#general
1 messages · Page 249 of 1

Hello @tired dirge ! Please check how-to-video-bot to learn how to generate videos.
Cool, deleted my message? Then where should I look for people to do this? Besides the AI server?
Same was when it was spotted in antrophic APIs and didnt release
Also Releasin just Opus and not Sonnet would be lame of them
It kills me that ai doesn’t understand basic stuff
As an alternative, you can go to the correct channel here https://discord.com/channels/1340554757349179412/1344733249628541099, post your creation, and ask for feedback. Please do not redirect users to other platforms to view your creation. Thank you.
Is it error again!
Sent, thanks for help
Gato13?
Where r u from?
Hello community, greetings and blessings.
My question is, what is the monthly subscription price for this platform? And can it be used to create faceless videos for my YouTube channel?
Hello
No there isn't price
hey , can someone help how can i make proper motion control video
It's free, but limited because their intentions are about try AIs to see which one is better
@echo aurora pls address quickly and tell the company's technicians to solve the problem! Still having major issues after the rebrand is extremely embarassing!
Guys how good is the new seed
thread with my best Gemini 3 GA results
https://fixupx.com/DavidSZD1/status/2019442686600610027?s=20
Gemini 3 pro GA is available on LM Arena
︀︀
︀︀here thread with multiple one shot result
︀︀
︀︀1) OS
Its this model, without google logo
is Nano Banana not working anymore? I get error message for two days already, "Something went wrong with this response. Please try again,...
Eh
It is
Why doesn't the site open at all?
all other models work fine tho..
hello
Help
Hi everyone. I need to know of an AI tool for developing websites, specifically for creating portfolios, and it needs to be completely free. Any suggestions? Thanks.
Guys I have a suggestion, you know how sometimes the ai pauses at a line for TOO long and not sends anything? They should add a "Stop Generation" just a button with a rounded cube in the middle, it could be useful, because if the ai is taking too long paused at a line and we refresh the game just screws up, so this would be very useful
@echo aurora [says this may be a bug](#codename-discussion message)
Fake
Do you see my result.
Its another model
Just need one test to see
That not 3 preview
what if i told you this isnt GA
if you went into AI studio's build tab
and asked for the same
it would be identical
is there a site like Arena where u can test Genie 3 for free?
no genie 3 doesnt have an API
and wont have
any ultra product never got an API, deepthink,genie3
ic. Thx.
Here your ai studio build
Hi everyone
I assume this is for a Something went wrong error with gemini-3-pro-image models?
This is an ongoing issue that the team has been looking into. The error rates late yesterday started to get much better.
Hi
What's good bro
opus is out
Ong
sonnet 5 delayed
Arena release it so the poorons can use it
@quartz light yo opus 4.6 is out
ratelimit prob like 5 or 2 messages
but you gotta keep the money stable
Better than nothing
waiting for twitter to start posting opus 4.6 things
waiting for arena
Ok
NAH
I’m testing it not thinking rn
WHAT I SAID?
What the
Of course it's not big upgrade
its not even a upgrade
its only better for terminal coding and not agentic
Cool
opus 4.6 dog svg
exactly
I'm still waiting for sonnet 5
Maybe it got nerfed?
So it means they will have to nerf sonnet 5?
Or make it Paid
Because it will be strange if it's better than 4.6
still no actual coding examples
The heck nerf means
Why would they do it bruh
gemini 3 and its insistence we're still in 2024… somedays I wish that were true.
It’s 2024
What do you mean?
🥲
the textarena model has no grounding
its bound to knowledge
I know
@honest verge @zealous sparrow
I mean you can enable search but that disables image reading
thats where i got the bench img from
Then wth is 20260203 date?
Sonnet 5 doesn't exists?
it was fake
Wow
But if sonnet 5 is not fennec
GPT 5.2 beat opus 4.5 at knowledge work the heck
Then it's not even ready yet?
Bro everything on internet is fake don't believe anything
alright theres an example of a growing tree that opus 4.6 did good
Maybe they folded under the pressure of a new model releasing soon or something
But gpt 5.3 comes out soon
Again GPT 5.2 beat opus 4.5
not today for sure
Again on agentic coding
they made opus 4.5 worse in the bench on purpose trust
But maybe next week
I think they saw GPT 5.2 beating their model and decided to release opus 4.6 😭
CLAUDE OPUS 4.6 vs OPUS 4.5 - Growing Tree benchmark
4.6 is on the left and 4.5 on the right.
We can see 4.6 is much better than 4.5
Prompt: "Create an interactive animation showing a seed growing into a full tree. The animation should show: seed sprouting, roots forming, stem
you only get 4.6 video here
4.5 was slightly worse
Still doesn't look like fennec tests it was better
5.2 was a benchmax
It's so boring the video just generating but never appear
What does that mean
Finally 1M token support
Maybe not fully benchmaxxed but it's slightly is
coding was a huge benchmax
What does benchmax mean
Is Claude 4.6 out ??
is the gemini 3 pro stealth model still on arena
5.3 codex is out apparently
GPT 5.3?
Dang
im unsure
Claude is basically the god for code this shi gotta be fake or sm
I mean they saw GPT 5.2 beating opus 4.5 then decided to release opus 4.6
Unless it's not fake and Claude is falling off on one of the best things it's good at
FR?
He said not sure
Im not getting the gemini 3 pro stealth model, did they delete it lol
Sam got scared lol
some twitter users are saying its in codex CLI
I've tried Claude and Kimi on Arena, with the same prompt. Claude was definitely much more impressive. Thanks for replying.
But I think even Opus 4.5 is better than 5.2 codex
Sonnet 4.5 Extend thinking is free on the app or web.
Kimi qwen glm all thees open weight models are definitely good but they just can't compare to closed models and if the benchmarks are better it means nothing
I think it was free
But it became paid
And now free again
I think This could be a bug because when I enabled the extend thinking and then send a message it disabled extend thinking so I had to re enable it and it works now
The reason I don't really like Claude is the locking out of the frontier models like I know it's flagship so obv not for free users but on Gemini you at least use the pro version with limits and basically unlimited in google ai studio
Gemini is by Google and they are selling your data so yeah
It's not unlimited
But anthropic is different
But it has high limits
Every model does this even gpt
opus 4-6 out in Arena DIRECT CHAT
I don't know about Claude though
Where did you see this?
devmode discord
is owned by legit_api and others
Link to it pls
dms
Yo Link too
Tbh what's the difference between 16k, 32k and 64k thing
Like it had on opus 4.5
64k?
Idk I saw it somewhere
It's the limit of tokens ig
Ohhhhh your right
Opus 4.6 just dropped
Still waiting for sonnet 5
its in direct chat mode now
I don't see it
refresh the page
Guys no abusing it or they will make limits 😭
Claude opus 4.6 park
It's almost the same for 4.5
But it's not 4.6 thinking version
Search opus it's at the end
It can Only dance
Wow insane
Where's thinking version
Maybe in a bit
I see it now
why the end tho do they not want people to see it?? 🥀
It just released ✌️
Make sense ig
Let's not abuse it guys
Idc I'm abusing it
The heck is that
NOOOOOOOOO
Wow
HELL YEAH
openai cooked
It's really today
Anyone can use video gen whoithout any problem?
Ig they rushed it
Difference
IT BEAT OPUS 4.6
XD
It's benchmaxxed
Wth...
What does everyone think about Opus 4.6? I was waiting for Sonnet 5, but Opus 4.6 came out instead, maybe that's even better 😁
Holy
Blud gonna make 3 Gemini work on 1 project
5.3 is not tested yet
When that Happened?
That's literally a 10%+ difference 🫀
Maybe they took Sam's Tweet on Free Access Personally
Just now did they removed it?
@echo aurora okay this is a bug
oh wait literally everyone already reported this lmao
???
Make them work on 1 project
This was flagged, thank you for sharing. 
Is still there
Well hopefully they won't remove it for now
This is like kimi agent swarm but Gemini
Gemini swarm
Probably not
Model name?
Idk lol but I think there is
Gatekeep it bro
I never thought I would see open ai cook Claude in my lifetime
K
Looking into and will report if I'm seeing something
huh
At this point anthropic will release Claude opus 5 tmr
It's reported chill
Buddy put the announcement up that you can use Claude 4.6
ayt
Refresh, should be fixed now.
They still haven't fully released it
Pal I'm waiting for something to be added
Hmm ig it's opus 4.6 thinking
Even if the thinking mode gets benchmark it won't beat. Open ai 5.3
guys i have a theory
are you a dev ? or just a community manager ?
regarding the gemini-3-pro stealth model

its just gemini 3 pro but
GPT 5.3 CODEX?
codearena system rpompt and react
Check his roles
community manager
Either that or Claude opus 4.6 thinking
I shall not confirm, nor deny
Nah that's just a theory
This
it won't be unsecure
gpt 5.3 just released lol
Codex only
We don't care
basically more secure than the webapp for them
OAI sucks
It could happen, our feedback thread on this is here: #1441588701472882759
Bro he is a Community Manager not a dev on arena so don't ask him questions
Hmmmmmm if it took like 40 minutes for when Claude 4.6 got in arena and it's been 20 minutes since chatgpt 5.3 codex
Sooo 20 minutes later
Ask me all the questions.
bro, I think he know the dev team and he can easily talk to them
that's it
It's on arena rn
thank you.
Just cuz he's manager doesn't make him be answered by every higher up
Real
I mean yeah it's on arena rn
Ama abuse the hell out of gpt 5.3 codex
Build Something Good
Make him cry because of the amount of times I will BOTHER HIM
Chill bro
yes but at least they hear him, more than a simple member
The one that released is GPT 5.3 Codex version
Not normal one
So it turns out, this was indeed an issue on Arena's side, and this was not gemini 3 pro GA it was just Gemini 3 pro..
GEMINI 3 PRO IS ANTHROPIC MODEL?
Lol
Actually they didn't know whom they are actually as the api end point have just configure to knowledge to respond. Not configured for self identity
Okay hold on last theory and one i will stick to rn
Gemini 3 pro GA was actually on the arena
but was pulled quickly
Where is gpt
yes but it should be on code arena
code arena has codex
Been waiting ages
is 4.6 up?
🫀
It's insane that oai and anthropic dropped their basic models like sonnet 5 and gpt 5.3 normal and just released opus 4.6 and gpt codex
Probably Trained on Data of it
Still looks kinda weird
Like ai is only programming and agentic now?
That's why Gemini on top
It has only 2 models
Flash and pro
Because that's where the Money is SaaS
@echo auroraare we getting opus 4.6
It's currently there, announcement going out soon. cc @summer shore
yippie!
Kling 3.0 is finally out, and it’s already being called one of the most advanced AI video models available today. In this video, I break down what’s new with Kling 3.0 on Higgsfield, how the workflow works, and why this update is such a big deal for creators, filmmakers, and AI video enthusiasts.
Kling 3.0 introduces a unified multimodal vi...
great
lol
I was doing side by side
hot
it doesnt think tho
Thinking version not there yet
Yes
yes
now we just need gemini 35
Gemini 4
no im pretty sure in stealth system prompt they are trained to lie
also the model i showed there is long gone
GPT 5.3 codex beat opus 4.6
in a parallel universe
what
people only said bad about it
from code examples
Where is 5.3 codex bruh
Facts
openai
hate
c1rclejerk
why is the above word banned
try type it
without the 1
clrclejerk
please tell devs:
"A new max effort level provides the absolute highest capability on Opus 4.6"
PLEASE
I was talking to lamyk but ik OpenAi just sucks at this point
Codex 5.2 didn't existed in the arena for a long time
it doesnt
just popular to hate on
Because ai is supper controlled
That’s why most stuff made by AIS mediocre
In order to make something nice you need to take risks and you need to explore categories and areas that aren’t so heavily regulated
btw openAi flopped again
opus 4.6 no think cooked
where are you from hare
lookstras h
oneshot webapp style stuff is just so meaningless
Fr
why do people still care
Thinking is on yupp ai
i guess its easy to see but like it is so useless
as any form or indication
of it being actually useful
Tf is yupp ai
it thought so much that it broke the website
it gets so laggy and then you cant do anythiing
Just like arena but you need credits to talk to it
Yeah happened to me
holy but dude i got permabanned on all of my accounts just for signing up on an alt with my referral
😭
instantly

i got banned from their discord for talking about openai
js cz i was stuck at 0 credits and couldnt vote
Lmao
I got banned from Gemini discord server because I never said something
.m.
lmfao
They gave me no reason why
Man so much for freedom speech
I got 4,818 yupp credits is it good?
How
I never used it 😭
i have 23k
What the hell 😭
Where these credits
Opus 4.6 thinking takes 350 credit the hell
At this point Arena is better
Always
Oh yeah 170 credit
So is it like each time I send message?
yes
5k?
I mean it is new
Holden let me do math
eh
i dont even listen to that anymore
gpt is benchamaxxed to the brim now
Its on all models
28 responses
theyll say "we beat every model!!!!!!!" and then you ask it to make a simple mc clone in html and it dies
Yeah Ig arena on top
i have a lot of credits lol
Arena just has a rate limit per hour wayyy better then credits
used to use yupp so much when nano banna was available
I really don't know
so i farmed up to 25k points
What I did
I need Gpt 5.3 - non-codex
You have 135 responses
my theory right now is that when opus 4.6 came out they rushed out the new model
all they had was 5.3 codex
they didnt even know what to say 😭
"you can just build things."
openai is so cooked
its literally 1% better on swe bench pro
albeit at far fewer token usage but this was definitely rushed
Hi!
LMAO
finally a model that gave me a decent terrain with one script
Openai is king of rushing atp
Like 5.2 was also rushed
OpenAI will head for banrkuptcy atp
The projection will come true unless they compete with antrophic or gemini
please explain what the hell the difference is
Just build stuff don't think about anything else
Build stuff with codex 5.3 use gpt not Claude loser
🫀
one is called 5.2 and the other is 5.3
true....
Claudes more impressive
bro 5.3 sucks so much
For building and design, Claude's models can be far superior, even to older models
OAI lost subscriptions due to 4o retirement
OAI lost hype due to bad frontend
OAI lost everything
Kling 3 on Arena When?
What
yep it aint in API yet
Opus 4.6 is actually good
is opus 4.6 out?
Not tried 5.3 yet
yes
But 4.6 seems very good
Then how will arena get it
its like
$5/M Input
$25/M Output
expensive but again this is actual SOTA
not lame GPT "sota"
yeah
waits till API releases
what's the new update
they normally get it like at most a day after
Is opus 4.6 good at front end?
no
Idk
I still don't get it is Claude 4.6 better or codex 5.3
wait isn't opus 4.6 relaunched?
🫀
by a LONG shot
claude
codex 5.3 is benchmaxxed to hell and back
im guessing it was the only version of 5.3 stable enough to be launched hastily
Always use Claude for coding, it wins nearly all of the time
Openai rushes their model for every time
Guys any news when the hell bana pro is gonna work
Like it just can't match new models from anthropic and Google
That's not always the case
nano banana pro already works
Gk try doesn't work for me
Im a huge google fan when it comes to models but they really lack coding skill. Except for frontend, they are the best at that (havent tested claude on that yet)
yes gemini is good for vibe coding
best at vibe coding
but also best at destroying real code bases
it will turn clean code into slop
Gemini for Front
Opus for Back
yeah
fr
Great Shots
Ever since gpt 5 I felt like open ai was tired at actually releasing good models and now they just rush
Front and Back
We love those opus 4.6 refactors
they need to for investors
REAL
gpt 5 was a mess but instead of fixing it right away they let it sit for like ages
Codex 5.3 already at arena?
Really excited for the time Claude Opus 4.6 gets an arena score
We are 2
wont be for a while, API is not available
K thnx
Sorry to say I don't have an update to share. The team is aware of this high error rate.
Would note it is a bit better now, but it's still something we're monitoring.
pineapple can you share how long it takes for the new models to get a score
5.3 codex is so good for me
I don't use it for the front end anyways only the backend
show some examples
don't have anything to show yet
it's still thinking
are you using it on codex cli?
Difficult for me to say as it all really depends: the modality, the amount of votes we're getting
so um
insane revelation
crazy
Where you get this
we dont know if its true but actual 5.3 results arent good too
Damn I didn't expect to get a new opus today
Chatgpt beats it on terminal bench but nothing else
oof, opus 4.6 has been struggling for me, keeps getting the formatting on math equations messed up -- i might wait for the thinking version
wait for arena to give us thinking if we ever get it
It's already in yupp.ai
Arena should have it soon
https://x.com/chetaslua/status/2019453031478182394
Apparently Gemini 3 Pro GA in arena?
Gemini 3 Pro GA available in Battle Mode in @arena
Opus 4.6 vs Gemini 3 Pro GA vs GPT-5.3 ( my first youtube video will be this - https://t.co/YMfvKtKDyl )
Follow our team for back to back updates :
@synthwavedd @DavidSZD1
@marmaduke091
Here is one shot demo :

i don't know how to use the site can someone pls help me out ?
I'm going to followup in a different channel.
what did you mean?
Just use it
I pinged you here: #1469063054963638436 message. Would like to keep #general chat for general chat and #ask-here for questions.
Do u know what is that SVG mode? Idk
it was
then got removed 15-30 mins later
what even is GA, this is the first I'm hearing of it
general availability
What does that mean
Sonnet 4.6 when
Gpt 5.3 dropping soon this month i think
Waiting for Gemini 3 flash lite
And Gemini 3 flash image
opus 4.6 thinking is live on arena
thank you
💀
same lol
gg 😭
Nvm
Team is aware of this, we're lookg into cc @frosty shuttle
good luck at fixing
Free opus 4.6 for next 48 hours
made with opus 4.6
Wish they compared to Kimi, but luckily Kimi published quite extensive benchmarks too
apparently its back
Let them take time they are trying to make sure it fully work
How is the model? Is it like "wow-level" good?
Yes but we heared GPT 5.3 codex is better than it
5.3 came out when???
Only codex version
Also I Don't trust OAI anymore after how disappointing 5.2 is
pls add claude opus 4.6
HOLY CRAP\
It has been added, announcement should be going out soon.
WIAT ITS IN THERE?????
Yup 🙂
yeah and it works now
Only in one
And that's terminal bench
Claude Opus 4.6 doesnt work without thinking
Claude thinking works
kjhjjkhkjhdfghjkfdhgdfhghdfkjhgkdjfhgjdfhgkjdhgjkdhgkjdfhgkjdf
It’s probably been a year since I last tried Claude lol
I think I’m gonna keep that streak going
your team is so impressive for bringing these new models out that quickly
Shitt
Wait a minute
W claudius thinking
I wanna try Claude
holy
The new one
is claude opus 4.6 just a renamed sonnet 5 so Anthropic can keep high api price with cheaper model ?
@meager harbor i dont think so XD
I love you pineapple
This made my day! I'm going to share with the team. ❤️

YEAHHH
yeah
pineapple for president !
Yes way
Way

The team gets the credit!
!!!!!;)
i cant use claude opus 4.6 but i know your team works hard so take your time also thanks for you guys making arena.ai so i dont have to pay for the models becasue im broke
Man we all broke that's why we're here 90% of the time
Yea
lol
100% *
There is no API yet so it might take a while for codex
By the way, the new Opus-Thinking model, is the thinking set to 16K or 32K?
someone tested opus 4.6 yet? if so is it really good?
Is Opus 4.6 any good so far?
Opus 6 thought for about 5 minutes and then gave an error 
why no sonnet 5 then ? while everybody was predicting sonnet 5 to come out this week, it didn't, just opus 4.6 dropped out of nowhere, that is very sus
- Very kind
i cant use claude opus 4.6
Why not?
😩
Something went wrong error?
May be a bit unstable at the moment.
i can use the thininking because it gives me the something went wrong eerror
Hey pineapple, why do none of my websites work on my phone?
Made with Arena
Can someone fix that?
As in the preview links? Can you share some of them?
(if you'd prefer to share in DM that's fine)
it works for me but it dont think
No, I mean like the apps I make in general
It's good some people says gpt 5.3 codex is better in terminal bench
Wait yoo it doesn't think the 4.6 thinking
Or it just doesn't show it??
I told you in DM’s
@echo aurora does lmarena have any influence over API response times or will we be able to tell how fast codex 5.3 is compared to opus 4.6 by doing a comparison side by side?
@echo aurora it works again
for sql i have used gpt 5.2 and sonnet 4.5 and gpt 5.2 was way better. sonnet 4.5 queries didn't even work. and I prompted both ai the exact same way
worst is always gemini 3
I don't think so
Gemini 3.5 is coming soon hopefully
for sql for my use case it was, always hallucinating tables and column name
I mean it's your opinion but I don't think it's the worst
Oh, now it works!
He thought for 5 minutes and ultimately failed.
you mean 3 GA ? apparently more chance it being called 3 ga than 3.5
What is ga?
Same here
It thinks long and probably triggers a time out
Gpt being back on top is funny
General Availability, it still experimental for now
How is opus 4.6 in terms of personality
do you often make sql query ?
Too many people want this Ninja
oh yeah if you say it like that it makes sense
also opus 4.6 got 1 million context windows like sonnet 4.5 while for opus 4.5 it was 200k
even more sus
long context windows for fat models is very expensive
I need opus 6
I need opus 7
Alongside glm 5 kimi 3.0 Gemini 3.5
Gpt 5.3
You know fck the gpt numbering system let's g back to 3.5 turbo
kinda pointless unless you're using it for programming
Why can't they release gpt 6 already
Their numbering system sucks
bro, the whole business of anthropic is centered around devs, they are the one paying lot of money for api usage

The way google handles the Gemini models needs to be studied
Fast
Thinking
Pro
10/10
XD
But I still don't know what thinking is
It's just flash but it can think
can't be worst than openai with 4o and o4

Openai sucks at giving models good name
And they still can't make better text model than 4o lol
Open ai model naming is a fever dream
Like is it really that hard to make a text model which can write better than 4o
one of the best i used was opus 3 for writting, didn't use opus 4.x for this so
To this day I don't understand what the o meant
omni
Omni
Oh
captcha getting annyong bra
why is my generates failing
But still like gpt 4 gpt 4o gpt o4 like their names suck
Just make it like this gpt fast
Normal
And pro
Essay
gpt 4.1 being more recent than 4.5 also was something

they were going backward
Huh
How does that work
🫀
Gpt 4.5 doesn't exist now
can anyone explain why the generates are failing
gpt 4.1 released in april 2025 while 4.5 release in february 2025
I still don't know what was the point of deleting 4.5
Also what's the rate limit for Claude 4.6 thinking
like gpt 4 doesn't exist anymore, what's your point ?
maybe cuz alot of people are using them
cost
ok
i kept getting errors too
Then why didn't they delete 4o
its getting deleted in less than 10 days
But #keep4o
this is the website's limit lol
At least now the free users have access to gpt 5
Because there are no more models after they will be deleted
Gpt 5 mini
Then why the tf are they deleting them
Money
cleaning my as.s
That's why #keep4o movement exist
Nobody wants 4o deleted
It was the one who talked like a bro
Not a robot
Not a clanker
It's almost 2 am
Damn
why do i gotta do a captcha for every prompt 😭
Like 4.5 but smarter
when will opus 4.6 be on the leaderborad?
Actually not a big upgrade
Like it's impressive but not a new level for sure
So is sonnet 5 still coming soon or was that anthropic telling us that they are still working on it
Coming soon
I believe
It will
I think but for now idk
I think coming until February 14
Or late February
But I hope not march
I wouldn't think anyway
Dead internet theory
Waiting for sonnet 5 now
how do you like 4.6?
Ehhh could have been better lowk but good
Because it's rumored sonnet 5 is better than opus 4.5
It's like 2 am
And am not sleepy
Lol, love trying opus 4.6 just to get hit with "Error - something went wrong" after having it think for 8 minutes 💀
3½ hour wait for that single generation, so I try again, now I am told I've used the single attempt for today. 😿
My generations are failing...
Me
It's good ?
For some reason opus 4.6 thinking always fails generations
It isn’t for me
It looks like it's not meant for arena
It probably tries to use tools that only app/website Claude has
Nvm
It works for me
Idk
But like it barely does like under 4 seconds
Sometimes it doesn't even think at all
Same
Like idk how it can think for minutes
Also I think in the app it's called extended thinking and could be different then the normal thinking in arena idk
So it has bugs
is anyoen else getting captcha every promt they put ?
2nd failed attempt
Same
its so annoying bruh
For more advanced prompts is when it'll show the thinking
Like it can't build
Ofc
is the chapcha issue gonna be like this forever?
Opus 4.6 thinking does not function at all
Let me send you a recenent message, I just need to go find it
I mean what cloud best at?
Yo pineapple
This Ui Is waaaaayyy Better than before bro
And code and personal writing ig
No entiendo inglés
Alr
This message has some more information - #1451574502369656842 message We are looking into the captcha system. The frusteration it's been causing the community is felt.
I think it’s something more than that
oh thanks! i thought it was permanent or something . preciate the clarity
Woot
What part specifically?
Fix opus 4.6 thinking
I like y'all actually listen to bugs feedback that we say
Amazing team tbh.

It's a helpful tool for preventing bad actors, but we've got some changes to make at detecting authentic users better.
It does this every time
The Ask Anything... Color And Choosing the Model
It's going to be a bit unstable atm. I'll keep an eye on error rates and escalate to the team.
Its looks like Claude
Very kind of you, I appreciate that a lot 
We've got a lot more we want to build
I’m not able to use it; this error keeps appearing a lot. I’ve already tried switching browsers and nothing works, and I’m also logged in with my email.
This is happening with every AI I’m using.
Yes, this is what I was gonna talk about earlier
Cookie reset?
When will opus 4.6 be on the leaderboard mkay?
a very thoughtful response
How did you even get it to work without crashing
was it a simple prompt?
woah thats a wall of text sorry
You cant be clearer than this tbh
Couldn't say for sure, but we will put out an announcement.
All that and it thought for 1 second
if you expand it actually thinks for a very long time
Bro was like I aint reading alat but good idea
enable image upload for opus 4.6
Maybe it crashes when it tries to do coding with opus 4.6 thinking?
Im just praying its gonna be #1 before feb 7 (kalshi bet lololool)
Oh yeah why can't I upload images to it
Or is that normal
same error
Is it working?
how to fix it?
Give it a few days
probably on Monday/Tuesday next week
This is expected. Claude models need uploads enabled, this is something we're working on.
Oh ok
@stray aspen what is the reasoning tokens limit for 4.6
Can you try these steps - https://help.arena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message
Were you using this model before it started to error out?
I think it's less than 32k
Generation still stuck.
And while I'm limited to one single generation per day.
I still see multiple account users generating all over the place. 😿
Opus 4.6 thinks way more longer than 4.5. they changed the architecture.
I've tried 3 times now with 4.6 thinking and it crashes every time after thinking for a few minutes, just seems to be unstable
This image is blocked by arena lol


