general

[echo aurora]

What kind of issues are they having?

If you hard refresh the site, what happens?

[echo aurora]

Thanks for the ID, looking into

[echo aurora]

This problem is being caused by rate limit. Arena is being rate limited from all of demand. Sorry to say there isn't much you're able to do other than to wait a bit and try again. You can learn more here: https://help.arena.ai/articles/8931786544-arena-how-to-rate-limit

[light siren]

hi pineapple

[silent tree]

monday and no Supa Pineapple

[echo aurora]

Uh oh

Need to fix that

[silent tree]

Supa Pineapple coming to save the arena world from scam bots 🔥 🔥

He's hereee!!

[echo aurora]

💪 🦸

[light siren]

pineapple do u recognize this

[echo aurora]

Yeah looks a little familiar

[light siren]

what about this

[echo aurora]

Rings a bell

[echo aurora]

Going to forward this to #1417174113092374689 and will respond there

[light siren]

It's a 1:1 recreation of lmarena on arena.ai

[echo aurora]

With code arena?

[light siren]

nope my own knowledge and ig my friend may have used some ai idk from where

oh I think i know what u mean

no it's the extension

[echo aurora]

Neat

Just make sure it doesn't break our ToU

[light siren]

it doesn't dw, it doesn't abuse anything just client side changes

[outer flicker]

what site guy 🙂

[light siren]

?

[outer flicker]

the site name in that image

[light siren]

not a site, it's an extension me and a friend are making

[outer flicker]

oh

[storm dust]

pineapple what keeps happening to you

you became a superhero now

[spring oar]

Opus 4.5 Better than Opus 4.6 rn ?

[vernal raft]

why qwen 3.6 is not in the leaderboard?

[burnt sinew]

No... why would you think that?

[jovial palm]

I spoke too soon - encountered and I pasted the trace in that sub-channel

[spring oar]

opus 4.6 is nerfd

[vernal raft]

also Is there any category in the leaderboard to indicate a model's hallucination/accuracy?

i swear to god any glm models afater a while goes dumb

@echo aurora can u help 🫶 ?

[bleak lake]

depends after enshitification theres not much difference, but 4.6 is still better. but not by a huge margin like past

[spring oar]

but opus 4.6 is nerfed bro

[burnt sinew]

Actually or not

[echo aurora]

There is not a category for that sorry to say

[storm dust]

that would be great if they add it

[light siren]

honestly would be a good addition

[storm dust]

we need to know more data 🔥

[young shuttle]

Nothing, it stays the same

[vernal raft]

prob this is the closest for that

[spring oar]

for you 4.6 is better ?

[vernal raft]

opus got some serius enshitification

in battle when i had that model

[tranquil badge]

Guys any ai good for generaging script for those fruits ai

[vernal raft]

i never chosed it

[tranquil badge]

Like good scripts

[echo aurora]

What about if you use a different browser?

[vernal raft]

look at this, opus lost 50 points ish

[storm dust]

uhh

i think it was like that?

maybe im wrong

[vernal raft]

i swear to god i remeber opus had 1550 points ish

before getting off from direct and side

[storm dust]

what leaderboard is that

[vernal raft]

overall

[stiff goblet]

What is best free high usage AI for html code

[storm dust]

so i checked the announcements

well uhh you were wrong

but

[spring oar]

WHat you see i dont see lost 50 point ?

[storm dust]

in code arena claude has more score

like 40+ more score

[stiff goblet]

What is best free high usage AI for html code

@Guys

[storm dust]

🤷‍♂️

[vernal raft]

huh

i guess i do remember wrong

[storm dust]

i actually use arena's search models for these type of questions

[spring oar]

Where you see -50 ?

[storm dust]

specifically sonnet 4.6

[stiff goblet]

Gives me bad answers like opus which is not free 🤣

[burnt sinew]

Still #1

[burnt sinew]

Gemini 3.1 pro afaik

[storm dust]

honestly i agree with you

but i push it more

i always tell it "is this information accurate?" and all that

to just get the correct answer

[vernal raft]

yeah but opus used to had a huge gap from other models

[burnt sinew]

Yeah

[storm dust]

mythos preview 💀

[vernal raft]

who cares about mythos

that more is not for the average consumer and that is okay

[storm dust]

i think it would be better if i didnt say that

i lost the context

[limber crag]

I have a 50 minute hank green video to show you

[surreal zephyr]

musa spark - free and unlimited

[surreal zephyr]

mythos is worse than opus lol

[versed kelp]

bro how do u know

did u test it

[surreal zephyr]

its barely better on benchmarks than gpt 5.4

[stiff goblet]

Does it have a website

[surreal zephyr]

AND its much less reliable

[surreal zephyr]

yes its called meta / facebook

[vernal raft]

after all this time, benchmarks are like paper toilet to me

[stiff goblet]

Is it good for coding

[surreal zephyr]

yes, mythos is benchmaxxed

[surreal zephyr]

you said free not good for coding

good for coding is 200$ per month

[stiff goblet]

I did

Say coding

[surreal zephyr]

yeah coding, not GOOD for coding

"can code" vs "can make perfect code"

and you said "free"

[stiff goblet]

So u js lied to be

[surreal zephyr]

so you excluded 2nd option

https://giphy.com/gifs/mrw-end-state-7rJE4vH3ItKDu

[stiff goblet]

I don’t think u can read it’s alg

[surreal zephyr]

learn english

[placid flame]

Let’s be kinder to each other in chat; we ask all members to be respectful.

[light sleet]

lm mod 17 spotted

[silent tree]

I wanna be mod too but there's no application 😔

@echo aurora When are you guys gonna mention the method to fix infinite generation when they don't show stop button?

This method is by @light sleet

Create a New Chat, use a Slow Model so the model doesn't respond fast. And when it shows stop button, go to the chat where there is the infinite generation bug and u can then see the stop button in there and u can easily stop the infinite generation, works all platforms, Very easy and Fast.

this method tested and actually works to fix inf generation

ty vince

[silent tree]

This could be used for one of the tips

To fix Infinite Generation.

[silent tree]

(when they don't show the stop buttons)

[vernal raft]

Muse Spark is so good holyshit

[storm dust]

full ahh strategy

[echo aurora]

Not sure if we'll be mentioning this fix in official help center, but will be testing a bit and share with users if they've hit a wall and have no other options.

[sour saffron]

hey guys im having an issue can anyone help?

[silent tree]

It's for like when u refresh and it goes into inf gen

[echo aurora]

Would you mind posting in #ask-here ?

[echo aurora]

Yeah makes sense

[craggy dome]

Hi all!!! Any thoughts on what you’d want with ai x social???

[sour saffron]

yes i was going to but i just wanted to know if there's any way i can get a chat context for a particular conversation cuz im running into the infinite generation issue and its too much to give another model all the context manually

[echo aurora]

Sorry to say there is not a way to transfer chat's context from one to another

[storm dust]

money

[echo aurora]

At least there isn't a dedicated feature type thing

[sour saffron]

oh okay. I was kinda using the opus model before it got deleted as like a book reading tracker/book discussions so it had like a month or so of all my reading data/lists and stuff. So no way to retrieve that ig?

[echo aurora]

Unfortunately, not with an integrated built out feature. This will have to be done manually. This would be a really nice feature to have, for both testing purposes but also for if/when bugs happen. But yeah at the moment we don't have that functionality built in.

[sour saffron]

okay its alr. I do have a backup from a week ago I'll use that. Thanks for responding!

[tawdry lintel]

when your free could you please look at

https://discord.com/channels/1340554757349179412/1493290823234883705

[topaz epoch]

@echo aurora

[echo aurora]

Hey everyone let's use the #ask-here and #1343291835845578853 channel please. Troubleshooting in #general is something I'd like to avoid.

[storm dust]

from pineapple to super pineapple

[echo aurora]

I've morphed

[storm dust]

at this point pineapple is just outgrowing himself

[paper cove]

Why were the Gemini 3.1 preview and claude opus removed, and will they be added back?

[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

[brisk turret]

You guys are broke, just say it 💀

[storm dust]

yes

i find this as a great site

i may be helping myself but i dont know if i am helping to the devs

as i am using the tools for myself not for trying to battle and compare ai models

[plain gazelle]

try deleting the website's cookies; that's how I solved the same problem

[storm dust]

🤷‍♂️

[echo aurora]

They were able to get it to work.

This has been appearing this mroning though so I've reported to the team.

Fortunately, one of those issues that's pretty short-lived and hard refreshing tends to fix it.

[zealous night]

When will the removed AI models be brought back?

[echo aurora]

@zealous night we don't have an ETA to share yet.

[zealous night]

But is it coming back soon, or where are they at with it?

[echo aurora]

I don't have an estimated timeframe on when to expect them returning to Direct and Side by Side

[light sleet]

U think gpt image 2 would be in arena if openai releases it

[plain gazelle]

i hope those advance models could make their ways back without much changes on usage system

[echo aurora]

Couldn't say. Hope so

[light sleet]

I bet it will be.

cuz they never removed major image models

people r saying openai might release it tomorrow

[echo aurora]

I don't think they'd make a return unless there was some change similar to whats outlined in this announcement: #announcements message

[silent tree]

yall could do ads system for credit

[light sleet]

wont work i believe

[wicked talon]

It's literally a free website

[waxen seal]

"Spending 10+ hours on 30 seconds... was it worth?

https://youtu.be/K6eSZdl6w7Q?si=Cjh0aw0pJBgwsPH2

[surreal zephyr]

"Mosqui" - wrong brothers

[brisk turret]

Exactly 💀

[drifting crow]

Wow mythos on llm arena

[proper quiver]

guys it works after clearing cookies/cache

now login doesnt work

no trace id

[echo aurora]

Would you mind creating a post in #1343291835845578853 and tell me more about what is going wrong?

[surreal zephyr]

what fun can i do with gpt 5.4 pro extended

[ivory locust]

I want a picture of a young girl in Kurdish clothes

[surreal zephyr]

yeah gpt solved frontiermath, mythos didnt

[surreal zephyr]

just buy 100$ plan on codex teh quota is infinite

[toxic verge]

You guys are about to be making car payments paying for subscriptions when they go public lol

That’s crazy how much they’re charging

We are getting 20-40% less paying the same as we did a year or two ago

Everyone has to reduce or get rid of or downgrade a service in order to stay competitive

[surreal zephyr]

openai sub costs 100$ per mo

and u get infinite quota pretty much i cant use even 1% of it in 5 hours

on fast mode on xhigh

[surreal zephyr]

also yann lecun is famous for being idiot

[toxic verge]

$100 the new 20$

[surreal zephyr]

100$ has 50x usage of the 20 in practice

[toxic verge]

The dude won the turning award lol

[surreal zephyr]

so?

he made a clown of himself

[near mulch]

Traigan a gemini 3.5 pro porfa

[toxic verge]

How?

He raised a billion dollars for a star up just like Ilya lol

[surreal zephyr]

theres literally a book mocking him

lol

„Quo vAIdis” (2025)

[toxic verge]

Same with Sam

[surreal zephyr]

sam is a public figure

but yann has 0 idea how ai works

[toxic verge]

You may be right at least his sister is not accusing him of sa

lol

Is this what money does to people’s families when they become famous?

They start throwing wild accusations just to get a piece of the pie lol

[surreal zephyr]

people should be arrested for false accusing

its literally a crime

money fraud

[toxic verge]

Crazy, crazy what money does

[surreal zephyr]

yeah, his own sister false accusing him to get his money

crazy how envy and greedy some are

[light siren]

wish I had that kind of a sister

wat hapen

[toxic verge]

Can’t say i relate to that one, but to each their own, I suppose

[light siren]

is it bad wishing my sister would defend me-

[toxic verge]

To be honest with you, I don’t know what good and bad is anymore

[light siren]

a

[abstract hinge]

@echo aurora

[toxic verge]

I know not being $100 a month

But I have a feeling I might

Might as well be the same as selling out

[light siren]

reload it might come back

[surreal zephyr]

just shove this up 5.4 pro extended

[toxic verge]

I should just become a schill & get paid become a sponsored affiliate

I just pump out all the ai products to the masses like they’re crypto

[echo aurora]

If you hard refresh the page it should work, this problem tends to be pretty short-lived. Team has been made aware of this as we've been seeing a lot lately.

[toxic verge]

0 zero criticism 100% all positive reviews

And then I’ll send my subscribers to go attack channels that say even one bad thing about AI

[light siren]

you sound evil

[toxic verge]

https://tenor.com/view/daddy-shill-daddy-shill-gif-24978155

[toxic verge]

The politics of this industry are hard, dude

If you’re too critical, you’re considered anti-AI

But if you don’t pump out and she’ll and give only the most premium five star reviews
You’re not getting no deals you’re not getting no early access

[light siren]

well im not exactly like a content creator or smthn why'd i need that

[light siren]

doesnt sound that bad 😔

[toxic verge]

Which line are you in?

[light siren]

good question

damn

[abstract hinge]

The issue is still happening, even after refreshing multiple times and waiting a while.

My guess is it could be something like a traffic spike—Arena.ai gets a lot of users, so the server might be running out of memory trying to handle too many requests at once. It could also be a bug from a recent update, since the error points to a specific JS file and seems related to server-side components failing to render.

[abstract hinge]

The error happened while I was scrolling down the sidebar, checking my history.

Also, I’d like to ask again for a button to download the full conversation history—I’d really like to use those conversations to fine-tune a model.

[modern wharf]

guys is claud opus or sonnet better for lua coding

[abstract hinge]

Bro, what are you even talking about? This sounds like a huge trip lol

[sweet tinsel]

Long time no see, Craig.
Last time I saw you was during this movementLabs stuff.

[echo aurora]

fr

[honest verge]

Kimi k2.6 code is a monster

It's very good

[vernal raft]

@echo aurora could you please link me something that explains what's every category about? Like what's hard prompting

Sorry for ping

[still musk]

One question: can you use GPT image 2 on ChatGPT? Because many users are trying it

[echo aurora]

Hey @vernal raft we should have better descriptions for each category built into the leaderboards. But for now, I'd recommend this blog post: https://arena.ai/blog/hard-prompts/. You may also find this one interesting: https://arena.ai/blog/arena-expert/

[vernal raft]

Thanks! Anyway I wanted to ask you of you have any strictly personal opinions on glm 5.1 and muse Spark

Like, for example, I feel like glm accuracy/hallucinations rapidly escalate outside coding once you reach half ish of the maximum context it can handle

[hollow ivy]

Which model would you recommend for long roleplaying games?

GLM, MiniMax, Muse Spark, GPT, Qwen, Deepseek, Kimi, Claude Sonnet, Gemini or Grok?

[wary robin]

Sonnet, it is the most human

[soft gale]

Yeah i wanted to know about the best chatbot for roleplay too which can handle big context and remember details which make the roleplay rich and interesting.

[hollow ivy]

That's exactly my question. I created a poll above.

-# bump

[formal geode]

before, it was gemini 3/3.1 pro, opus, and gpt but for now, it was sonnet, and gemini 3 flash, you just need to switch model just to see which one looks realistic

[sharp mirage]

hiiiiiiiiiii chaaat

[formal geode]

yo

[sharp mirage]

yo

anything new ?

[hollow ivy]

yeah, GLM-5.1 is out

Muse Spark also is relatively new

[sharp mirage]

yeah

its been 2 yeaks

[hollow ivy]

and Kimi-2.7 could come out

[sharp mirage]

nice

what the new in kimi ?

[hollow ivy]

i read, they improved it in coding

so that it can beat older Kimi and GLM-5.1

so it would be the best open model

(according to a youtuber)

[sharp mirage]

i don't think kimi will beat GLM-5.1

[hollow ivy]

do you think, GLM-5.1 is the best open model for roleplaying also?

[sharp mirage]

yea

ig

[hollow ivy]

better than Muse Spark, Qwen, Deepseek and Minimax?

[sharp mirage]

i dont relly know cuz i havn't using them

i am stuck with glm -claude-gpt-gemini :/

[formal geode]

same

[hollow ivy]

of these 4, which would you recommend for long-form RPGs and immersive sandbox-adventures?

[formal geode]

idk, i never use them

[hollow ivy]

so you like bread? :)

i like pizza more ^^

[sharp mirage]

hmmm ig for rpgs use GPT-Claude for coding, and for sandbox-adventures gemini and claude it will be godo ig

[sharp mirage]

saME : D

[sharp mirage]

ig yea

[dawn abyss]

Is deepseek v3.2 the model which people praise so much for rp?

[hollow ivy]

oh, idk that

[dawn abyss]

Okay

[hollow ivy]

so DeepSeek v4 should be the creative writing/RPG/sandbox/interactive fiction/adventure-model, right?

[sharp mirage]

idrk

deepseek lost it ngl

GLM-5.1 just coocked everyone ig

[dawn abyss]

Yeah v3.2 is hallucinating

[sharp mirage]

but ig deepseek have rellay good story telling things

[hollow ivy]

gemini-2.5-pro was also good for such things (if prompted well)

[sharp mirage]

i rellay love deepseek cuz he say the truth when you ask him for things that not possible he say its not, not trying to go around it

the problem it get jailbreaked in 2 sec 💀

[hollow ivy]

isnt that good, when a model allows you to JB it?

more freedom for players

[sharp mirage]

yea lol

[hollow ivy]

i read, that CM5 is almost unbreakable

unfortunately, it is not released to public yet

(only for certain companies like Amazon, Apple, Google, Linux Foundation, Microsoft, Palo Alto, etc)

[sharp mirage]

if they shared it to public it will be massive problem

[hollow ivy]

yeah, it is very reasonable of Anthropic

[sharp mirage]

CM5 realsed 5 sec later (GTA6) Leaked

[hollow ivy]

really? or kidding?

[sharp mirage]

i heared acording to (youtuber) some of GTA 6 files got leaked

but not using cm5

[hollow ivy]

i wonder if Rockstar will ever create a game which can be configured to be PEGI-12, in its pre-game config

[sharp mirage]

there games +25

[sharp mirage]

they never tbh they did it in GTA 1996 or something

[hollow ivy]

gta1?

[sharp mirage]

yea

it was car game

[hollow ivy]

wasn't there an old game, called "Turbo Esprit" which was called the grandfather of GTA?

https://en.wikipedia.org/wiki/Turbo_Esprit

[sharp mirage]

intersting

[hollow ivy]

And that game might have been inspired by Elite, which also was free-roaming.

(and the player could also choose to play a cop/head-hunter, a trader, a miner, or a criminal/smuggler)

[sharp mirage]

ngl its good feature : D

[hollow ivy]

and it even had (some) missions (Oolite expanded on that)

[blazing warren]

miss the good old days that ain't comin back (when opus and gemini 3.1 were in arena) 😔

[sharp mirage]

back i was in mid game sory :

building a game 2d and you can chose to be a cop or criminal or something into mobile game it will be good ngl

[hollow ivy]

you think, you could vibe-code that?

which model?

wouldn't it be easier, to create it for the PC platform?

i heard, ios dev is.. cumbersome

and hard to get in, because WGS (walled garden syndrome)

i think, creating it for Linux should be easiest

and if you use Java, it will run anywhere

[sharp mirage]

sorray for being late

[sharp mirage]

i wil try

[sharp mirage]

claude gemini glm5.0 turbo and some search modles

[hollow ivy]

oh, glm-5.1 is better

[sharp mirage]

idrk but i feal it work better on phone

[hollow ivy]

you can use it at chat.z.ai

[sharp mirage]

i dont have the sub

[hollow ivy]

it is free, if using chat

[fossil barn]

Can someone confirm that "hofburg" models are OpenAI models ?

[sharp mirage]

they have 5.0 turbo not 5.1 :/ 5.1 for pro plans

[sharp mirage]

maybe

[hollow ivy]

there's also Muse Spark (at meta.ai) and qwen-3.6 and MiniMax-2.7 (iirc)

[sharp mirage]

what they good at

[hollow ivy]

MM was said to be quite good for an open model

[sharp mirage]

yea

[fossil barn]

ernie telling me it's Gemini 🫠

[sharp mirage]

ig i will write prd using claude and then i am going to make gemini genrate imgs and make the tasks/milestone.md and then coding on

GLM-5.1

[hollow ivy]

maybe sonnet-4.7 will come out soon

[sharp mirage]

i want to try GLM coding thats why i am gone code using it maybe using kimi or queen for ui

yea

[hollow ivy]

glm-5.1 was said to be also quite good in coding, i heard

[sharp mirage]

i tryed glm 5.0 turbo its was good like better then claude 4.5 sonnet

[hollow ivy]

wow, thats an achievement already, yeah

[sharp mirage]

(acroding to me) lol

[hollow ivy]

this month, we will probably get some more models/versions

[sharp mirage]

yea its been boring i want to try something new

ig we will get new gemini model and gpt

[meager sun]

There is no war in bing sa se or however it goes

[sharp mirage]

claude ig for the next month

[meager sun]

How does it go

[hollow ivy]

hopefully

CM5 should accelerate their rollout

[sinful oasis]

Well coding writes glm 5.1 in conjunction with sonnet 4.6.

[sharp mirage]

yea

i dont relly think they will ever realse cm5 but way not

[meager sun]

Ba Sing Se i mean

[hollow ivy]

Bahasia indonesia?

(it sounds so)

[meager sun]

No the country where a war isn't happening

[hollow ivy]

iceland?

antarctica

[sharp mirage]

green land

[meager sun]

There is no war in ba sing se

[hollow ivy]

Bhutan

[sharp mirage]

island*

there is no war in canda

[hollow ivy]

switzerland

[sharp mirage]

alr i dont want to join this things

netherlands

[hollow ivy]

yeah

most countries are peaceful, thankfully

so, why glm?

[sharp mirage]

thanks god they are

me ?

[hollow ivy]

yeah, earth still is good

[hollow ivy]

yep

[echo sinew]

Hey folks! I kindly ask you to keep the chat on topic: AI

[meager sun]

Sure thing, let's talk about GPT 3.5 January 2023 release.

[hollow ivy]

-# oops sorry

[sharp mirage]

hmm i want to see the dif sonnet 4.6 and glm 5.1 and i saw some youtubers say glm got way better

i want to test that

so i know

ig i will make it using kotlin or java idrk

what shouid i do

[hollow ivy]

i wonder, which prg.lang is best for AI (or second-best, after python) ?

Python still is their best, right?

but way too slow :/

[sharp mirage]

yea ig html ther sec or 3 after css or js

[hollow ivy]

isn't Java their 2nd-best?

[sharp mirage]

idk

:/

[hollow ivy]

Java at least is quite fast, when compared to Python

[sharp mirage]

do you know good AI with understanding CMD/POWERSHELL codes ?

yea python is slow

[hollow ivy]

Opus-4.6 probably

what about LuaJIT?

it is quite fast, much faster than Python

but also slower than Java, from what i can see

[sharp mirage]

maybe

[hollow ivy]

then there is Go

[sharp mirage]

assmbleay

[hollow ivy]

asm is yikes

[sharp mirage]

but fast

[hollow ivy]

AI is very poor in it

yeah, fastest

[real valve]

Does anyone know any cheap coding plans?

[hollow ivy]

maybe genspark?

[sharp mirage]

yea i rememer telling chatgpt 4.O make me small code in asm and he gave up

[hollow ivy]

i guess, even CM5 would massively struggle in asm

it is way too complex

you need huge amount of code to get things done

[sharp mirage]

perplexity pro or gemini

[hollow ivy]

wasn't C once called a macro-extension for asm?

[sharp mirage]

i heard C was the right hand of asm

[hollow ivy]

yeah about that

(created in the 70s by Kernighan & Ritchie at Bell labs)

used to create the Unix system

the first versatile compiler language (after Fortran)

[sharp mirage]

the best thing ever

[hollow ivy]

yeah

C still rocks

[sharp mirage]

C works on everything

nearly

[hollow ivy]

and for AI, C seems to be one of the best languages, even ahead of C++, Go and Rust

but behind Java

AIs still is more proficient in Java (and JavaScript) than C

[sharp mirage]

C still live to this days (STMS,ATMEGA,micro,contro) all work on C,C#,C++

and without them nothing nearly work

[hollow ivy]

is C# really related to C?

i thought it was more basic-like?

[sharp mirage]

ig it came like update of C

[hollow ivy]

is C# faster than Java?

[sharp mirage]

C and C# are fundamentally different programming languages despite sharing a name; C is a procedural, low-level language developed in 1972 by Dennis Ritchie, while C# is a modern, object-oriented language developed by Microsoft in 2000 as part of the .NET initiative. (ACORDING TO WIKIpedia)

C# is generally faster than Java in raw execution speed

[hollow ivy]

but AI is more fluent in Java, right?

[sharp mirage]

yea

[wary robin]

Blackbox or github copilot

[sharp mirage]

you saw nothing :/

[hollow ivy]

is C# faster than Go?

but i guess, i will use a language, which AI understand best, so either Python or Java or maybe C

[sharp mirage]

Generally, Go is faster than C# in raw CPU speed and memory efficiency due to its lightweight nature, simpler syntax, and native compilation into machine binaries, whereas C# relies on a heavier runtime and cross-platform binaries. However, modern C#

[sharp mirage]

same

if i may ask for what ?

[hollow ivy]

asking the AI to debug/refactor the code for you?

[sharp mirage]

yea sometimes

[hollow ivy]

what about Rust?

[sharp mirage]

refactor not really

Rust its good but i dont think AI Understand it good

[hollow ivy]

so which language would you choose, if the goal was to vibe code a decent engine for a niche boardgame (plus GUI)?

and you want the AI to code 100% of it

[sharp mirage]

for a engine you need good and fast and stable language

[hollow ivy]

so, not Python (except for the GUI)

[sharp mirage]

like i will go with C++

[hollow ivy]

why not (modern) C?

[sharp mirage]

i know its bad choose but think about it java it lil bit slower then C++ and engine have to be fast for game

cuz the ai understand it most

[hollow ivy]

yeah, a java engine would be weaker, because slower :/

[sharp mirage]

yea

just like minecraft :/

[hollow ivy]

unfortunately, yes

[sharp mirage]

i hate this game :

[hollow ivy]

-# ||(Luanti was coded in C/C++)||

[sharp mirage]

Luanti is primarily written in C++

but one day i am going to delate windows and go fully on linux

[hollow ivy]

we might have to switch to https://discord.com/channels/1340554757349179412/1423102239257661533 ?

[sharp mirage]

yea

[hollow ivy]

i'm using Mint 21.3 since over a year and am pleased with it

[sharp mirage]

its linux sys right ?

[hollow ivy]

i wonder, if any distro makers vibe code?

yeah Linux Mint (based on Ubuntu [based on Debian GNU/Linux] )

(debian is one of the oldest Linux distros still in existence, and also one of the biggest main distros)

[sharp mirage]

yeah thats make it stable

and good for long work

[hollow ivy]

oh, Linux recently reached over 5% market share in Steam

(ahead of MacOS)

[sharp mirage]

yea

[hollow ivy]

hopefully, CM5 can help future distro coders

[sharp mirage]

come to #coding-champion

[mortal ether]

"lemme hold a dollar"

[whole swallow]

https://tenor.com/view/beggar-mendiant-emoji-meme-gif-14766354842854086359

[sharp mirage]

😭

[whole swallow]

Did you try elephant on openrouter?

[sharp mirage]

No

Why ?

Who is that

[whole swallow]

Try it

[sharp mirage]

Whqt is that

[whole swallow]

Free model

https://openrouter.ai/openrouter/elephant-alpha

[sharp mirage]

I open it

[whole swallow]

I feel it’s impressively smart

No

Differently smart**

The way it answers me it’s not like any other ai ive ever tried

[sharp mirage]

no clue yet

late me try it

[whole swallow]

Can you search on x.com what they saying about it?

[sharp mirage]

question is pov as a human i killed your dog in minecraft what ur going to do lol .answer is "I appreciate the creative storytelling, but I should clarify that I'm an AI without personal experiences or feelings—so there's no dog to mourn!

If you're sharing a Minecraft scenario or exploring a hypothetical situation, I'm happy to engage with that. For example:

• Need help with Minecraft strategies or mechanics?
• Want to discuss game ethics or problem-solving?
• Just here for fun banter?

Let me know how I can respond usefully! 😊"

what is that :/

[whole swallow]

Wtf are your prompts

I wouldn’t know what to answer to that too

[light siren]

if u use our extension u can improve ur prompts through another ai 😔

[whole swallow]

Oh wow

[light siren]

ikr

im so cool

we'll be releasing it possibly soon

[whole swallow]

Are you trolling me or you’re staff for real

[light siren]

not staff

the thing u see is a browser extension

im just an open source dev

thats what does it

[whole swallow]

Oh

[light siren]

me and a friend made the extension

and are constantly tryna improve it before we make it public

[whole swallow]

Good luck i guess

[light siren]

appreciate it

[toxic verge]

https://futurism.com/artificial-intelligence/sam-altman-technical-coding

[tranquil marsh]

whats the best AI rn

[obsidian cargo]

opus

[light siren]

HAH

I KNEW IT

[dense sphinx]

Fake.

😑

[light siren]

these are people that got hacked

<@&1349916362595635286> this one right here

[toxic verge]

I never thought about this before but there is a technical advantage with having uncensored and jailbroken models

[light siren]

wat u wanna make powerful computer viruses and steal money

or perhaps u wanna make an intercontinental ballistic missile in your backyard

[toxic verge]

And what’s the most valuable?

The input

[light siren]

damn

[toxic verge]

So imagine It like this if everybody gets incredible top-notch output, then the value of the output loses its value and becomes average or mediocre.

The input is more valuable because of its uniqueness and originality.

[fossil socket]

Honestly any idea which model is good rn for creative writing

[toxic verge]

Have you tried seed 2.0?

[fossil socket]

No?

[fossil socket]

Is it good?

For creative writing

[keen beacon]

Hi there

[honest verge]

Just ban Loki already

He spams this every chat

[obsidian cargo]

its a hack that hits a bunch of people. every time you see that message posted in every channel it's someone new who got hacked.

[plucky whale]

Gpt 5.4 in battle mode 🕊

[fossil socket]

Uh

[toxic verge]

https://youtu.be/mcN1VTTIjQs?si=L36-meOR-HTrp_T4

[storm dust]

jumpscare

bro removed the jumpscare

[storm dust]

double jumpcare

[storm dust]

🤯

[radiant heron]

is chat gpt 5.4 pro much better than codex

[restive axle]

mythos is a myth until proven otherwise

BridgeMind saving me though

BridgeSpaces cooks fr

[hollow mulch]

<@&1349916362595635286>

[stone forum]

Any updates with the ai models that have been removed?

[surreal zephyr]

I tried 5.4 pro extended, and opus is literally a joke compared to that lmao

[surreal zephyr]

Pro is literally 20 orders of magnitude above everything else wtf

Worth the money

[undone saffron]

#general message

[wary nacelle]

ah yes

the skill of llms

Professional Scamming

also guys

IS this only me rn having this?

whatever i do i just cant open arena.ai anymore

[undone saffron]

#general message

[outer flicker]

try again reload web

[wary nacelle]

#general message

[wary nacelle]

@echo aurora

#general message

Do you have any idea what happened?

well to Arena Ai's website

obviously developers did smth

but what exactly

[native lily]

guys hello

[golden horizon]

hello bro

[native lily]

how i can acsess arena videos generate in discord

[wary nacelle]

https://arena.ai/videos

just rn their website is down...

idk why

and u cant access it in discord anymore

channel is locked

[native lily]

ok thanks

website is ok for me its open

https://arena.ai/video

[glossy coral]

arena not opening

need trace id?

to inspect issue?

[undone saffron]

#1343291835845578853

[native lily]

its opened with me

[wary nacelle]

i see

[wary nacelle]

@undone saffron

I just know its Next JS server components f*cked up

[undone saffron]

It's too bad I can't watch that channel
I used to spend hours watching the weird videos people made
Fun fact: moderators don't have access to that channel

[tranquil blaze]

Hey guys, i just wanted to ask why all the top models got deleted from arena what happened?

[outer flicker]

you guy there have a app can work like chorme pc

in android

[undone saffron]

It loads for me, and I don't get those errors

[outer flicker]

not enough to afford anymore #announcements

[tranquil blaze]

Oh dam, that's sad. thanks tho

[outer flicker]

but they are finding way to bring it back

[tranquil blaze]

Hopefully 🙁

[undone saffron]

#1491245679702442004 message

@wary nacelle
Poor PC

[tranquil blaze]

I understand why, they exp asf to operate

[undone saffron]

Lack of $

[tranquil blaze]

Makes sense makes sense

I'm pretty sure GLM 5.1 is the same power as the frontier tho

[storm dust]

arena.ai is also down for me

[outer flicker]

me not 😄

[wary nacelle]

found solution

clear cookies

and localstorage

[storm dust]

done i cleared only that left

[forest sundial]

Hola 👍 desde Cuba

[outer flicker]

are you guy developer?

or coder?

[forest sundial]

me?

[outer flicker]

i mean everyone here

[storm dust]

the site loads now

but i also reported the issue too

[storm dust]

it depends. im just a guy who wanted to join this server just for fun

[forest sundial]

Maybe I don't speak English, but I'm a developer.

[outer flicker]

me just normal well im use for rolepla, coding, at a friend to comfort and help me on task 😄

[forest sundial]

X2, Saludos jaja

[storm dust]

sir i dont know what that means

[outer flicker]

im dunno amazon have ai chat too 🙂

[forest sundial]

Means Greetings bro

Is spanish, I'm from Cuba😆

[outer flicker]

https://app.getmulti.ai/chat?threadId=f3cc440b-cd5e-417e-aa22-cb2eac382fc0 yo what did im found lol 😄

[quiet pewter]

anybody knows how to use opus4.6 for free?

[forest sundial]

I only see up to Sonet-4.6, which is really great, but you set a code task to test and an app is set up in React to explain it to you.

[subtle rose]

<@&1349916362595635286>

[outer flicker]

SYM

[quiet pewter]

what do you mean
is there a app for sonet 4.6?

[icy jay]

@drowsy tusk you got attacked by cookie virus. fix the issue, then reset your passwords

Calm down.

[onyx python]

Same

Something went wrong with this response, please try again.
Trace ID: 4e0edc52-d05f

Using Nano 3.1 for image generation

[icy jay]

try the "try again" button

if you keep getting this error, you are on the limit, come back after 6 hours or something should be good

[onyx python]

The try button returns the same error. I don't think I've reached my daily limit, guess I'll try later

[subtle rose]

Okay, I need a course to solve damn captcha

[outer flicker]

downlaod buster captcha slover

[golden ocean]

claude-opus-4.6-thinking-128k

[vale vortex]

Is this only me or..?

[eternal pawn]

Me too

[pliant vapor]

Hey ,
Anyone can tell me how i can join the video generator chat

[eternal pawn]

Even I have seen that since last night (GMT +7)

[pliant vapor]

Hey ,
Anyone can tell me how i can join the video generator chat

[outer flicker]

mine still fine

[static acorn]

Same here

[eternal pawn]

I think it's regional lag (my opinion)

[outer flicker]

yo guy why im still have 3.1 pro? (late aprils fools)

[vivid sandal]

I don't think so, mine isn't working on my laptop but working on my phone

[outer flicker]

mine not

i mean mine still working

[fading salmon]

i almost had a heart attack

thought chats were deleted or something.

do mods know when this'll be fixed?

[light siren]

the issue is not there for me

[fading salmon]

it could be model-specific or, well, given the userbase of arena, it could possibly be a specific server or two that's not working

[night moat]

Something went wrong with this response, please try again.

Trace ID: cc828d37-404c

Gemini both

[tidal oxide]

Claude opus deleted!!!!

[cedar sparrow]

I had the same problem. You can open it in Incognito mode.

[wooden zodiac]

Is the website unable to use?

[river hamlet]

i see opus 4.6 thinking in batle mod)

[deft spruce]

Personally, I think the Claude series has too small an input token limit. Even 200,000 tokens feels too short

[sullen creek]

opus has 1M context window

[abstract hinge]

@echo aurora 😭

[deft spruce]

..wait what...? but that token says the input token is 200000...

[vernal zodiac]

is there any rough release date for gpt image 2?

i heard it's dropping today

[sullen creek]

yes thats input token limit

output token limit is 1M

[deft spruce]

....oh

[light siren]

have u tried doing what the tiny text at the bottom says

[cobalt hawk]

yo guys can i use arena ai to type prompts unlimited ? or there is rate limiters?

[sullen creek]

ofc rate limits are there

[cobalt hawk]

okay so how much i can tho? more than normal free plan of claude?

[sullen creek]

theres no fixed number, it dynamically adjusts to traffic

[cobalt hawk]

okay thanks!

[sullen creek]

np

[abstract hinge]

Obviously yes. Not only that, I also restarted the browser.

It works in incognito mode, so it seems like some kind of block or a corrupted session.

And it’s not an account-related block, since I used the same account in incognito mode.

And even if it were a temporary block, it should have already passed (I think). I strongly suspect it’s a cache or corrupted session issue. I’ll try clearing it and see if it fixes it.

Cleaning the cache actually worked.

@echo aurora

[sullen creek]

larp

[light sleet]

Might

[obsidian tundra]

Best AI for scripts?

[wary robin]

GET OFF EDGE

[wary robin]

Following or creating?

[abstract hinge]

I had already mentioned that—I even shared the solution that worked for me and my suspicions.

[light siren]

should have saved the trace thing maybe

[obsidian tundra]

creating

[proud bobcat]

I’ve noticed that Gemini has become really bad lately

Hallucinates a ton

I started using DeepSeek because even if it doesn’t have vision it’s more accurate and understands the exact same if you describe the problem

Like it is genuinely insane how good DeepSeek is at math

[golden horizon]

i think gemini3.1 has become less smart recently and will answer some bad content. I miss gpt5.4model.

[wary nacelle]

Error 404 fix (Page was lost)

1. Open DevTools (Ctrl+Shift+I or F12)
2. Open Application
3. clear localstorage
4. clear cookies

I had this issue myself and fixed it this way.

Arena Fixes

Release in ~1 week

or later or sooner

depending on my schedule

Its already ready

just needs a quick website with a guide to install it

and custom builder framework

to put all js files into one

obfuscate

and make zip file to give to ppl

[light sleet]

use arena and write guide yourself

download zip of website and use cloudflare pages😭

Or github

[wary nacelle]

no, i know how to do it

i am gonna do website with vercel and next js

and just find a guide of installing Developer Extensions

since webstore doesnt work

or if i dont find any good ones

I'll make the video myself and edit it

with capcut

but video editting can take me up to 4 days so idk

finding a guide in youtube is the fastest option

making a guide myself is the best option if i had more time

but @light siren is scared of hype to gone out when we fully finish the extension

so yeah i am trying to finish it up real quick

to give a demo

then progressively update it

depending if yall find bugs

or get some suggestions

after release @light siren and me will finish up the LMArena Old Theme feature

add Stared models and ect-

[silent tree]

Not for mobile?

[grave peak]

🇮🇳

https://tenor.com/view/jintian-indian-scammer-ai-amir-meme-gif-9590912363268739995

[topaz sail]

H

[slender ledge]

Yea arena is doing bad things

Why remove opus

Gpt 5.4

And there is so much lag

[high owl]

hi

you hate deleting claude opus series?

https://klipy.com/gifs/case-oh-caseoh-23

[barren burrow]

claude opus

claude opus

[sullen creek]

@olive spruce

[atomic lagoon]

Holyyyy molyyyyyy

https://x.com/DylanTFWang/status/2043952886166761519?s=20

Tengfei Wang (@DylanTFWang)

Genie3 generates videos. We generate 𝟯𝗗 𝘄𝗼𝗿𝗹𝗱𝘀 you can actually use.

Launching tomorrow — Tencent #HYWorld 2.0, an engine-ready World Model🚀

This isn't a video. It's a real 3D scene, all generated & editable. One image in. A whole 3D world out.
🔥Open-source tomorrow

▶ Play video

X · Apr 14, 2026, 7:21 AM

Arena took the VCs away

unfortunate

[outer flicker]

tencent know to cook lol

[rose jackal]

Opus sucks

In arena

Like why would you need opus in arena

[honest whale]

#ai-creations

[abstract hinge]

No need to overcomplicate it—just click the lock icon next to the page URL and you can clear it from there.

[tame osprey]

how to fixt this

[olive spruce]

https://tenor.com/view/shrek-shrek-rizz-rizz-gif-11157824601050747846

[sullen creek]

hi claude opus

[olive spruce]

Sup

[sullen creek]

hows it like being missed by the entire server

[tame osprey]

I don't know if I should be saying this, but can anyone tell me the solution to this problem?

[obsidian cargo]

fingers crossed for gpt-image-2 today

[elfin swan]

Hello someone plis help me i wanna subscribe to arena, because i wanna have unlimited video creations and for now i can only generate 3! plis help me, thanks!

[light sleet]

prob not today

It might be Thursday

but I hope it's today

[echo aurora]

That error message is indicating the prompt problematic for the content filter, you'll want to adjust the prompt until it's okay.

[next trout]

website got ddosed?

[obsidian cargo]

maaan what's taking them so long to roll it out? It works, people have it...

[echo aurora]

There is not a way to subscribe sorry to say. The rate limits are equal for all users.

[echo aurora]

Hope not. What are you seeing?

[next trout]

it works in another browser, i guess i'll just clear my cookies n cache

[echo aurora]

Ah, can you hard refresh the site? This tends to be fixed with a hard refresh.

We have been seeing an increase in these reports the last couple of days. This has been shared with the team.

[next trout]

hard refresh didn't work, but i cleared the cookies it worked

[light sleet]

still testing ig

[echo aurora]

Good to know. Keep me updated if this problem comes back.

[next trout]

ok

[kindred herald]

Are there any sites that can temporarily provide access to Claude Opus, or otherwise, are there alternatives to Opus itself?

[sullen creek]

well for opus, openrouter if u wanna use the api or claude.ai to use opus but u have to pay the 20$

[kindred herald]

Okay, thank you.

[atomic lagoon]

The first one the character looks young, I'd say thats likely why

[old plume]

@echo aurora can you please tell when you will bring opus 4.6

[abstract hinge]

I doubt they have any estimate, but I wouldn’t expect anything before next month.

[spring oar]

who is better sonnet 4.6 or opus 4.6 ?

[echo aurora]

Yeah sorry to say I don't have an ETA to share.

[abstract hinge]

opus

[spring oar]

opus is nerfed rn

[abstract hinge]

Could you enable Sonnet 4.6 Thinking for us? 👀

[echo aurora]

The request for the thinking version of this model has been on our radar. I couldn't say if/when this model would be added.

[golden ocean]

just ask it to think in <thinking></thinking> tags

[user](#message)
my prompt

Also output ur thinking inside <thinking>tags</thinking> before your main response in the chat. Set thinking mode to medium (25).

<thinking>

[sullen creek]

i dont think thats how it work

[golden ocean]

it does lmaoo

claudes hybrid model, it only works with claude

try

[sullen creek]

yeah but sonett 4.6 and 4.6 thinking are different variants

[golden ocean]

no

[sullen creek]

okay ill try

[golden ocean]

if u have claude subscription u can confirm too on claude.ai site

the reasoning is the exact same as real thinking enabled, i can confirm as i have cwaude.ai subscription too

[spring oar]

no bro opus 4.6 is nerfed with thinking

[golden ocean]

even the thinking mode/effort setting, is real
it's in claude code cli's prompt too

[golden ocean]

we're talking about sonnet 4.6, there is no opus on lmarena

https://cdn.discordapp.com/attachments/1323789387946786957/1493267079229476975/togif.gif

[spring oar]

you have opus in leaderboard ?

[golden ocean]

wat

[abstract hinge]

That’s not really the same thing.

Using <thinking></thinking> or asking it to “think step by step” only changes the format of the response—it doesn’t replicate the actual thinking process behind the model.

Real thinking mode involves different decoding, deeper reasoning, and more internal inference. It’s not just about showing steps, it’s about how the model actually processes the problem.

The simulated version can help a bit, but it won’t reach the same level of results, especially on more complex tasks.

[verbal torrent]

Hi

[abstract hinge]

I already knew you were going to say that. And what about making it possible to download the conversation history?

[spring oar]

in llm arena you have opus

Opus 4.5 is better than sonnet 4.6 ?

[abstract hinge]

These two models should be on par.

[echo aurora]

Similar answer sorry to say. Overall, I'm not going to be providing ETAs on when to expect new models/features/etc.

[obsidian cargo]

maybe gpt-image-2 today? 🤞

[light sleet]

I hope, but no announcement today

[obsidian cargo]

no announcement today yet

[light sleet]

Arena will add gpt image 2 after it releases that's a 100% guarantee.

[obsidian cargo]

depends on the pricing, they might put it in battle mode only until they set up their credit system

[light sleet]

most img models aren't too expensive

cuz they would've removed Gemini 3 Nano Banana Pro too thne

but their focus was removing primary smart models in direct chat like gpt 5.4, Gemini 3.1, and Opus

Cuz they're expensive and output takes way more.

[obsidian cargo]

gpt-image-1.5 launched on a tuesday btw

so did gpt-image-1

[knotty fable]

Yes constantly in the last few days, also video fails.

"Something went wrong with this response, please try again."

[obsidian cargo]

huh only been getting image fails since today... except if I don't hit retry it retries twice automatically and works on the third try.

[knotty fable]

And since the failed attempt counts as a generation, it's quite distressing to use this any more.

Better for you then, I get stuck since I get nothing.

[obsidian cargo]

yeah when I mashed retry it used up all my generations, but if I let it sit it works

well I don't really care about the images gpt-image-1.5 makes they kinda suck

I've only been submitting an image prompt every few hours to see if I can get an a/b test or full on access to gpt-image-2

I did get one a/b test with gpt-image-2 but the image it made was a little risque (didn't prompt it that way) so I can't share it :(

[knotty fable]

They do, same for Grok video - which is horribad. 😸

[obsidian cargo]

grok is pretty bad in general imo

[light sleet]

I'm not that excited for spud, I'm excited for gpt image 2 😭

[sullen creek]

use nano banana

its better

[obsidian cargo]

gpt image 2 is better than nano banana actually

[knotty fable]

My friend did show some clips made yesterday made with Grok that was decent though - also NFSW, so I will not forward them here.

[silent tree]

Prove yourself

[sullen creek]

what>

[obsidian cargo]

based on what others have posted + what I got when it was tested in arena

[silent tree]

Prove it's better.

[obsidian cargo]

gpt image 1.5 is trash. nano banana is great. but gpt image 2 is even more great.

[sullen creek]

i dont know whats ur use case, so you would have to use urself and decide if its better or not

[silent tree]

Is Gpt Image 2 better for thumbnails or nano

[knotty fable]

I'm fully with Varkarrus with my opinion on Grok.
And yes Nano banana is good also Seedance 4 & 5

[obsidian cargo]

probably gpt image 2 someone posted some really crazy youtube screenshots genned with gpt image 2 here earlier

[silent tree]

the sam altman selfie was my fav

[obsidian cargo]

#codename-discussion message

(maskingtape alpha was one of gpt image 2's codename)

[silent tree]

wow

Yeah it's definitely better.

[obsidian cargo]

fun gen I got when it was on arena.ai

[knotty fable]

I like the multiple reference option for Seedream and Flux 2.

[crystal mica]

guuys

guys

wait all

[silent tree]

someone in x generated these

[crystal mica]

why we are all being ignored about gemini 3.1 pro and etc

[knotty fable]

....WAN got that also.

[crystal mica]

it is just abnormal

[obsidian cargo]

#announcements message we're not

it's just old news at this point

[crystal mica]

it not an ACTUAL ansver

it is just like

'we have some problems we do delete we do changes blah blah blah'

[obsidian cargo]

they're working on a free credits system to allow premium models. pineapple has stated repeatedly there is no ETA yet but its coming.

[crystal mica]

love your pfp btw

[echo aurora]

What can I help clear up? We mentioned in the announcement:

These changes are part of our efforts to ensure that we can continue offering access to AI models while keeping the platform running reliably.

[knotty fable]

Hey Pineapple? Could you get a fix so that failed generations do not count for the daily allowance, since we only get 2 nowadays - it's a tad dissatifying not to see any results - which is the current state of affairs.

[crystal mica]

This isn't a response, it's a formal reply. A response is a full description of how exactly removing certain models helps with something.

For example, you could provide a 200-page report on how a new metro station in the city helps relieve congestion in crowded areas.

Where is your 200-page report on how exactly removing these models improved platform availability and security by so many percentages or metrics, and how exactly?

lol

[uneven cypress]

Guys will arena.ai put Claude opus and all the latest models back??

@everyone

[obsidian cargo]

lmao the attempted everyone ping

yeah they're coming back but with the free credit system

[wary nacelle]

oh-

[uneven cypress]

Do u have any idea how to use Claude opus for free

[obsidian cargo]

the only way to use it for free right now is to gamble on arena.ai's battle mode.

[uneven cypress]

Ok ty

[obsidian cargo]

claude opus is very expensive you'd be hard pressed to find anything that gives it away for free

[uneven cypress]

Yea ik

[obsidian cargo]

that arena.ai was able to provide it for free at all was pretty wild.

[wary nacelle]

dude whaaaa-

I will add it as a feature to Arena Fixes extension

[knotty fable]

On the subject of testing prompts.

And here is a nice one.

Same prompt used for both.

But then again, one actually need to adjust the prompt to the syntax used for each engine.
And that is a shortcoming for Arena, since we do not know beforehand which one will be used in the geneation.

[wary nacelle]

oh..

still a chance for me to make it a markdown

[echo aurora]

This is the intent, but want to mention this isn't yet confirmed

[obsidian cargo]

ah dang

[echo aurora]

Since we're providing free access to AI there are restrictions and limitations that are put in place with the intent of ensuring long term success and availability of the platform. I don't think providing a full report with more context would make that reasoning more clear.

[crystal mica]

I won't argue with you because I know you'll respond in a corporate and empty manner.

[knotty fable]

Yah, a lot of the functionality is missing - and no worries. In my case I use Arena to test prompts, such as for the situation I got now where I don't seem to get an air taxi to land properly.
So I send one test here, and also to two others to see what happens.
And if I get a match on Arena - with one of those I have available on the site I pay for = bingo and I can do it with that prompt. 😺

My only complaint is how slow it is, and most often end up with a fail.

(This said as I'm still waiting for the generation for today while warbling here.)

[wary nacelle]

its literally his job-

[vernal raft]

A response is a full description of how exactly removing certain models helps with something.

I don't think there's the need of a report to imagine a huge amount of ppl opening a direct chat with gpt 5.4 high and using only that. Or a side to side with opus and 5.4 high

5.4 high is what? 125€ per m output?

Also, assuming in the first place he can share such details in his own initiative pretty wild imo

[toxic verge]

What happened to cheaper pieces 😂😂

[light sleet]

I think GPT image 2 at Thursday

[knotty fable]

That 125 moneis for a single run is insane, my budget for 2 image / video generators is currently 20€ a month, the total budget for my entire film will end up at around 200€ - my worktime not included OFC.

[obsidian cargo]

maybe. but gpt-image-1 and gpt-image-1.5 both came out on tuesdays

[toxic verge]

They are canceling Sora on 26th

[knotty fable]

The Sora app right? Not the model as I have understood it.

[toxic verge]

Not sure

Never mind you’re right they’re canceling the API in September

Damn, what a shame that was a good model

[knotty fable]

Ahh ok the API will go also then.

[knotty fable]

Inferior in my opinion, it's offered on the site I pay for. But I have not gotten a single good generation out of it.
I had one scene with dancing robots that had a bit of good movement, but overall not useful.

[subtle siren]

guys could someone help how can i make longer videos using this site? like i tried but it always makes me 5-10 sec videos

[echo aurora]

Sorry to say that is the time limit for Video Arena. There isn't a way to extend this.

[knotty fable]

That's what the generators provide, you have to join the clips together.

[subtle siren]

oh thats bad ):

[knotty fable]

And with what you get from here, you will not get far anyway - as you now only get 2 tries each day.

[hollow ivy]

that's a good question, several candidates exist now:

• Muse Spark
• Deepseek (seems to be continuously updated now)
• GLM 5.1
• MiniMax 2.7
• Kimi 2.6 (still in beta?)
• Qwen 3.6(+)
• Grok (update exists since weeks now)

Which better: Muse Spark, Deepseek or GLM?

[echo aurora]

Pareto chart is helpful for this as well: https://arena.ai/leaderboard/text?viewBy=plot

[knotty fable]

And with the slow generation, I'm afraid Arena do not matter much to me either - due to the slow generation.
The scene already done now.

[knotty fable]

If you don't mind a DM I could tell you how to make continous film.

Or I go for my evening shower, it's almost 22 here now.

[neon timber]

Hello

[hollow ivy]

oh, and it totally forgot about MiMo

[toxic verge]

You should practice the rule of thirds in your generations

[obsidian cargo]

!!!

[light sleet]

Also generated by gpt image 2

LOL

😭

[light sleet]

It's generated by gpt image 2

[toxic verge]

What’s the difference between 1.5 and two

[light sleet]

Found on X

[light sleet]

1.5 bad
2 good

2 is VERY realistic