as good as 2.5 pro
I know this will be a classical one but.
Yup
Grok has dropped out a bit
lately
xAI*
But we got Chinese labs
GLM and Kimi
Deepseek
And Qwen.
They're especially good if you are into local stuff
Probably they will add Sonnet 4.7/8 when it comes out, yeah. But whether Opus 4.7 will be added is really unknown at the current point.
does lmarena agent mode have preview for site requests
Agent mode good
Guys I had to ask how can you bypass the third party content filter. I cannot create Batman from LEGO.
what kinda pc did you have to run 675B parameters bro 
Hii
guys any thoughts about Ernie 5.1?
Nope
how to create videos here
BTW why is my agent mode not working, it has stopped replying after 4 replies
is there a hard limitations everyday for this mode?
How to prompt qwen-image-2.0-pro-2026-04-22 to generate portrait images? Adding 9:16 aspect ratio to the prompt still generates images in landscape orientation.
Havent used it at all sorry
Upload a blank image with the desired AR
Welcome
why does my gemini 3.1 pro chat have a longcat icon
new agent logo
<@&1349916362595635286>
Look guys running locally
122b parameter models...
You need atleast 80gb gpu
Thats bare minimum
anyone know a site for website designing prompts
I recomend you use gpt-oss 120b on nvidia.build
its free and fast
nah
nice
@hardy swallow Thanks for your reaction, you are a cool moderator.
Mistral is good there too
I use glm 5.1 from ther
I have free glm 5.1 api key π and its super fast
i want one too
since glm 5.1 normally overthinks everything
in a public area
<@&1349916362595635286>
I cant read
thanks
**β
Treat others with Respect. Be kind, assume good intent from others, and keep disagreements respectful. Itβs encouraged to share your disagreements, but only if itβs done in a respectful and productive way.
**
It wasnt.
It wasnt. Trust the mods
*He/She is just ai expert
Indeed
Yea totally
Finally I got agent mode
Yayy congrats!
Thanks!
My custom appπ€
congrats!
Whats that?
My ai app that allows me to chat with models (You can see all models in the screenshot)
Browser based!
So its a site
Cool. U using your own api keys?
What is the name of the app or website?
No, but I did try using paid api's/self hosted and theyre expensive or slow
But now I found a thing on github thats free and fast
Could you elaborate?
Cant tell π° because of rules and its not public yet
For example like what kind of github "thing" you are using
Okay so:
I found a github repo that has free api and endpoint for ai
Its super fast and good
Will that count as promoting?? Mods
Search in github for "Free ai api"
Dont think so
You can send it to me in dms if you want
Okay ill try
didnt worked
searched 30+ repos
nice
this, right?
Lemme send the url I used
Count as promoting
Not an promotion because I didnt put the repo name/link
thx
np
it opened finally
Still advertising itβs like saying hey, I have this this is that would you like to see but I canβt tell you the name but I do have, but is what it is though keep it up nice web site
Theres like 3 hidden gems in that topic
oh
But Iβm not here to judge π§ββοΈ anyone. Isnβt my job. Just be-careful donβt be ban for that lol.
Also nvidia.build is good for small projects but its slow
You need to run it on your local hardware itβs way better
Thanks :> for warning
I dont have gpu in my private server at home π’
Sounds like 1 token/hour
Depends on the gpu
?
just to reply a simple, 'yo'
it took ~3 mins
Iβm running off of one of my laptops, AMD and I have one of my own programs, and he uses my GPU CPU V Ram and ram
Its 27b model π’
I have my own running on my private server and I have 24 models some of them filter. Some of them unfiltered some of the restricted some of unrestricted work perfectly fine.
well yeh
And itβs only on a laptop. Iβll get you the specs in a few. Iβm not on the computer at the moment doing some errands.
I have xeon 3rd gen and 24gb ddr4
i expected ~20 tokens/sec
I have laptop as my main device π’
Same
me too
GTX 1060 (Laptop edition),Core i7 gen 7, 16gb ddr4 2400mhz
oh
Wow
i have LOQ
4 Core cpu btw
2.8 ghz
My server is almost better than my main device π’
i could have bought a legion 5i
they coasted me ~same
but 5i was giving me 5050 and loq had 5060 and 24gb ram
5i varients were 16gb
price?
in $?
Maybe I will try to earn money to buy new laptop or pc
hp laptop victus I just forgot the specs I know itβs a AMD and it has two GPU one built-in and has 32 gigs of ram. Iβm pretty sure the cores are 6 to 12.
RTX 5050,R3 3600/R5 5500
~1600$
WHY SO EXPENSIVE???
if u can find ddr4 for pc
You could buy 5070 desktop + ddr5 and 7800x3d
Some parts will be used
currency exchange
But the better hardware GPU ram youβre gonna get better results and depending on the program that youβre using if youβre using your own custom tool, you need to send prompts or have a brain system for your AI
why this captcha keeps going to next, im solving it for the past 15 mins
thats 100x better than new, ram r good in old market
I agree. I happens to me a lot too and Iβm over here clicking.
u r bot, be happy
150$ for 16gb 5600mhz (1x16gb)
ok
damn
Used*
is there any extension which can fix it
thats a 3050
try reloding the pge
yeh, but it has atleast vram
even 6gb vram is something god tier in 2026 for budget builds
I have a 3070 ti but it sucks so much that it only has 8gb of vram
Oh, and a heads up if youβre running your own and your server, make sure you donβt have nothing else running in the background like web browser or virtual studio code or any program I eat a ram you just want to keep it just that program running youβll get v better results and depending on the model if you donβt have your program to train it to respond correctly and learn, theyβre gonna be all over the place sometime you can say hey, and Iβll give you a long paragraph
try finding old ram, and gpu too, if u get 4060ti 16gb one
5050 new or 3070 ti used? or 2080 ti used or 5060 used 8gb
My budget in under 900$ :<
yup
and 5060 without dlss and 4x frame gen is ~ same to 3070 ti in games
Yep
Who here has created their own program yet
If I donβt get to you you guys give me a second I got things to do you guys did DMM if you like if not, Iβll try to come back here and check it out
Ai app? if yes then me
grab some old 26gb stick <100$
a 5060 if possible will be ~400$ now
some 5600 series, x/g or non g is great too
buy some 450m motherboard, dont spend much on motherboard, ram is more necessary
An app they use for your phone or desktop version a Web version
β‘
Bypass like
US $299.99 = RTX 5060 Used 8gb
US $298.00 = RX 6700 XT Used 12gb
EUR 315.00 = RX 6800 16gb [NON Xt]
hi
Wsp π
Yo banana
6700 xt is the deal
nvidia is better
Good!
yup
Lemme check benchmarks
new vid site is witty
5 Minutes!!!!!
like AMD videocard is so annoying
u need vram for llm
not much
<@&1494389630630625400>
using the FSD even might crash your system
Also I am aiming on 1080p
π
π
6700xt is perfect for it
how strong will gemini3.2pro be?
5060 gives u 6x frame gen, basically 30fps to 180fps
only if u use it for gaming, and 6x will be supported on new games like re9
so, 6700xt is better for llms and raw gaming
noo im back
not better than gpt 5.5
its here?
my predict
oh,
Its gonna be like 3.1 100%
i mean maybe they already changed the gemini 3.1 pro in arena
like they did to 3 flash
6700 xt is worse at 1080p
"I need it for gaming"
maybe its going to be much better
we need to stay united and create a colorful bond
melondaman is back
But the delivery + 5060 gonna cost 350$
what
So Rx 6700 xt is better
?
I'm back!!
bc it have CUDA
kinda yyeh
How are you guys liking limited use
Okay so
RX 6700 xt
Mini atx motherboard/Micro mini atx case
@echo aurora 4 recovered, 2 more to go
it has but 4gb extra vram beats the cuda
Pineapple hi!
no
yup
Cpu?
CUDA beats AMD
cuda is great for optimization
Yes but not on value per dollar
5600
Lemme check prices
yup
e
US $95.80 = R5 5500
US $133.00 = R5 5600
US $139.90 = R5 5600x
So 5600x is better for almost same price
Now the hardest part to find
earlier ppl got 5600 cuz 5600x was 20$ more
storage.
MSI B550M PRO-VDH WIFI
99$
thats for extra random files
oh yeh
thats it
But I dont know the brand of nvme
its gen 4.0 nvme
So lets say 256gb nvme new
my laptop is from 11 years ago
no comment
wd/ kingston/ some other too
thats the hardest part
just dont buy samsung nvme
I didnt know it existed
Corsair
3200hmz is default
check if 5600x supports 3600 or not
Dual channel
how much
it does
$
lemme check
and motherboard
it will be scarry
HELL NAW
US $249.99 = Kingston FURY 32GB (2x16GB) RAM DDR4 3200MHz KF432C16BB12AK2/32
USED!!!!
Lemme check more
Nvm I looked not for the cheapest
oh
Corsair Vengeance RGB Pro 32GB (2x16GB) RAM DDR4 3600MHz CMW32GX4M2D3600C18
- US $159.99
Phew
use the amd one
nah, see the 6700xt first
MSI MAG A750GL PCIE5 Black 750W Fully Modular 80+ Gold Certified
- $79.99
its 250w i guess
lemme check
I buy used cpu
thats a gold
oh,
TDP
230 W
then grab any for <$50
ARCTIC Freezer 36 (Black) CPU Cooler - $30.54 (Air)
or
Thermalright Peerless Assassin 120 SE Air ARGB 66.17 CFM Black - $36.90 (Air)
or
Thermalright Aqua Elite V3 Water 360mm Black - $54.90 (AIO)
20$ more isnt worth it for the aio'
5600x runs cool
anything works, its 5600x
only if u dont live in arizona
Thermalright Peerless Assassin 120 SE Air ARGB 66.17 CFM Black
Is the perfect choice
alr
lemme add it to my baskeet
agent mode straight up gives Something went wrong. Please try again.
after 5s
u have 100 for case, thermal paste, rgb kid stuff
I need a case
mine works fine, report it #ask-here
micro mini atx
grad anything
Thermaltake View 170 ARGB MicroATX Mini Tower Black Tempered Glass Side Panel
Fish tank
oki
u wont be taming a shark anyways
be sure to leave 100$ more
Oh and I need monitor
cuz ram will be gone next hr
wow
fr
90% chances in 2026
160$ for 32gb is basically a gold nugget on road
- AOC Gaming 24G42E 24-Inch Full HD Monitor, 180 Hz, 1 ms, FreeSync Premium, G-Sync Compatible, HDR10 (1920x1080, 1x HDMI 2.0, 1x DisplayPort 1.4), Black
wow
take a loan for 1000$
and buy a 10million $ nuclear reactor
24'' is perfect
for 1080p yes :>
yup
27 inch >>>> for everything
well yeh
maybe sata ssd?
10$ less than 512gb nvme
u need a reactor first
why now
Hello !
maybe 650W psu downgrade?
sup'
yo
works fine
use pc builder app
on phone
its better for psu calculation
MSI MAG A650GL Black 650W Fully Modular 80+ Gold Certified ATX
perfect
or get a 750w bronze+
Nah
lemme ask ai to build a gaming pc
hmmm
it suggested me a h610m for am4
π
Maybe 5050?
lemme check benchmarks
and price
or 4060/3070 ti/non ti
I have 6gb vram.
4060ti 16gb
oh
RTX 5050 + r5 5500 benchmarks 1080p
and ~10% difference in raw performance
Says 60 fps stable in cyberpunk on ultra
decent
very easy
lemme check more games
70 fps in almost all games on ultra
?
dlss 4.5 is here, dlss5 is coming by 2026 end
4x frame gen is already here, 6x is coming by end os the yr
Dlss5 is not confirmed on all 50- series cards
So 5050 is perfect because of frame gen
yeh, but 4.5 is
yup
lemme remake the build
k
And 5050 is better than b580
keep the motherboard and ram same
not in vram
yup
but b580 got 12gb vram
the same deal with 6700xt
US $279.99
- Delivery US $35.44
used?
ye
doesnt matters tbh
MSI GeForce RTX 5060 SHADOW 2X OC 8GB GDDR7 Graphics Card MS-V537 = US $299.99 + US $36.09
been ~1yr since 50 series dropped
hmm, $20 more
go with 5060 if u want
it has more chances for dlss 5
nah bud
30 series is too old
u will regret
cuz devs now depend on ur gpu's AI
RTX 40-
so u might get cooked fps on latest series
US $300 or less for delivery and the gpu
if u get at least 60$ less than 5050 on 4060
MSI GeForce RTX 4060 8GB GDDR6 VENTUS 2X BLACK OC VIDEO CARD
βNo Retail Box, Card Only, No Cables Included, Grade A condition, 1 Year All-State Warrantyβ
Ebay
then ebay is better
try some gpu stores
like
or nah
they r expensive
go with ebay
I asked my ai to make a adult network logo and it did good
nice
US $250 for 3070
US $300 = GIGABYTE GeForce RTX 3070 Ti GAMING OC 8GB GDDR6X Graphics Card
lemme check benchmarks
vs 5050
4060?
it will use more resources for less results
maybe
4060 is WORSE than 5050
So 5060? cus the price is almost the same
dont search gpu
And rx 6700xt is good but no frame gen
yup
yup
then 5060
or if u want to search more
try 7600XT
it will give u vram
C $349.99
+
C $67.00 (Not funny)
XFX Speedster SWFT 210 RX 7600 XT 16GB GDDR6 - TESTED
WOW
hmm,
5060 is better
254,44 + 48,71
like
300$
u will have 32gb ram
30-50$ less
dont worry bout vram
for 7600xt
oh
Cheaper for better?
it supports frame gen
its perfect for <$1000 builds on pc, and 1500$ laptops, tho it coasted me ~1600+ cuz of currency exchange
i cant sign up for oracle to host the bot π
same
me too
because some pro developer decided its good idea to not accept unicode in the city of residence
π€£
btw @sleek urchin
π
87 fps avg on cyberpunk 2077 ultra 1080p
Ryzen 5 5500 + RX 7600 XT 16GB
and it can run gta 6 cuz of 16gb vram
yay :>
yup
if u have some more money, buy a 5600
but 5500 is great too
Money?
5500 wont run gta5
Hii
get 9950x3d
hey
how do u see my wallet
9950x3d is cheaper than 16gb of ddr5 lol
twin β₯οΈ
we dont own Nvidia yet
yup
Nah why not 9950x3d2
i have 24gb ddr5 5600mhz
cuz that costs 2x more
just wow
my laptop, ddr5 4800 mhz
well, laptop coasted me $1600+ just for nothing
The built got more expensive I think
prob
$1208,02 WHAT???
thats actually kinda affordable
nah
Pineaapple did you review my application for champion?
good ram mobo cpu
π
256gb is decent for roblox
256 gb of what?
?
ram
for roblox you need a beast of cpu lol
I though vram
ram is fine at 32gb
i dont play that brainrot for now
yeh, 256gb will cost more than a 5080
I never played it
me too
i will do on that websides 10 ais fight eachother in mafia
yeah but i got mine for free for school from school
read the reviews of this product
plz
slow as hell
Nvidia B300 is perfect for roblox
yup
gb300
just get 4x rtx 6000 blackwell pro
Its worse
yup
B300 is the most expensive the best gpu for ai in the world
it will run my mario on 10million fps only
Brev.Dev Instances Console
$86.47/hr total
yeh
nah low single core
if u sell 192gb ram
9950x3d mogs it
u can play there for a whole day or 2
Maybe
for ai get rtx6k pro
chech 6k pro
yup
a sec
its 5090 with more vram lol
here
Budget gaming pc: https://www.buildcores.com/builds/Qt0P2FXrx
Only 200k$
we can buy this, if i pay $1 and u pay rest
we can use it 50/50
Pretty cheap
hi
Hii
sup
how y'all doing today
Good^^ hbu?
yup,
i will need just few 10million $ nuclear reactors to run it
im good twinβ€οΈβπ©Ή
somehow alive
Yea frr
gotta keep pushing fr
wow
wont fit
Amazingly good thank you for askin Beanbag - considering I had surgery yesterday. I'm in unexpectantly great shape.
oh
Is anyone here using chatllm from abacus ai?
just buy this
ima go eat see ya
No worries at all - surgeon said things went great. And I can feel it did. Not even on painkillers right now - but I prolly will need some later.
Did you get knee surgery?
Why did you have to get surgery..?
A small cancer, beneign, no biggie.
thats great
I hope you feel well!