general

[proper sun]

Basic supply and demand

[wicked talon]

🙂

[frosty lava]

Lol if there is only few job and elictrician left the whole system will change

it's nonsense to say this

[proper sun]

It wouldn't happen immediately but obviously the wages would tank eventually

Electrician has a safeguard behind a license

[proper sun]

AI isn't the most accurate

[wicked talon]

But it's one of the most accurate

It likely wouldn't be more accurate then someone working in economy but it's likely more accurate then the average human

[frosty lava]

If it replace a big part of jobs then it will already change everything and people will find way to automate the few that's left

cause automation = efficiency

its basic

that's what the whole world focus on

that's why ai exist

[frosty lava]

robot at start cost alot and then with bigger production the price will be lower than human

and improvement

[frosty lava]

and your talking with an ai but it don't have enough context

[frosty lava]

its fully theoritical

[proper sun]

Posting AI slop doesn't help your case

No need to spam it

[wicked talon]

AI slop lol

[frosty lava]

yeah.. but the standardization do increase, and many thing your ai is saying is already happening

yes full nonsense cause its not based on new information

and doesn't research

and doesn't do anything

that help to answer correctly

it have knowledge cutoff

you just don't give it enough context anyway

it answer like we're still in 2024

but many thing changed from that time

especially for what we're talking about

its simply missing many crucial information

[frosty lava]

Yes job will disappear and then new job that require human will appear

and then we will automate those too

and we don't know where it end

[frosty lava]

yeah exactly you never heard of universal income theory yet ?

[toxic verge]

This current ai ain’t taking no one’s job.

[frosty lava]

i was talking to you not to the ai lol

that was a question for you bro

[frosty lava]

who said the current one will

[frosty lava]

but it already help for many thing

[shrewd citrus]

they say that ai won’t take people jobs

but it’s actually the people who know how to use ai would take their jobs

[frosty lava]

History just repeat itself, people see new technologies and are just scared

ai will create new jobs

that's the part many people miss

actual job, disappear

and new job appear

but then we will try to automate those one too

[frosty lava]

and that's why the universal income theory might happen

you know its all about theory right now we can't know

what exactly

[frosty lava]

will happen

[haughty timber]

Hwo to crack veo

[frosty lava]

you are trying to prove a point but ai are evolving and you can't deny it

[wicked talon]

😭

[frosty lava]

and also for some jobs it'll require specialized ai not LLM

[toxic verge]

I’m not denying that if they’re not I’m saying I don’t believe in the promise of tomorrow. I’m living in today.

So until then

[frosty lava]

LLM is large language model

[wicked talon]

I was literally about to say there will be specialized ai

[frosty lava]

but you can do reinforcement learning

[wicked talon]

Current AIs are built for general purpose

[frosty lava]

for basically everything

[velvet furnace]

OMG,I saw GPT 5.5 in the battle model

[wicked talon]

Oo

[toxic verge]

All right well now they get there when we get there

We can discuss it as of now

[frosty lava]

even right know ai are very usefull

for most job most people actually use ai

to be more efficient

[toxic verge]

That’s different from taking people’s jobs lol

[frosty lava]

cause it don't happen instantly

[toxic verge]

That’s what I’m saying

[frosty lava]

but the fact that this is happening

show the potential

[toxic verge]

Derrick Rose had potential where did he end up?

[frosty lava]

that's just pure ragebait and unrelated thing

[toxic verge]

Potential doesn’t mean guarantee

[frosty lava]

who said its guaranteed

its all theory

[toxic verge]

They could go both ways

[frosty lava]

but it's most likely happening yes

for now nothing show it won't happen

its becoming more efficient

smarter

[toxic verge]

Right, but until it does as of now and today, it’s not

[frosty lava]

we can do specialized ai

[short sluice]

fallen in love with making random youtube videos of latin american churches that dont exist with gpt image 2 tbh

[frosty lava]

Yes, today and for a year your job is safe

we don't know about the future

[toxic verge]

So why would we assume that it will if we don’t know?

[frosty lava]

that's theory

and its most likely going this way

that's why

[toxic verge]

OK, and if it doesn’t?

[frosty lava]

then it won't ?? what you want me to answer to that

but nothing show it won't right now

its going to this way till now

and its evolving faster

[toxic verge]

Actually, I’d argue that everything at a showing and it’s showing that it’s incapable of doing a job with the tools we have now lol

The way they’re working as of today and now

[frosty lava]

Yes, now its making people that have a job more efficient fordoing the task

it also automate some right now

[toxic verge]

Which is different from actually taking someone’s job it’s assisting with jobs which is different altogether

And actually doing someone’s job

[frosty lava]

you seems somehow blocked on the idea that we said it's taking job right now

when we talk about potential way it evolve

in near future

[toxic verge]

Whose job is it taken right now?

[frosty lava]

its not the same thing

[frosty lava]

almost no job right now cause that's why we talk about potential way it evolve
in near future

right now it help people

that do the job

[frosty lava]

not doable right now also doesn't mean it won't ever be doable

the fact is that its evolving

very fast

[toxic verge]

Look G

This is from 2024

Now this is 2026

[frosty lava]

Yes ?

[frosty lava]

yes openai literally closed sora but you know what ? its not the only one making text to video model

and you know why they did it ?

to focus on Chatgpt

LLM

[toxic verge]

You completely missed the point

[frosty lava]

probably otherwise i just see that you are trying to prove im wrong with just random information and without the context its need to be in first

[toxic verge]

All I’m saying is I’m not trying to be the next Tyler Perry

You’re free to believe what you want I’m free to believe what I want and it’s not black-and-white for me. It could be both ways. It could be ways. I don’t know not speaking for certainty.

I’m just going off based of what I see now today and currently

I’m not living in tomorrow

[frosty lava]

Exactly but somehow you want to prove that its impossible that it'll go the other way

[toxic verge]

Never once said that

[frosty lava]

you didn't said it but your doing it

[toxic verge]

Show me where.

[frosty lava]

just the whole convo is you saying its impossible cause right now its not taking job

[toxic verge]

I just told you I never said that

[frosty lava]

that's what i understand from what you say

that's what you seem to do

[toxic verge]

Well, I apologize for the misunderstanding.

I could see why what I said could cause confusion.

[frosty lava]

Yes no problem its just that i wanted to make you understand why people say it can take job and probably will, its cause ai is evolving fast and showing many improvement, there is also specialized ai that we can make instead of just using llm (automated car and thing like that) for example but applying it to job and more

[toxic verge]

Yes, and there’s also a lot of research that’s putting all that into doubt

[frosty lava]

its not happening right now, but its going this way or it seems to

[toxic verge]

Well, yeah, I guess those people live in reality of today. Just don’t see the promise for tomorrow.

[frosty lava]

yes right now its not doable with the actual LLM and ai.

that's why its the possible outcome in the future

not right now

[toxic verge]

Yes, and one day AI is gonna cure cancer end world hunger and somehow make us live forever and and solve all of mankind’s problems

[frosty lava]

some of them might happen but definitly not solving all of mankinds problem neither immortality

[toxic verge]

Do you know what the philosopher Stone is?

[frosty lava]

its cause of those answer from you that i actually see that your trying to prove im wrong using nonsense and unrelated things

[frosty lava]

imagine going many years ago if that was possible and saying to people that we will be able to fly with plane

to walk on moon

stupid isn't it

[toxic verge]

Here’s what’s gonna happen

Either one day in the future, as you say when this does occur, I will wake up and realize what a fool was for everything I said

[frosty lava]

If your talking about ai taking jobs then i can bet on it honestly

its the thing that will most likely happen

doesn't mean there will be no jobs left

but the actual one getting replaced yeah

many of them

[toxic verge]

And if it doesn’t, one day, you’ll wake up and just realize how naïve and gullible some of your ideas were.

It’s one of those things and I’ll bet on it also

So until we meet again, my friend

https://tenor.com/view/hats-off-józef-the-hourglass-sanatorium-taking-of-my-hat-greetings-gif-27453787

[frosty lava]

go back to 2021 and say to people that in 2026 we will have ai capable of doing things like its actually capable of

only 5 year ago

imagine it and how they would react

now you better understand i guess

honestly even in 2022 or 2023

[toxic verge]

Good day sir

[frosty lava]

Exactly

i didn't insulted you in any way

good day

[velvet furnace]

is kimi2.6 is good? >glm5.1?

[frosty lava]

I like arguing with people that actually doesn't spread misinformation, and don't use nonsense and unrelated example to defend themselves, we defend our point of view with facts and theory if talking about future, not by giving info or sending links by deliberately not stating important part of the context to prove our point.

its how arguing should be

this one didn't felt like it honestly.

When doing selective omission to prove our point, it's pure manipulation

just a reminder

[tardy geyser]

generate video

[narrow sigil]

I got logged off my account suddenly

[tranquil current]

since lmarena is removed top tier model to direct test. i think at least they need to add a new subscription to access it all but can access every model

[gleaming kettle]

hello what is the best ia no filter for cybersec? (web and local with olama ou ia LM) (sorry abou my bad inglish)

[velvet furnace]

mybe opus4.6

[soft obsidian]

but they havent in direct

only one ai here -

[gleaming kettle]

no have filter

[velvet furnace]

u can ask the same question to the battle model

[gleaming kettle]

no have ia no filter in areana battle mode

[rocky geyser]

What rate limits for images are there right now? Like the new ones start on monday the 27th but like what are they currently?

[stone forum]

Is it me or most deepseek chats never FOLLOW instructions properly

Can someone like literally note that TO THEM

[errant rover]

Any mod help ?

I keep getting this error

[grand raft]

gpt 5.5 is a goat model

[night moat]

After gpt, now Gemini

[safe cloak]

will arena.ai survive long term or it will shut down like yupp?

[steel steeple]

No way i need lm arena😭

[errant rover]

Nothing lasts forever, appreciate it when it's still here 😂

[weary thistle]

@echo aurora i think the team should review the new update of terminating a conversation when token limit of a conversation is reached. I think it takes out any incitement of using arena. Instead of terminating completely the conversation, maybe you can at least block the conversation for a period of time. Or, if you a implement a number of messages or tokens per day per user, the user can spend these tokens the way he wants. And the next day he'll continue from where he stopped. But completely terminating a conversation is a really bad idea in my opinion. The whole arena goal will be unreliable . It will affect the quality of the leaderboard. People will only be using the site for "small unimportant" conversations.

[rocky geyser]

Wasnt the original goal of arena to test models and rank them not to build large applications?

[weary thistle]

Maybe. Why?

[safe cloak]

if arena ai shuts down,goodbye to access of all models forever

[rocky geyser]

Because like i dont think it was the point of aippy to build really large projects and it would be really expensive for them to do that

[rocky geyser]

I believe it wont, ai companys kind of rely on arena.ai

[safe cloak]

i hope it stays forever

[rocky geyser]

forever is a big word but it will definitly stay for a long time unless some big changes happen in the ai world

[weary thistle]

One the goals of arena is, in addition to ranking overall models, is ranking models by categories (legal, construction, coding, etc). I don't know much about coding, but in legal and construction, i rarely give a vote from the first answer. I need to have a "long" conversation, before i decide which model is better.
That's why I'm saying that, in my opinion, it will affect the leaderboard. Results will be kinda rigged.

[tame mason]

Guys is opus never coming back?

[rocky geyser]

That could be too but the longer the conversation especially in coding the larger the context gets and ais have a context window which isnt that large often and even if it is the price is really expensive for example claude opus is 25$ per million output tokens and arena would have to pay for that. It has many active users and no paid subscription they would have to raise much money, they even had to remove opus from direct chat because of that.

And have to decrease image generation limits

[weary thistle]

On twitter, The big companies like ChatGPT for example, show Arena leaderboard results to "brag" about their models.
But if the leaderboard ranking start getting less reliable, it will affect them.

[rocky geyser]

Nope the larger the community, the more people chatting, the higher the usage and api costs so as long as the community scales it wont come back and it will probably get restricted even more

[obtuse smelt]

man i retry generating image like this

Something went wrong with this response, please try again.

[tame mason]

Damm

[rocky geyser]

Its pretty sad it was a reliable way to use opus 🥲

[inland quest]

Majority of all models on arena is free for now.

Frontier paid models removed from normal chats.

[tame mason]

My older brother somehow managed to tell an ai to make an website, and ut has all the opuses, no limits. But idk how he did that

[inland quest]

There is no any use for lmarena for now

Like, only maybe for gemini 3 and claude 4.6 sonnet

And images

Other models are not worth to use (stupid) or available free with more features on providers web sites

[rocky geyser]

I dont rlly believe that is possible unless he is using his own api key, the models arent really opus, he found a leaked api key or he is using some kind of secret unlimited free site🤣

[inland quest]

Like glm 5.1, there is it basic. On z ai it at least can search and read pdf, which basically lmarena can do, but it doesn't.

[strange lynx]

/image-to-vide

[toxic verge]

Thats very strange way to look at it

[stone forum]

welp pineapple and their team are trying their best to fix the arena, but can we also like help them too????

[toxic verge]

If opus is what you’re looking for or anybody in general why not just go get a subscription to Claude?

[inland quest]

Free Vs Paid 😱

[toxic verge]

Well, that’s exactly the issue then

people aren’t willing to pay for it

[inland quest]

The whole point of lmarena was

[strange lynx]

How do I create videos here now? It was there few months ago and I recently came and its gone (the video arenas)

[stone forum]

it's no longer available on discord

[strange lynx]

then where is it?

[toxic verge]

If that was the whole point of the arena, they would just make it all, opus

[stone forum]

in the website

[rocky geyser]

I wont get it because i love vibe coding as a hobby but like on claude even with a subscription here is a 40k-100k token limit per 5 hours and there are weekly limits etc. The limits are so strict😭 Id rather get gemini pro then

[strange lynx]

Ok

[toxic verge]

Well, starting on the 27th they’re gonna give you 15 credits only

Completely drastically reducing

[safe cloak]

i think open source models are the future

[stone forum]

how many limits of tolken is for Gpt pro?

[rocky geyser]

It is only available in the website

[toxic verge]

People are gonna have to start start paying for if they want to continue using it one way or another that’s the sad reality

[inland quest]

Also its visible how lmarena devs trying to reinvent solutions for basic problems

[stone forum]

wait, how many token will be wasted and limited for Gemini Pro?

[toxic verge]

After 2027 AI’s probably double in price lol

[inland quest]

Oh no, our models are used too much
Add stronger rate limits: ❌
Create some shity credit like system from idiotic sites "free 3 uses": ✅

[toxic verge]

We’re probably not even gonna see any unlimited plans also

[toxic verge]

But what do you recommend who pays for it then?

[safe cloak]

only top 1% people will be able to use the best models and we will get only peaseant models in the future

[toxic verge]

Who pays for people to use it for free if people aren’t willing to pay?

It has to come out of someone’s pocket

[inland quest]

Lmarena providing system based on providing to model hosts usage info, datasets and else

Evaluation things

[toxic verge]

Yeah, but people aren’t reasonable. They want the best models for free.

They don’t use it to rate models. They just use it however they want to use it.

They just want to use opus and they could care less about the rankings lol

You said it yourself nobody even uses the other models

[inland quest]

There is no problem to do to claude opus and Gemini 3 the same thing which going to be done for image arena models

Like in last announcement

Thats the point

[toxic verge]

Oh yeah, people are abusing it

[inland quest]

But they are doing "new credits system"

[toxic verge]

People want something for nothing

[stone forum]

what is the new credits system

[toxic verge]

They can’t be grateful or thankful even if it is 15 free credits

It’s free credit that people didn’t have before

Or would you rather them just shut down? And no credits for anybody even if it is 15.

I mean, I’m not happy with the credit system either. I think it’s gonna suck but it is what it is.

If we wanna play, we gotta pay

[inland quest]

Probably you are not catching what i mean

[toxic verge]

I might not be in I apologize

[inland quest]

I mean i dont understand why they are doing credit system if there is fine working "rate limit" system

It was like 5 messages to claude opus per hour

Okay, do it per 3 hours, idk

[toxic verge]

Yeah, but times that by 1 million users

If not more that adds up quick bro

Daily hourly

Imagine their bill at the end of the month, bro

[copper ferry]

https://www.youtube.com/watch?v=IuYL-15jQvQ

[inland quest]

Rate limits is basically designed for that specific problem, lol

[toxic verge]

Well, I hear ya, but obviously they’re not cutting it

Or maybe they are I don’t know. I don’t know behind the scenes.

I’m just taking their word that their face value

[inland quest]

They should just make it stronger

Like done for image arena

[toxic verge]

Looks like they already made up their mind

I just have to accept it either we move on or I don’t know I don’t know what else to say

[inland quest]

Their mind is so much strange

Like whole rebranding

Lmarena -> Arena

[toxic verge]

It’s called evolution

ChatGPT started with a nonprofit

They start with ChatGPT and now they have codex lol

[inland quest]

Change specific name to a generic one is not the good marketing, if that could be applied for that

Lmarena is lmarena

[toxic verge]

They don’t need to market it. They have a business model outside of the marketing.

[inland quest]

Arena is uh... just google it.

[inland quest]

That outside business probably also doesn't care about wtf is that site called or styled

[rapid merlin]

from lmsys to arena

[toxic verge]

That could be strategic also

[rapid merlin]

I mean it kinda worked for them, they grew

but the costs from the growth are eating them alive

[toxic verge]

Honestly, the thing to blame here is the AI industry for increasing the prices

it was a complete lie when they said ai was gonna get cheaper

It didn’t got more expensive

By the time models get cheap, they retire them

[inland quest]

And not really much equalent smarter

[toxic verge]

Ai is going to get more expensive

27/28 we’re gonna be paying double if not, triple

[inland quest]

That doesn't applies for small parameters ai

10B in 2025 and 10B in 2026 is absolutely different

[toxic verge]

The golden days of ai died with chatgpt4o

All that free stuff unlimited access

https://tenor.com/view/gone-gif-24439860

[inland quest]

Also all "multimodal" ai sucks with image recognition

The only ai model who is capable of normally "see" something is Gemini.

Other else is like
* image of the cup*
- What is that?
- Hmm... its a cup
- OMG PEAK

[toxic verge]

Man yall ain’t got no one to blame,w but yourselves yall thought the good times would never end lol

This whole AI industry is predatory in nature. The whole point is to extract as much money as they possibly can.

This is the only inevitable outcome

They don’t care if we complain they don’t care if it gets too expensive and regular people can’t afford it they don’t care

https://flaminghydra.com/sam-altman-and-aaron-swartz-saw-the-future/

Cause none of y’all are interested in the ethics of this

They know y’all are thirsty that’s why they could jack up the prices

Cause they know you’re still gonna pay for it 😝

[ocean ferry]

Bye arena

[surreal zephyr]

gpt has 5x the usercount of all other combined btw

[frosty lava]

The reason why many people actually don't like anthropic is the cost and low usage

so no people ain't paying completely absurd price

only some

[frosty lava]

Just like everything else

the point is to make money

[surreal zephyr]

and the models being dumb as hell

they are like "ill lie so i get liked more" with 0 brain usage

[frosty lava]

And mythos is overhyped considering the api price it'll be

it will cost as much as you'll pay for gpt 5.5 pro api price

or maybe more than this

it won't be something you can use often

it's literally a pro model

atleast the price is the same

would anyone actually use gpt 5.5 pro daily

its the same thing

Pro model are usefull but not as much as normal thinking model

[slender ledge]

Why

[frosty lava]

Its a whole ai race with alot of competition, so increasing the price mean giving people to other companies

[grim canyon]

yeah i can only use it on my previous chats with 3.1

normally there is sonnet 4.5 as u said

[ocean ferry]

Check announcement

[clear spear]

Ouuu I hate not having all the models

[slender ledge]

Uhhh not that bad

[echo dome]

opus is back?

[light sleet]

it's only in battle

It was never removed from battle

[shrewd falcon]

4.7 is out?

[median smelt]

Why is there just random people in the arena stage

[shrewd falcon]

will they add it to direct chat

i jus joined so idk

cuz its only claude 4.6 sonet rn

[slender ledge]

Noooooo

Bever

Mabye 2027

It’s too good

[toxic verge]

Wats the model?

This model is sooo good wow

[quartz light]

@echo aurora reporting this from someone else's behalf

[fossil socket]

I'm legit just waiting for any more news on this usage system

[stray aspen]

The credit thing?

[fossil socket]

Haven't really been using this website on the fact the latest models are well, battle exclusive

Can't really use battle to document stuff now can we

[robust cobalt]

prompt?

[shrewd falcon]

🙁

hopefully they add it

or ad dclaude 4.6 opus bc

back

[vernal raft]

https://tenor.com/view/spiderman-lol-jonahjameson-gif-4589953

[night moat]

Gemini still error

[sour spear]

ChatGPT 5.5, I am quite impressed. 🤣

[night moat]

Yeah. Four prompt max.

[wicked talon]

When you're chatting with Kimi and this pops up

[severe ruin]

Hello

[wicked talon]

At least clicking the message and clicking send works lol

[sterile tartan]

Then you switch to Qwen or Deepseek

Free Kimi is not GOATED anymore...

[wicked talon]

How does bro know my installed apps

https://tenor.com/view/say-whaaat-gif-18261283115155956121

[wicked talon]

True

And £16.99 way to expensive for me

For base

[sterile tartan]

Because im U

Nah don't pay chinese ai subs

[wicked talon]

When was I born

[sterile tartan]

I can't disclose that in public

[wicked talon]

What's the 1st letter of my name

[solemn osprey]

How do I call opus 4.7? I can't find this model.

[wicked talon]

Opus 4.7 got removed from arena.ai unfortunately.

[sterile tartan]

A letter between A to Z

[wicked talon]

https://tenor.com/view/dayum-gif-1189676832747532102

[sterile tartan]

You pay it

[slender ledge]

Gpt 5.5 sucks 😂

[wicked talon]

Why is Chinese ai just built different

[wicked talon]

Correct

[slender ledge]

Gpt 5.4 even better

[slender ledge]

Wym

[wicked talon]

Sounds like my ex, chatgpt

[slender ledge]

Lol

Im not an gpt fan

More Claude

[wicked talon]

I'm a gpt hater

[solemn osprey]

Can I only pay on the official website? Arena doesn't support it?

[wicked talon]

I'm an even bigger Claude hater

[slender ledge]

Same only pic is good

[sterile tartan]

Because it's Chinese

[slender ledge]

Why

[wicked talon]

https://tenor.com/view/china-gif-26678009

[wicked talon]

Way overhyped

[sterile tartan]

For now yeah

[wicked talon]

Ok when tf did chatgpt become rude

[surreal zephyr]

what u say

[visual vigil]

how to use the ai here to generate videos

[wicked talon]

Fu Bi

[solemn osprey]

thank 有

[sterile tartan]

Don't Sub Claude

[sterile tartan]

Their Rate Limits are Sheet

[wicked talon]

Chatgpt is actually so rude

Did openai stop reinforcement learning or smth

[frank junco]

State your request.

[frank junco]

maybe it just hates you bro

[silent tree]

yeah and when its not straightforward yall are like "i hAtE gPt sO muCh It AgReeS wiTh WhatEvEr i sAY!!"

[light sleet]

real

[stray aspen]

tuff

[solemn osprey]

ok

[severe ruin]

I want to learn how to code

Hw can l start

Using ai

[quartz light]

turn off personalisation

[outer flicker]

guy whta best deepseek model use for coding?

[quartz light]

???? do u even know what that means

[quartz light]

v4 pro

[severe ruin]

I know deepseek

[outer flicker]

and they not add gpt 5.5 to direct right?

[outer flicker]

thinking or no?

[severe ruin]

I don't know hw to use it to code

[obtuse heart]

this response a, response b randomly appearing on direct chat is so annoying

like oml i don't wanna switch to gpt 0.022 for sonnet 4.6

[stray aspen]

no

its in battle

[gray ibex]

When will GPT 5.5 be added to the leaderboard? ^

[gleaming roost]

has anyone found opus 4.7 / gpt 5.5 in battle mode?

[stray aspen]

yes

[gleaming roost]

🤔

[proud bobcat]

Holy overpriced

[slender ledge]

Yes

[gleaming roost]

I've been trying for a while now and nothing, I must be really unlucky

oh, so I was saying 🤭

[stray aspen]

damn you got 5.5 high

ive only gotten 5.5

imim gonna develop a serious roblox game with mimo 2.5 pro

lets see how it turns out

[quartz light]

thinking obviously

[outer flicker]

alright

how many token v4 have?

[quartz light]

tokens?

u mean parameters

[stray aspen]

1 million

[outer flicker]

better than gpt 5.5 right?

[stray aspen]

no

5.5 smokes everything

but deepseek is still a good option if you dont wanna pay

[quartz light]

not really

kimi and glm are still better

[stray aspen]

cap

mimo is better

[quartz light]

while being smaller faster

[quartz light]

no lol

[quartz light]

its not open

so its more expensive

because many services can provide open weight models

as they can host it without overpricing

although mimo 2.5 pro is good

but then you can also say qwen 3.6 max is better than mimo

but 3.6 max isnt open

[stray aspen]

just no

[quartz light]

i bet you havent tested it lol

[slender ledge]

Broo

[stray aspen]

i did

[quartz light]

where?

[stray aspen]

its not bad but its not better than mimo

[quartz light]

where did you test it

[slender ledge]

Genimi 3 flash says it’s 24may 2024 💀 💀

😭 🙏 🙏

[stray aspen]

on their website

[quartz light]

are you sure it was max

[stray aspen]

yeah

[quartz light]

it released recently

[slender ledge]

💀 💀 💀

[stray aspen]

cant wait for deepsek v4 omni

[limber hound]

no tools ig

[slender ledge]

Wym

Im fr

[stray aspen]

nosearch

[slender ledge]

It really said

[limber hound]

it can't call any tools that let it know the date

[icy frost]

gemini

[limber hound]

AI ain't all-knowing yk

[stray aspen]

theres no way people voted slopus bruh

[limber hound]

4.7 by far

[abstract heart]

Image To vidéo

[slender ledge]

Gpt 5:5 is bad

Opus better

Fs

[limber hound]

not that good as opus, but still really good imo

[stray aspen]

just no

5.5 is insanely good

[compact flame]

Nah it's mid compared to gpt rn

[spring oar]

what is the problems of opus ?.

[stray aspen]

its not better than 5.5

[slender ledge]

4.7 still the best for me

[fading heath]

hello i greet you all i am new here
i work with comfyui hope to learn more

[spring oar]

Opus is very good

[stray aspen]

how did deepseek get so bad

like god damn

cant even make a simple F to toggle flashlight script

[frosty lava]

4.7 feels like a backward step comparing to 4.6

Alot of people feel like this

[plucky sparrow]

Besides the first Llama when has Meta had a larger community

[stray aspen]

never

who even cares about meta AI

[topaz epoch]

Gpt 5.5 high is good tbh

[solemn osprey]

just got hofburg_2_alt

awful model

[sterile tartan]

5.5 Mogs All

[stray aspen]

this is just the begininng

wait for the spud

[toxic verge]

48 hours left until the new update to the arena

[sterile tartan]

GPT 6.7 will Absolutely Mog All

[toxic verge]

Yeah, except none of us will be able to afford it

It will cost higher than a gallon of gasoline

Rip to that man.

The People’s champ

[stray aspen]

whos that

[scenic holly]

eliy5

[toxic verge]

He’s a hero, bro a legend

He was for the people

[stray aspen]

what did he do

[toxic verge]

He devolves the web feed RSS and was one of the comfounders of Reddit

He did so much dude he’s a martyr honestly

Aaron Swartz. Swartz had a prolific career as a computer programmer: At the age of 12 he created The Info Network, a user-generated encyclopedia widely credited as a precursor to Wikipedia. Swartz’s later work would transform the internet as we know it. He helped co-found Reddit, developed the RSS web feed format, and helped lay the technical foundations of Creative Commons, “a global nonprofit organization that enables sharing and reuse of creativity and knowledge through the provision of free legal tools.” In 2011, Swartz was arrested and indicted on federal charges after downloading a large number of academic articles from the website JSTOR through the MIT network. A year later, prosecutors added an additional nine felony counts against Swartz, ultimately threatening him with a million dollars in fines and up to 35 years in prison. Swartz was found dead in his Brooklyn apartment from suicide on Jan. 11, 2013.

You can read his manifesto: https://ia800101.us.archive.org/1/items/GuerillaOpenAccessManifesto/Goamjuly2008.pdf

[stray aspen]

what the sigma

[toxic verge]

Yeah, you should look into his story, bro

[toxic verge]

@deep adder is this the guy you wanna trust?

[stray aspen]

@quasi atlas

[visual galleon]

Forest

[stray aspen]

oh god

<@&1349916362595635286>

[inner relic]

@quasi atlas

[visual galleon]

Generate image beautiful forest

[void shore]

..?

[noble quarry]

Arena is too much censored.
It all the says this content violets our Terms of Use

[void shore]

it depends on what you are prompting

[visual galleon]

Hey anybody can tell me how can I use the model AI models and new and beginner please help me

[void shore]

yeah, so you go to https://arena.ai

and in order to make images you have to sign in

and then type the prompt into the box AFTER you choose direct mode

battle mode does 2 random models

which you may not like

[visual galleon]

Okay I know but there is no option here showing a i models

Anyone is there

Hello hey please reply

[void shore]

on the top bar there should be a dropdown, where you can select different modes

choose either Direct , or BattleMode from the list

[visual galleon]

I did not know how to use it please give me full information

[void shore]

(i dont see side by side nessesary)

[visual galleon]

Are you saying about the LM arena the platform okay you are give me a interference but in I did not telling about from I want to use in this server so I am download the discord

[void shore]

OHH

alright

i dont believe there is a way to generate in the discord chat

standalone

[visual galleon]

What do you mean about stand alone you mean there is no option to chat I models in this server means I was so many times see you too there was showing the discourse join and free access

[void shore]

The youtube video might have been talking about a differerent sever, like Midjourney

but Arena is a website you have to go to

its not a discord bot nor server

where you can create images straight from the chat

sorry about that

:3

[whole sundial]

it was likely an old video from when arena had the video arena, it was pretty popular

<@&1349916362595635286>

[void shore]

yeah

they did get rid of it

[visual galleon]

Hey thank you about this information your kind person but can you confirm me or you real human

[void shore]

I am a real person, not an ai chatbot

dont worry!

[visual galleon]

Thank you I want so many times avoid social media like Instagram to chat anyone but I will first time chatting you are the first person in my life and chatting

Hey I was want to talk with you more anyone is there

Please reply

[void shore]

sorry for being gone, im active in other discord servers, so i cant be here 24/7

[visual galleon]

This is I am not asking for bad reason Just for curiosity what is your gender

[late rivet]

Why can’t I use any good models

In direct chat

[void shore]

on someones profile you can see their pronouns via clicking on their pfp

[void shore]

wdym

[late rivet]

Gemin 3.1 pro

Gpt 5.5

Claude opus

They’re all gone

[visual galleon]

Hey but reply my question answer about your gender

[void shore]

Probably to cut costs

[void shore]

by clicking on the users profile picture

the image next to their name

[late rivet]

Where’s the website for like

Developing websites

The coding arena

[void shore]

coding arena is in arena.ai by defualt

by the code button

i believe

[late rivet]

I rmb there was one for developers

[void shore]

if thats not what your talking about

let me know

[void shore]

hmm.

ill look into it

[late rivet]

It was dev.lmarena

But it’s in the lmarena.aai website now

[visual galleon]

Any anyone hear me okay leave it anyone can tell what is the re interesting habits in their life

[late rivet]

Cause it redirects me

[visual galleon]

Hello please can you reply

[void shore]

the </> button is the one you are looking for

[loud herald]

@echo aurora Is there anywhere I can tell you about a person you banned? They’re complaining and arguing in my DMs about it

[void shore]

ive heard them say this but you should notify ModMail instead of pineapple

as ModMail redirects to him

for some reason

[void shore]

just dm ModMail

<@&1349916362595635286>

[loud herald]

Alright

[red sluice]

Been testing GPT 5.5 search on the arena. It’s disappointing. For my different uses it is far below Claude 4.7 opus 🙁
not only that, it’s also producing inferior results compared to 5.1 & 5.4.
My usage is pretty hard and the result is objective: 5.5 is disappointing

[void shore]

it depends on hosw you use it

GPT 5.5 is good in many ways

like agentic and coding

[stray aspen]

bro no way

4 in an hour

[void shore]

but if you use it for something it isnt really trained on

[stray aspen]

4 scammers in an ahour

[void shore]

it will often make mistakes

[red sluice]

Yeah but it does not excel in many uses while opus 4.7 is very polyvalent

[void shore]

ChatGPT has been very up and down lately

[wary helm]

love it when both video gens fail 🙂

[void shore]

amazing

[late rivet]

How did you test it

It’s not available

[void shore]

in direct mode

it is on battle mode

i think

i actually havent checked

[void shore]

antrophic made minimax

[late rivet]

Gpt 5-5 is available

I know

[void shore]

STOP

NO ONE IS CLICKING THIS

<@&1349916362595635286>

[red sluice]

You'll find it easily if you turn on search mode on arena.ai

Both opus 4.7 and gpt 5.5 have high chances to be the ones you'll have for test on arena.ai/search

[loud herald]

No, minimax was just partially distilled from claude

[void shore]

i know that

i just find it funny it thinks that

[vale quest]

A

[loud herald]

Yea, thats how people find out what models are distilled from

[stray aspen]

use chatgpt.com then

[vale quest]

Some idiot said they generated an image with gpt 5.2

That idiot is me

Just kidding its some random old woman that has insanely bad hygiene named toni

And thinks she is a star

Actually, I made that up

None of this happened

[haughty timber]

👽

My id is delete

Skkskßodkd

[obtuse heart]

why's this guy pinging random people

[minor pulsar]

Why..🥲

[obtuse heart]

js accept it

arena has grown a lot and they cannot maintain it for free like this forever

that's why they took down high-end models on direct chat..

oop

guess they got banned

[toxic verge]

I find it a coincidence that this growth happened at the same time as AI company started cutting the rate limits and increasing the prices. For example many video models like Sora (shutting down) Grok, even Veo (only offered 1000 months credits, now 60 daily) have all started heavier restrictions behind paywalls. Claude & GPT have severely increased the usability with lower rate limits and making up new tiers of subscription plans, mind you that they are also planning to go public by the end of the year.

[wicked talon]

What AI generator made this

[brisk turret]

4.6 is better

The testing methodology of arena.ai is flawed

[smoky ruin]

what model is this??

[brisk turret]

The multi turn approach is bad

Amd yet multi turn is the main real world way we use llms

There's a much better way to do it

[hollow ivy]

maybe Chad?

[loud herald]

Was it a good model?

[smoky ruin]

yeah ive had it show up a few times
i would say on par with opus 4.6

[loud herald]

in what exactly

Coding?

[smoky ruin]

yea

[loud herald]

I see

Hmmm

You never know, could be a gemini model

[smoky ruin]

i saw someone saying gpt 5.5 in #codename-discussion but i dont think it is since gpt 5.5 has its official name, and its worse than 5.5

[smoky ruin]

true

[loud herald]

Nah no way

[tardy veldt]

anyone know when 5.5 is coming to the leaderboard?

[next trout]

grok suddenly wondered about electrolysis

[proper sun]

Will 5.5 beat opus?

[foggy igloo]

@echo aurora I thought you said leaderboard was coming yesterday. When is it coming g

[stray aspen]

Yes

[tardy veldt]

easily

[stray aspen]

but ti doesnt beat mythos

<@&1349916362595635286>

so many scammers today

[wicked sage]

th

[urban herald]

wdym i got robux from the 2k dollars that mrbreast gave me

[storm dust]

mrneast

[void shore]

No you didn’t bro

[urban herald]

yes i did bro

[compact flame]

I feel like that is a ragebait

[urban herald]

yes i dod bro

[compact flame]

What

[urban herald]

https://cdn.discordapp.com/attachments/1304973426297012264/1487481742083232072/togif.gif

[storm dust]

i got baited hundred times already

stop ples

[vast kiln]

When will gpt 5.5 be on the leaderboard?

[tardy veldt]

artificial analysis had it the second they announced it

arena needs to up their game

[stray aspen]

deepseek is on another level of stupid

[minor pulsar]

Will this pass? Deleting cookies, filled out the form, did everything as it should and writes this. Nothing helps. Today I sent a request for 2753 lines of code and everything worked. Now there are 1231 lines and this... It’s just a shame. Sorry..

[stray aspen]

why is it suddenyl answering in russian

[minor pulsar]

I saw your request)

[raven otter]

Here’s a idea I think NO one has tried: recreate this INFAMOUS commercial: https://youtu.be/v7fTeT3g1AA?si=MDnXAmnIqTP-eiPj

[stray aspen]

are you russian

[minor pulsar]

no. Im use a translator

[dawn path]

Just convert your code into pdf

[minor pulsar]

Thank you)

[prime karma]

@echo aurora Hello

[echo aurora]

Hello

[prime karma]

Why is cloud opus not showing in my chat section

[prime karma]

@echo aurora please help

[echo aurora]

Recently some models have been removed from Direct and Side by Side mode. This was done to help ensure reliability and availability of Arena in the long term. You can find more information in this announcement: #announcements message

[whole sundial]

a good way to reduce these questions is by either putting a message in the models list when people search for these models or just put them in the list but gray them out and then have the removal reason show up when someone hovers over the model

[mossy mesa]

I lost all my chat history. Didnt deleted cookies or anything. Was there some update about non-registered users getting their history deleted? (had no account)

[pallid crypt]

If youve been using lmarena for a while you'll know there data saving is jank as heck

[mossy mesa]

yes was using since many monthes, everything gone....

[pallid crypt]

I'm talking like idk 1-2 years

[mossy mesa]

don't remember when I started using lmarena, probably at least 1 year yes

[spring oar]

Adaptative Thinking with opus is weird ?

[pallid crypt]

I haven't had the issue since I logged in tho I think it's more unreliable for logged out users

[toxic verge]

What’s really ironic is the models that are left some of them are the most notorious for being the most easily jailbroken

[pallid crypt]

Well they are cheaper so theyll have lower parm size so of course that will be the case ig

[toxic verge]

There needs to be some research done on platforms like this in terms of safety. Although on the surface may seem more moderated, arena has their own guardrails in place on top of the models built in safety, the complexity and the sheer scale of the operation with being able to choose from various models at any given time, creates some of the strangest opportunities that would normally not be possible.

[hollow ivy]

lol, the smallest incarnation of it has 235GB

(the 1-bit quantization!)

which model would you recommend, for a 6GB gpu?

why not qwen-3.5?

and llama3.3?

and llama4?

are there any indie models which are good?

ty for the tip!

oh, does a .gguf of it exist?

[stray aspen]

what the hell is this

[hollow ivy]

me too (lmstudio)

i read, they trained it on math, code and such stuff, so it would suck for roleplaying and creative writing, right?

[hollow ivy]

yolo

life is supposed to be fun, you know

yeah, but roleplaying can be kind of kathartic

like reading a book but better

[stray aspen]

arcee is ass

its so bad

[hollow ivy]

Craig, do you watch movies?
If yes, then imagine playing the main character, using only text
that is, in essence RP

maybe you are just not used to it?

have you ever played text-adventures?

(ok, maybe it is not every person's type of tea)

i heard, you can even play D&D with AI now

giving them a ruleset, and let it be a GM

dungeons and dragons

[stray aspen]

whats that

[hollow ivy]

the ancient classic which inspired the "Rogue" game, iirc

also first computer-RPGs were inspired by it

a tabletop game, played with real people, in the eighties

not even, if it's local and hidden on your own computer at home?

(so that nobody else can ever read it)

even if you had CM5? (Claude Mythos V)
(np)

lol

i think cm5 should be stellar for RP

..running on a X-1000 processor ^^

(if you have watched Small soldiers, you got that referenec)

[stray aspen]

<@&1349916362595635286>

[hollow ivy]

if AI gets smart(er), it could become more fun to interact with

i'm a pragmatist, regarding electronic entertainment

if it works, i use it

[stray aspen]

gpt image 2 generator is so good

[hollow ivy]

and i mean, if it gets humanlike, then why not trying it out?

[languid mulch]

I love getting error codes when I'm cooking up something

:d76db87f-bc35-

Anybody know what that means ._.

[hollow ivy]

WoW?

(that game [not warcraft lol])

[stray aspen]

how did you do this

it messed up

[red sluice]

Might be the next Grok model

<@&1349916362595635286>

[uneven cypress]

Nahhhhhhh

Guys does anyone have an api key that has credits

[hollow ivy]

which of their model is best for creative writing, roleplaying and realistic sandbox universe simulations? #ai-news message

unfortunately, that is proprietary, so not included at Nvidia site

-# (i agree, that it would probably be best, if)

is ds4 better than the largest qwen-3.5?

and they even have kimi-2.7

and glm

[stray aspen]

they are all ass

[stray aspen]

wdym 2.7

[hollow ivy]

the AI model "Kimi", version [K]2.7

which is surprising (because it is big, so not cheap)

[inland ember]

I have now got jebel-1

surprisingly good.

[hollow ivy]

how good? top-5 of all models?

[inland ember]

Its nice

[inland ember]

Definitely a top tenner

[hollow ivy]

-# (All exist in that Nvidia website.)

[red sluice]

So specific, and I'm pretty sure no one has tried all those models and can genuinely say one is better than an other for this specific use

[drifting surge]

I'm having a problem right now, where when asking Gemini 3.0 flash, or other AI models, a captcha window is popping up, but it takes so long to complete, that when I do, the AI times out and gives me an error, anyone know how to fix this?

[red sluice]

Played a game like that some time ago, that had fast LLM models... It wasn't really good, and yet it was Grok-fast or something

[red sluice]

Use another IP (i.e LTE/5G/4G whatever), turn off vpn if any, and last resort you can try using free residential VPN.
It is linked to your IP, most of the time.
But yeah that sucks, happens to me as well from time to time, and captchas are too long to solve in time...

<@&1349916362595635286> sorry to annoy you again, it's this bot again

[loud herald]

<@&1349916362595635286>

[vast kiln]

btw. prob not a bot, it could be a hacked account. that hapened to my account a few weeks ago

[loud herald]

It is a hacked account

[drifting surge]

Thanks will try. Had tried a vpn, but didn't help. Happened a while ago and just stopped after a few days, thanks again.

[red sluice]

vpn will make it worst. The captcha system makes it harder if an IP solved a lot of them, as simple as that. A classic VPN can be shared amongst thousands of users daily.
The best way to counter that is to use some IP that isn't used to solve captchas often. Sharing your phone's connection to your computer can work.
Using VPN such as Tuxler (which uses IP like a p2p network, meaning when you use it, you put your own IP in the pool and someone can use it, wouldn't do it, and do it only when I can't do anything else, like I turn it on, turn it off once it's done) can work. It's free, but you share your own IP address, if you let it on, it'll make it even worse when you come back to your "normal" connection.

[light sleet]

of course its gpt spud

[smoky cradle]

Eu gostaria que voce gravasse minha voz, é possivel?

[fiery gull]

where qwen 3.6 bro?

[storm dust]

nuh uh

claude expanion

[silent tree]

no

gpt

[storm dust]

gr

jalapeno anger activate

ok i think i should not have said that

jokes aside...

[silent tree]

😠

[main nexus]

Next Gemini model in 2028, n watch it STILL be behind other ais 🙏💔

Gemini likes to downgrade models heavenly

After 1 week of release

Its kinda sad

[storm dust]

gemini number 1

who here makes a project

a react native one

im asking this because i made a package that finds packages for your project and recommends even better packages. for example: you use expo image it recommends nitro image because it is faster

i think the point is understandable

[storm dust]

it also checks if implementing those will be useful or useless if useless then it doesnt recommend

[main nexus]

Gemini #1 hater here

Watch them release a new model and lie abt their benchmark

Then downgrade it

Gemini 2.5 pro was the ONLY good gemini model

2.5 pro > opus 4 > whatver gpt model was out

Gemini should just quit making models

If they can't make a good model and not downgrade it

I hate gemini btw

[raven otter]

This is true BTW

[dawn mist]

Lowk I liked gemini 3.1 pro preview model I was actually kinda sad when they removed it the google ai studio is overly sensitive it isn’t funny 😭

[dawn mist]

Kimi

[loud herald]

Training is genuinely no joke

The cliff was me training then abruptly stopping it

Went immediately from 100% usage to 6%

[void shore]

Im making my own ai architecture.

no GPT

im calling it LightningMoE-SSM

[void shore]

That name is bad

im making AURA now

cool name

[warm mirage]

if u speak arabic, look this how the algeria telivision is! they are goats!
https://www.youtube.com/watch?v=z_2jQVDCeeo

[velvet furnace]

why no GLM

[hollow fossil]

are u using your gpu?

[loud herald]

Gpu and cpu, problem was that uh my code was set to 634% CPU usage overall. BIG problem

But I put the threads down to 3

So I’m good now

[void shore]

Training a new archecture, Nebula

Nebula is supposed to have minimal params

minimal compute

and fast generalization

[loud herald]

How do?

How so*

[void shore]

Its competition is LightningMoE, my previous

which got 100% on the counting test

I dont think Nebula will cook for its first gen

its practically guessing

[devout venture]

is there anybody who is creatin g ai ugc?

[void shore]

..?

[velvet furnace]

why there no gpt5.5 rank

[neat delta]

Hello

[smoky ruin]

see it didnt occur to me that lmarena would use an lm to handle basic support in #ask-here but thats like so obvious

[torpid oxide]

Hi Can anyone tell me the free unlimited ai video and image generator #every#ask-here

[steel stump]

Hey how to generate videos?

Guys

[wicked talon]

Lmao

[wicked talon]

https://arena.ai/video

[obtuse smelt]

nice is creepy cartoons

[strange lynx]

In the website, you have to sign in

[cedar agate]

Guys where is the video generation channel #ask-here

[echo aurora]

The Video Arena Discord bot has been removed from the server. You can find Video Arena on our site here: https://arena.ai/video To learn more about how to use Video Arena I'd recommend this article. More information can be found in this announcement about the removal of the bot.

[neon idol]

<@&1349916362595635286>

[iron laurel]

Is Direct and Side-to-Side coming for Video?

[slender ledge]

https://discord.gift/jZ7AqkJmmsfpA6zxhvc2NBE4

Lol

[solemn warren]

seems arena no longer helpful over the official model site!

[supple topaz]

Wait when did the image and the video generation channels get removed?

[crystal mica]

when the f**k we will get gemini 3.1pro back in direct

[light sleet]

who else watching pineapple stream

[dawn iris]

The neural network has an application

[vernal raft]

#announcements message

We will see tmr

[stone forum]

that's only for image

[light sleet]

@surreal zephyr @silent tree

[silent tree]

AHH

SO TUFF

TUFF APPROVED

WATERMELON APPROVED

100=

[surreal zephyr]

Tuff

[light sleet]

what if i generate chatgpt generated inside of Chatgpt generated inside of

[flint sandal]

i was wrong about deepseek

it is a really great, balanced model

like its not the best SoTA like kimi k2.6 or glm-5.1

but it doesnt overwork things

like yk

you ask for a landing page

and GLM will add like visualizations you didnt ask for

on the site

but deepseek just does the prompt

gives polished, good results

and doesnt overdo anything

best open llm imo

maybe mimo 2.5 pro is better