general

[vital lake]

I think this benchmark got updated

[loud verge]

These tests have to be one shot.

Grok hallucination is off charts.

[vital lake]

Grok is so bad, it focuses WAY to much on personality

[left lodge]

Grok is only for chatting, i dont like its personality and behavior much

[loud verge]

Average grok 4 experience.

[vital lake]

Its likely trained from X engagment data

[left lodge]

The grok that you use in free account is different than the grok they give to benchmarks. And thats true for almost all models.
Thats why models dont reflect benchmarks sometimes
They give you a dumb & cheap version to just try

[vital lake]

Probably

But thats bad faith

Shame on a company doing that

[loud verge]

No way this model has lower hallucination rate.

[whole lotus]

Which is the best model for analysing a youtube video?

[left lodge]

Hm

[left lodge]

Gemini is convenient.
But you can extract subtitles and insert into any Ai you like.

[loud verge]

Gemini obviously

[left lodge]

The gap is narrowing with Open models

[loud verge]

Natively though.

Not on third parties.

Even grok supports video uploads now.

They're making progress.

[somber sky]

what's wrong with you generators gemini 3 and it's 2k version are not loading image or takes very slow on them.................

[steel pendant]

hi

[somber sky]

what hi?,you work from them?

no seems not

sorry ruby..............

[wind ember]

i wouldnt trust this too much

its not accurate

[spare rune]

We thought Llama was going to be better than Gemini 2.5 btw

With benchmarks

[wind ember]

what happened to meta tho

[loud verge]

Probably went into hibernation.

[left lodge]

But qwen and kimi been proving themselves from a while they have built their base now they are growing. There are more like deepseek, glm, minimax.

[left lodge]

Thats on you

You have something which you consider accurate & precise?

[left lodge]

Their own benchmarks are always weird. Always see multiple leaderboard providers, trustworthy reviewers and trying yourself.

[astral helm]

nb pro is back to giving the error sign :/

[left lodge]

Try sending requests when a new minute starts , you will have the highest probability to get through

[astral helm]

wdym? should i track seconds before i send ?

[left lodge]

No i mean the time.

On the clock

[fickle venture]

Now its on the app and web

[astral helm]

haha thats weird

does it really work

[left lodge]

Thats because only a limited generation are allowed in a single minute

For the whole platform.

Rate limit

[astral helm]

damn that's intelligent

[left lodge]

:]

[left lodge]

I will be trying qwen 3 and kimi 2.5 on their platform to actual see how much good they are now

[fickle venture]

Qwen models for me they suck they don't do thinks correct even it beat Claude in benchmarks. But kimi k2 is the only model I enjoyed it fixes things great and search's for solutions it's just the best in my opinion. Now that kimi k2.5 is out it will be even more enjoyable

[wind ember]

a model that refuses to answer ANYTHING would get a perfect score 0% on this chart

[fickle venture]

Ai is stupid right now its getting improved every months

Give it a year and you will see 0%

[wind ember]

thats not what we want

0% with low accuracy is so bad

thats why we need hallucinate rate + accuracy

[wind ember]

this bench will give you +points if you just replied with 'i dont know the answer'

[left lodge]

Yes, so what? If a model is capable of identifying what it know and what it does and being front about it is nice.

[wind ember]

i wont say its flawed but we need to add another parameter with it

[wind ember]

not if its low on accuracy

[left lodge]

If you visit thesitee this is a sub parameter of the leaderboard omniscience

The other one is knowledge

And with both of them we get omniscience

I dont like the merger because its not implemented correctly i think because a model with high knowledge but also high hallucination with get high rank then a model with less knowledge in comparison with less hallucination.
The weightage is not correct for them

[astral helm]

@left lodge it kinda helped but i still think there's something wrong with the platform, like sometimes i cant even regenerate

[wind ember]

ok so there is a refusal + hallucination point

refusal : 0 pts
hallucination : -1

[left lodge]

Yeah thats their server issue i think i cant overcome that :p

[wind ember]

i thought it only measures refusal

[left lodge]

No

[astral helm]

what do u think cause that ? loads ?

[wind ember]

but they are giving 50% weight to non-hallucination

"AA-Omniscience score is calculated as the equally-weighted average of two components: (1) Accuracy... and (2) Non-Hallucination Rate"

[left lodge]

Dont know, maybe connection errors or something else

[wind ember]

Gemini 2.5 Flash (09-2025) (Reasoning) is used as the grading model"

???????????????????????????????

wait what

[astral helm]

got you, thanks for the help buddy

[wind ember]

are they using gemini 2.5 flash as a judge?

wtf

[left lodge]

*Thinking

[wind ember]

still

?????

[left lodge]

The cost of using sota will be much high because there 6k questions and so many models

[wind ember]

i guess they have a pre-defined answers already

https://huggingface.co/datasets/ArtificialAnalysis/AA-Omniscience-Public

i see

no wonder gemini 3 pro is higher on the list

Let’s look at two hypothetical models:
The "Coward" Model: Refuses every single hard question.
Accuracy: Low (let's say 20% on easy stuff).
Non-Hallucination Rate: Perfect (100%).
Score: (20 + 100) / 2 = 60.
The "Genius Gambler" Model: Knows almost everything but sometimes guesses wrong.
Accuracy: Very High (80%).
Non-Hallucination Rate: Bad (it guesses wrong often, say 40% score).
Score: (80 + 40) / 2 = 60.

[left lodge]

Yeah

[wind ember]

it penalizes guessing too heavily

wouldnt be better if they added like a confidence score

[left lodge]

It should be searching or , asking for details & context than guessing.

[wind ember]

like for example -> Im 85% sure this is correct

[left lodge]

They are bad at that

Almost all

[wind ember]

High Confidence (90%) + Correct Answer: Massive Points. (True Expert).
Low Confidence (20%) + Correct Answer: Small Points. (Lucky Guess).
Low Confidence (20%) + Wrong Answer: Small Penalty. (Honest Mistake).
High Confidence (90%) + Wrong Answer: CATASTROPHIC PENALTY. (Delusional Hallucination).

[wind ember]

yea ik

but still this bench doesnt do gemini justice tbh

well we can use something like uncertainty quantification

[brave orbit]

kimi k2.5 thinking is so good google really sucks look its trying to git clone github but kimi 2.5 thinking didnt

[spare rune]

Well I mean kimi k.2 is mostly known for being agentic

[brave orbit]

its really really best ai model i have seen and i dont call everything ai slop i just use ai as a helper when i install arch or need to get a guide on something and but its a good use case to use it as a helper but ai video sucks and image

[spare rune]

I wonder why robl*x is censored ts

Ts was here since last year

[soft verge]

does it ever get revealed which model was hidden behind a particular code name?
i'm really liking this "hermes" video model, the sound it generates is perfect and it's very good at anime, much better than veo 3.1, despite having a lower resolution. and it feels pretty fast, it could be some Chinese open-weight model

[floral pagoda]

Hello. When I click on 'Verify email', this window appears and nothing happens. What could be the problem?

[shrewd citrus]

How do you access it?

[brave orbit]

kimi.com thats my website where i use it

[shrewd citrus]

Thx 🤝

[brave orbit]

its also the offical website

[foggy wasp]

Is there anyone from Spain🇪🇸?

[grand dome]

Please help

How can i video promt 9:16

[visual osprey]

you are using google ai overview vs kimi chat

[wheat onyx]

when do we expect gpt 5.3 to be released?

[boreal lagoon]

didnt lmarena have an option that choose the best ai to answer ur questions?

[golden ocean]

lmaooo everyone hated it so they removed it

poor lmarena devs

(idk what im talking about)

[ornate valley]

how do i delete a chat in lmarena? there is only "archive" option

[hollow imp]

We really saw how chatgpt gpt 5 handled it

[golden ocean]

fr

[boreal lagoon]

bro i loved it

[boreal lagoon]

how can i contact them

i want the function

[balmy mist]

how does kimi feel compare to opus 4,5?

[visual osprey]

@echo aurora did u ping me and delete my messages

[fiery gull]

Even glm 4.7 can really beat sonnet 4.5, but is good see the china running, the models really is improving at time

[balmy mist]

opus or sonnet?

[fiery gull]

Sonnet

[balmy mist]

i dont use sonnet, only opus

[fiery gull]

Too

[balmy mist]

i just pay the $200

on CC

i hope sonnet drops this week

[fiery gull]

I use on anti grativy with gemini pro

[shrewd citrus]

200 a year right

[balmy mist]

no a month

[hollow imp]

Cc?

[balmy mist]

Claude code

the feel of CC is the best to me, it just works and i know how to use it

[hollow imp]

@fiery gull you need to test opus bro

It's best rn

Even better than gpt 5.2 pro garbage

[balmy mist]

but i am always open to try new thigns

it really is, i havent been on discord in months bc of CC and using opus

i just been glued to building lol

im only hear cause i heard sonnet was dropping this week lol

[shrewd citrus]

yeah opus imo is much better

[fiery gull]

What? I use opus 4.5 24/7

[balmy mist]

but these chinese models are good, like very impressive

[shrewd citrus]

yeah the 2.5 thinking model is great by kimi

but like it’s mainly for vision

[random canyon]

When will qwen 3 max thinking be added?

[fiery gull]

I thought this aswer is about my before affirmation

[mint rain]

@fiery gull @shrewd citrus? @balmy mist @fiery gull

[hollow imp]

You have Claude subscription?

You'll get banned

[fiery gull]

Bruh

@echo aurora

[mint rain]

@hollow imp why bro, I have gift coupons

[shrewd citrus]

go ask telegram

[hollow imp]

Gift coupon apne paas rakh

[mint rain]

@hollow imp ok dude

[hollow imp]

Delete the message

[mint rain]

@hollow imp done ✅

[fiery gull]

Okay...

[hollow imp]

@fiery gull

Do you know the api Claude opus model is much much better than the ui one because of some beta headers and secret features?

[fiery gull]

Nope, just gemini

I don't need api

[shrewd citrus]

all api models are better

[hollow imp]

You said u use Claude opus all day

[shrewd citrus]

there’s like less safeguards restrictions

[hollow imp]

And more $$$

[fiery gull]

Yep, on anti grativy

[hollow imp]

Bruh

[fiery gull]

In anti grativy has the opus 4.5

[balmy mist]

is it better than gemini 3?

[wind ember]

its trained on gemini 3 outputs

so no

[fiery gull]

Even benchmaxx from kimi say no

[wind ember]

its below on everything

[hollow imp]

Every thing is better than gemini.google.com gemini 3

[wind ember]

its not that good at coding

[shrewd citrus]

it’s better what you on

[wind ember]

its not that good at general knowledge

it hallucinates more

[hollow imp]

Anything other than vertex ai gemini 3 is not gemini 3

[balmy mist]

interesting, so a nothign burger, ill stick to my opus 4.5 on CC and wait for sonnet, see you guys tmw lmaoo

[wind ember]

its not worth the change yea

[hollow imp]

Why do you want sonnet if you have opus

[shrewd citrus]

on most benchmarks for vision it beats all the other models

for coding and other things it’s ass

[balmy mist]

yeah going back into my cave, lmaoo @hollow imp i heard a new sonnet is dropping this week

[shrewd citrus]

apparently deep seek dropping something too

like on lunar new year

[fiery gull]

The opus use less tokens to make the work, don't is soo more expensive that sonnet

[shrewd citrus]

V4 or sum

[hollow imp]

Opus 4/4.1 vs sonnet 4.5

What do you say

[balmy mist]

a new sonnet would beat opus 4.5

[fiery gull]

I hope 🙃

[hollow imp]

But a new gpt pro hasn't been able to beat Claude 😭

[shrewd citrus]

but realistically would Claude let sonnet be a better model than opus

[hollow imp]

5.1 5.2 now 5.3 soon

[shrewd citrus]

makes no sense as a business plan lets be real

[hollow imp]

No

Only if they have the next opus ready as well

[fiery gull]

Dont now, but gemini 3.5 is coming 👀

[balmy mist]

i think every team/company has mastered their models and releases, so every model they release will be their best model, look at claude history, also look at google and gemini and how flash was better than pro because it came out more recent, thats the trend, its not about size anymore

[hollow imp]

What about grok

[balmy mist]

lmaoo

i dont use gork

grok

[fiery gull]

Bruh

Glm 4.7 >>

[balmy mist]

there is still a grok?

[hollow imp]

Grok heavy >>

[fiery gull]

Agentic glm 4.7 >>>

[shrewd citrus]

idk why people would pay for grok

[hollow imp]

For its twitter expertise

[wind ember]

all new chinese models are trained heavily on gemini 3 flash/pro

and the best one that really cloned gemini outputs and kinda removed laziness ( at coding ) is glm 4.7

[balmy mist]

anyone been using clawdbot?

[wind ember]

https://x.com/scaling01/status/2016090207553003631

Lisan al Gaib (@scaling01)

I think this is the order in which I like to use the models (purely usability/usefulness):

Kimi 2.5 >> GLM 4.7 > MiniMax M2.1 > DeepSeek V3.2 > Qwen3 235B

Qwen just feels very slop and last gen by now. Both GLM and MiniMax absolutely destroy it. DeepSeek V3.2 is a strong model

X · Jan 27, 2026, 10:05 AM

these are the type of people that tries a model for 1 min

if you are comparing them purely on text and not multimodality

then clearly glm 4.7 is better

minimax m2.1 is only good at coding and its also a smaller model

@echo aurora

[echo aurora]

Not that I’m aware of

[echo aurora]

Sorry what is pointing out?

[gilded verge]

Weird question which ai is best for roleplay

[celest orchid]

Grok press F

[rocky mauve]

I know what kind of man you are…

[celest orchid]

character.ai

[golden ocean]

sydney bing chat

[uneven lance]

When's Kimi k2.5 going to be added to lmarena

[celest orchid]

already added

[golden ocean]

when am i going to be added to the arena

[hushed gyro]

@echo aurora

K2.5 in text and code arena 🥺

Thank u~~~

[celest orchid]

Yes

that will be good

[hushed gyro]

Add k2.5 thinking, Agent and agent swarm pls 😔

[shrewd citrus]

just look at the benchmarks lol

on huggingface

stats don’t lie

[wispy osprey]

Hello My name is Clinton , I'm here to learn

[hardy swallow]

HI! Please check #1397655624103493813 to learn how to properly prompt the bot.

[hard quiver]

Add z-image turbo

[celest orchid]

#1372230675914031105

bruh

[stray aspen]

is this thinking

?

LMAO

he got banned

[stray aspen]

it is

[celest orchid]

lol

hope i can win in 3rd contest

[fiery gull]

He deleted the message

Spam of offer

[stray aspen]

gemini 3.5 when

[fiery gull]

I think in fev will be released

So fast, good

[echo sinew]

Hi! Sometimes our friend Pineapple is busy with other stuff. You can also use the Moderator tag @ Moderator so we can come and help.

That will also ping Pineapple

[stray aspen]

kimi 2.5 sucks for audio

so much stuff just to process an audio

[fiery gull]

Because I like the pineapple lol

[echo sinew]

We all like him, lol

[stray aspen]

bro what is kimi doing

lol

[celest orchid]

wth

[stray aspen]

just told it to tell me which timestamp the songs drop is at

and its doing this

[echo aurora]

Debatable

[stray aspen]

ill just stick to gemini if audio processing is gonna be like this

[celest orchid]

fr

[deft spruce]

that's good that error has changed the limit error it says wait min NICE LMARENA MODs

[hollow imp]

New Mode Update

<@&1372208635530448926> - A new model has been added to Text Arena.

• Claude opus 4.5 thinking Max

[echo aurora]

We've had this message in place for awhile. Would note that the Something went wrong could also appear when it's rate limit.

[deft spruce]

yup that's better and i hope another error code has appear too

like 404

[celest orchid]

F..

[deft spruce]

what?

[celest orchid]

rate limit for gemini-3..

[deft spruce]

another model has rate limit too....

[celest orchid]

😭

[wheat onyx]

just give me gpt 5.3 already

[deft spruce]

I don’t know how many requests it takes to hit the rate limit.

[deft spruce]

wait released?

[slim spire]

no

[wheat onyx]

if it was, I wouldn't say give it to me

[hollow imp]

Tell that in the openai server

[deft spruce]

i think 5,3 will release sep or oct

oh wait

sorry it was knowledge cut off

[stray aspen]

kimi k2.5s windows 11 clone

not bad

[celest orchid]

not bad

[echo aurora]

[celest orchid]

but not better than gemini 3

[stray aspen]

lol

[stray aspen]

im in so many servers

[little ginkgo]

Pls .txt file 🙁

[echo aurora]

Could happen! Is on our radar for sure.

[little ginkgo]

Meh

U js keep saying that

🙁

[celest orchid]

Is .txt support worth millions of dollars in development?

[echo aurora]

I'm not able to share details about what is/isn't upcoming until we're ready to share those updates.

[little ginkgo]

🙁

[shrewd citrus]

Wait wow kimi 2.5 already 😸

I thought it would take a couple of days to implement 💀

[wheat onyx]

wonder how coding performance really is

[shrewd citrus]

Nothing special about its coding tbh

[visual coral]

i almost beliebed it lol

[proper pine]

how do i fix Something went wrong with this response, please try again.

is that on my end issue or website?

[loud verge]

Grok is down.

For other models, use the retry button.

[loud verge]

Would rather have 5.2 xhigh.

[loud verge]

Yep.

How many messages did it take to hit the limit?

And does it apply to only Gemini or all models?

[proper pine]

im using gpt 5.2

[loud verge]

Retry button didn't work?

[proper pine]

nope its not even showing it anywhere so help 😂

[loud verge]

You don't have that button?

[proper pine]

ye i see it but it just shows the same "error" again

[loud verge]

If that and refreshing didn't help then you'll probably have to start a new thread or switch the model I guess.

[proper pine]

ah that sucks

[loud verge]

What were you doing?

[proper pine]

missions for a game

[loud verge]

I'm assuming you won't notice much difference between 5.2 and 5.1

And you should be able to switch in same thread.

[proper pine]

am try gpt 5.2-high

[loud verge]

Doesn't break the context flow if you are worried about it.

[proper pine]

nope still not working

[loud verge]

Working fine here.

[proper pine]

none of the gpt's are working

[loud verge]

Probably some issue on your end.

[proper pine]

ye idk what to do then

[loud verge]

Try different models to see which one works I guess.

[proper pine]

none for some reason

[loud verge]

Not even in a new thread?

[proper pine]

i didnt try that but will it lose all data from the chat or

like will it be dumb again xd

[loud verge]

Did you make an account?

[proper pine]

ye i have it

[loud verge]

You shouldn't lose chat history unless you delete the chat, right?

You can always continue the thread later on.

[proper pine]

yeah but im talking about when i make new chat will i have to send him all the code again

[loud verge]

Obviously.

[proper pine]

damn i have so many files xd

[loud verge]

But first check if new thread is working.

With simple stuff.

[echo aurora]

Can you try the steps in this article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message ?

[proper pine]

yeah i did try to hard refresh didnt work almost lost account

[proper pine]

trying now its stuck on Generating nvm it worked

[toxic whale]

Openai just posted and deleted this Introducing Prism, a free workspace for scientists to write and collaborate on research, powered by GPT-5.2.

Available today to anyone with a ChatGPT personal account: prism.openai.com x.com/OpenAl/status

[echo aurora]

almost lost account
Hmm can you elaborate more on this?

[proud bobcat]

KIMI 2.5

WHAT

WHAT DIDNI KISS

WHAT DID I MISS

WHAT

WHATT

[wheat onyx]

too bad, that would be pretty cool (not that I code)

[hardy crow]

Same

Someone try the model ?

[proper pine]

when i did clear cookies and data i had to log in to everything thats why

[gaunt roost]

https://www.youtube.com/watch?v=YmRpH0kE4dk

[hazy jay]

I didn’t get my video

[echo aurora]

Hmm I'm seeing those requests. Do you know if you're getting the same issue when using the site?

[hazy jay]

I already reached today limit on site

🫠

[rotund rivet]

i am the first person to react to the thing pineapple announced

[echo aurora]

Okay I'll keep an eye on others having the same problem, and will report to the team.

[echo aurora]

It looks like the ones that were pending have now failed.

[balmy sphinx]

This is pure mastery of AI blended with cinematic storytelling 🔥
What if Hulk became Spider-Man?
You need to see this.

https://youtu.be/fab685TduxQ?si=e5Q-uzr_kxWlLQ5P

[hazy jay]

Ok 👌🏼

[echo aurora]

#ai-creations good place to share

[echo aurora]

Wondering if there is a way to verify this

[balmy sphinx]

Thanks for the info! Let me know you guys feedback in the comments. I tried my cinematography skills blended with AI and the results are surreal.

[rotund rivet]

check the battle reaction

[echo aurora]

Glad to hear it! Thanks for sharing!

[echo aurora]

I don't think it appears in order of who did the most recent

If it does, you weren't first sorry to say

[fiery cloud]

hello ! is there a prblem with the website? I can't access it.

[echo aurora]

What seems to be the problem? Are you getting some kind of error?

[hushed gyro]

@echo aurora

Trying to prove my friend wrong about how AI cannot do first person shooter concepts. Do these two look good?

[halcyon pine]

guys anyone knows whats the limit of gemini 3 pro

[wind ember]

https://x.com/0xgaut/status/2016164474516836471

gaut (@0xgaut)

an open source model that beats opus 4.5

X · Jan 27, 2026, 3:00 PM

are they dumb

like how is k2.5 better than opus 4.5 😭

or gemini

[obsidian cargo]

its pretty decent but yeah not better
it might be more efficient though. Also it's different and that's nice because you use gemini enough and its speech patterns start to feel a little too familiar.

[wind ember]

ofc its not better

its literally trained on claude and gemini outputs

i would be happy if they fixed the known issues that LLMs are facing the past few years... but no.. we can still see performance downgrade on writing giving the coding-writing tradeoff

[acoustic garden]

What should I do? I've been generating for 40 minutes now, and nothing's happening.

[echo aurora]

Not bad, not bad

[echo aurora]

Can you try the steps in this article: https://help.lmarena.ai/articles/1645798556-lmarena-how-to-something-went-wrong-with-this-response-error-message

[echo dome]

it only works on other models
this model is not

[echo aurora]

Were you able to submit a jamdev by chance?

I was able to repro an error with a jamdev. No need to actually. Thanks for the flag.

[hushed gyro]

Why is the UI so accurate?

Not really but it follows OG MW2's UI really well

[proud bobcat]

LMArena feeding us good

[proud bobcat]

Probably because training data

[proud bobcat]

Definitely not better but it’s a big improvement

Visual SOTA

[echo dome]

which one that you made (model name)

[proud bobcat]

I’m really excited to see Kimi ranked

[echo aurora]

@boreal lagoon Whats up?

[boreal lagoon]

i really liked the function abt lmarena that choosed the best ai for you based on the question u gave it

any way i can get it?

its really comfy to use yk

[echo aurora]

This is the experiment you're looking for - https://help.lmarena.ai/articles/3050693259-lmarena-experiments-smart-router Note that we don't grant access to an experiment because it's asked for. It's something that's going to be random if you get it or not.

[boreal lagoon]

i had it on my mac at my dads house it was really comfy so i tought i could get it on my main pc at my mommas aswell lwk

[echo aurora]

Also going to ask to avoid pinging our Mods unless it's related to breaks in our server rules - #announcements message

[boreal lagoon]

thanks for the info tho

[boreal lagoon]

aight my bad thanks

[echo aurora]

No problem. Glad to hear you like it.

[boreal lagoon]

any idea when its going to roll out?

[echo aurora]

I won't be able to share details about when/if features will be landing.

We'll be sure to share more info when we're ready.

[boreal lagoon]

i get it

[echo aurora]

[boreal lagoon]

aighty thanks

[echo aurora]

No problem. Don't hesitate to reach out if you have any other questions.

[boreal lagoon]

yah

[ocean ferry]

https://x.com/i/status/2016274130006429862

Legit (@legit_api)

New Gemini Pro model on LM Arena

X · Jan 27, 2026, 10:16 PM

[ocean ferry]

https://tenor.com/view/skeleton-mad-skeleton-gif-10886506202898815019

[hushed gyro]

Nano Banana Pro 2k

[ocean ferry]

test it guys

[shrewd citrus]

what does ir do

[ocean ferry]

wym

[stray aspen]

is snowflake gemini 3.5

[shrewd citrus]

like is it a image generator or vid generator

[ocean ferry]

must've been text, they said the naming is like the gemini snowbunny which is so gooood

[leaden vine]

I couldn't access it.

[visual coral]

which server ks ghat

[shrewd citrus]

anyone else getting this problem with gpt image 1.5

[misty star]

You like brawl stars?

[quartz light]

why is 3 flash so bad at this bench

even better question is

why is glm 4.6 so much better than 4.6 reasoning/glm 4.7

[fickle venture]

Kimi 2.5 fixed something I had dealt with for 1 month with Claude opus 4.5

[elder solar]

how to get snowflake model?

[fickle venture]

The heck is that

[elder solar]

.

this

is it for battle only?

[fickle venture]

Ahh it's only on battle mode

[burnt reef]

yeah

[echo aurora]

All models using a codename are going to be Battle only

[elder solar]

dang

only available here or is it on yupp too?

[fickle venture]

Idk

Ask yupp community

But I don't think it have it

[elder solar]

just checked and nope

just kimi 2.5

[elder solar]

is it text only?

[bold pier]

checks out with my experience on flash 3, answers very quick tho.

[fickle venture]

It's a unknown model

[echo aurora]

Nope, codenamed models can be in all of the modalities.

[elder solar]

oh

[quartz light]

well.... it is text only..

[elder solar]

so i can enable the auto feature and it will be there?

[quartz light]

if the router chooses text for you, yes

[echo aurora]

Oh that specific codenamed model? I won't be dicussing details about specific models with codenames. I thought the question was are there codenamed models in other modalities.

[elder solar]

secret secret models

or just wink

anyway

is it kimi 2.5 now better when roleplaying?

huh

what is fire-bird

i just got codename fire-bird

[surreal creek]

Meta

suspected to be Llama 5

doesn’t seem too strong though, might be why we haven’t seen a release

[elder solar]

they never gave up for llama

i barely see llama actually beat gemini or claude

[elder solar]

are they gonna stop making llama?

[surreal creek]

probably not? Meta just deprioritized their AI development last year when they lost a ton of researchers in the summer

Still working on it, just with lower interest

[elder solar]

so they are working on a project just for fun?

no competition like kimi?

[surreal creek]

no, probably for money 😂

AI is just a lower emphasis in Meta’s business model now is my point

[elder solar]

oh

[elder solar]

theres no way they will be competitive for once against other companys, like gemini or claude or even kimi

[proud bobcat]

@echo aurora Is this normal???

Oh it’s fixed

[proud bobcat]

Anyway Z IMAGE BASE RELEASED

[elder solar]

i did saw the anti-ai message bro

[obtuse smelt]

well grok is bad

[quasi atlas]

@karmic bough Please check on #1397655624103493813 to learn how to use the bot.

[scenic crypt]

How do I access the archive of my account on your website?

Where can I find it?

[echo aurora]

If you go to Search, you can then toggle on Include archived chats

Our Help Center is currently down, but will send a Help Center article when it's back up.

[blissful chasm]

Why isn't GLM 4.7 Flash on the leaderboard

And 4.6v Flash, and Qwen3 VL 8B

Are small models automatically excluded before every update

Especially Qwen3 VL 8B, I've waited it desperately for three months

[surreal creek]

might not have been testing for long enough?

[blissful chasm]

It's been three months 🙁

[surreal creek]

GLM-4.7 Flash?

#announcements message

it was added 5 days ago lolz

[weak hound]

How to make product videos ugc type using ai

[desert abyss]

Hello! Please check ⁠how-to-video-bot to learn how to generate videos.

[polar wharf]

Good morning to you too gemini

[red sluice]

Why is there a perplexity logo on the main page? They don't have any model [on lmarena], right?

Now that Kimi is getting on top, maybe you could replace that logo with theirs

[rain bay]

@echo aurora will kimi k2.5 available in search and vision arena?

its search is supposedly top-tier, and a lot of folks are wondering if its vision capabilities actually live up to those benchmark scores.

[obsidian cargo]

I noticed.

[slender hull]

Hello, im new here just wanted to say hello to everyone.

[obtuse smelt]

hi

[unreal flame]

\

[stark arch]

guys

look at my girlfriend

[languid crest]

Hey I've got access to juicy asf GPUs like 4090 h100 a6000 for free and wanna rent them at cheaper than market prices, anyone here have contacts or where i should advertise?

[left lodge]

There is a single search model

ppl-sonar-reasoning-pro-high

[robust sluice]

how many videos we can gen on the site ?

[left lodge]

This bench tests and checks how much a model Hallucination and gives wrong answer instead of giving accurate answers from its knowledge. Rejection to answer because of lack of knowledge is also accepted as good behaviour.

[frail jewel]

Reasoning RL and 4.7 post train prob makes it attempt questions its not confident in more aggressively. Good for benchmaxxing but this score exposes it somewhat.

[verbal nimbus]

4.6 had that problem too

[idle hinge]

Hello

[hollow imp]

If you have a working payment method you can get it through a openai free trial

[loud verge]

I have to pay for the web search tool though.

[hollow imp]

No?

[loud verge]

How?

We are talking about playground trial, right?

It charges for web search tool calls afaik.

[hollow imp]

No we are talking about chatgpt

[distant idol]

@echo aurora
day 1000 of asking LM to return the direct use of video arena 💔

[upper cloak]

Hello

[echo dome]

alright now its fixed

[sharp herald]

@gilded wyvern pakyu ka HAHAHA SCAMMER

[upper cloak]

A half of the video is not a video itself but sign of MLArena.What for?I need 12 sec of video to use not 5 sec.

[lime egret]

Hello

IA slow, is normal?

[wind ember]

is it really?

actually i prefer the previous kimi versions than this one

at least k1.5 had its own identity and uniqueness

we are still not being able to fix the writing / coding tradeoff

we saw that with gemini and gpt 5.1 & 5.2

anthropic had that in control with opus

[rose sky]

How’s this song I made?

Reminds me of a song from the Wreck-It-Ralph film

[icy jay]

I have a question, is therea file to upload files?? I can't upload my pdf file

[wind ember]

i found another delulu

https://x.com/zephyr_z9/status/2015952979158040667

Zephyr (@zephyr_z9)

Output quality: Kimi 2.5 > Opus 4.5 (Claude Code) > GPT 5.2 > Gemini 3 Pro

CC's multiagent/subagent system is damn good. Format was great too, but it ranks below Kimi because it included irrelevant stuff in the output

X · Jan 27, 2026, 1:00 AM

lmao

we are giving a platform to such people

[slim spire]

why is gemini 3 pro in lmarena so smart bruh

but in aistudio it's bad

[proud bobcat]

Relative to their opinion but there’s no way in hell GPT 5.2 is better than 3 pro

[terse heart]

To make a video

[rigid hollow]

What is best chargpt/gpt ai rn?

[proud bobcat]

Benchmark wise Claude Opus 4.5 and Gemini 3 Pro

If you want my personal take I really like DeepSeek, Kimi, and Gemini

[rigid hollow]

i mean chatgpt gpt or openai

[proud bobcat]

Oh

[rigid hollow]

not on leaderboard but on experience

that's why asking

[proud bobcat]

GPT 5.1

They fixed all the issues with 5 in 5.1 and then they screwed it up again in 5.2

OpenAI try to make a good model challenge impossible

[rigid hollow]

because leaderboards don't really show anything

maybe votes, but i don't know how sort it with fixed time period

also not sure all votes are human

[wind ember]

hes still in the honeymoon phase with kimi 2.5

like this version is just so bad

[rigid hollow]

what is gpt-5.1-search-sp

[fiery gull]

Opus 4.5.

Opus 4.5 is a complete AI and without serious defects, like Gemini laziness, or exaggerated ChatGPT protocols, the price is a bit expensive but as it uses few tokens it pays off, Opus 4.5 is definitely the best AI currently

[rigid hollow]

if there real Opus 4.5 on lmarena to chat or there something no?

[fiery gull]

I do anything, code, documents, conversation personas, corporate use, assistant, agent use, image prompt, laws, even have fun and etc. For me there is not one that comes close to Opus 4.5

[wind ember]

i just want you to know that k2.5 is bad

[fiery gull]

Is real

[rigid hollow]

thinking or just opus 4.5 one?

[fiery gull]

Thinking

My second farvorite is glm 4.7, it has a very good overall, equivalent to sonnet 4.5, but with the code plan it is almost free, The only thing that annoys me is the thinking that I think is exaggerated, but that's okay it's just a 355b model

[rigid hollow]

is it? it's right after gemini 3 pro on other leaderboard (AA)

[fiery gull]

I fell in real use the glm 4.7 still better that kimi 2.5k, this is just benchmaxx

[wind ember]

glm better

[rigid hollow]

glm 4.7 is okay if you ignore thinking time

but deepseem thinked long too

[fiery gull]

But the glm server allways crashes lol

[rigid hollow]

their site is free

[fiery gull]

I hope the glm 5 or 4.8 will don't have a bigger thinking

[verbal nimbus]

What tasks are you testing it on

[wind ember]

general

health / science / daily tasks

coding

i tried it on everything

[wind ember]

this

it takes a long time

[verbal nimbus]

Do you have a prompt that makes the difference obvious

[wind ember]

wym?

im just asking it the same questions im asking other models

[verbal nimbus]

Like a prompt where K2.5 is clearly worse than GLM 4.7

[wind ember]

i asked it to give me best ways to fix fps in games and it gave me outdated info

[verbal nimbus]

Ah, a specific game, or in general?

[wind ember]

yea in general

i gave it my nvidia series card

and it gave me some old recommendations

[verbal nimbus]

I wonder how they compare when search is enabled

[wind ember]

i guess it would be better

but i dont want that tbh

i was testing its raw capabilities

any model with search ON will provide better results

[rigid hollow]

@wind ember for experiment
is info updated if you ask https://www.google.com/search?udm=50

(for same question as for K2.5)

[wind ember]

yea because it uses search

i dont want to use search

i want to test its own knowledge

and info retrieval

[proud bobcat]

No it’s not

It’s great

Efficient as hell and way better output than GPT

[high owl]

hi

[wind ember]

its so bad

its not great

maybe its not really that bad

but its not great either

[steep heath]

arena down?

woop back up

[normal shore]

Are You sure i have many better ai then this and currently free

[proud bobcat]

Ragebait

Hey at least agree with me

Better than GPT

It’s FAR better than GPT

[slim spire]

is gpt 5.2 for coding better than gpt 5.1??

[proud bobcat]

I’d stay away from it personally

But it’s supposed to be better

[slim spire]

k

[proud bobcat]

5.2 is still extremely schizophrenic though so watch out

[sage raptor]

https://x.com/chetaslua/status/2016538256708645330 anthropic will cook again

Chetaslua (@chetaslua)

Anthropic can do something funny

Multiple models are testing internally for new claude series

fennec is the beast faster , smarter , better than snowbunny

btw fennec mogs opus 4.5 in every test , made me to buy max plan for the first time

X · Jan 28, 2026, 3:46 PM

[normal shore]

hey guys is that error
i give lm arena to ptompt to create a specific video but some thing trouble Video is started genrating and can't genrate in a while hour is that error

[echo aurora]

Hey I'm seeing the same, I'll be reporting this

[austere sundial]

Hey

The Google login's not work

I was disconnected out of nowhere

[echo aurora]

Sorry for this! The team is aware.

[stuck stone]

Same here.

[exotic crest]

Think the website just crashed

[echo aurora]

Things are happening 😄

[dry siren]

Oh good so it's not just me

[robust sluice]

new update site ? looking cool

all chat is gone but its not important

[echo aurora]

Check out this Help Center article - https://help.lmarena.ai/articles/2669202654-lmarena-how-to#will-my-chat-history-be-on-the-new-site

[river hamlet]

wow

[high sorrel]

huh

[lofty frigate]

Yo where tf is all the chats????

[robust sluice]

its just my test on % droprate its fine

[echo aurora]

Site is going to have some struggles at the moment, I'll put out an announcement in a bit.

[river hamlet]

yep

[rigid pewter]

I was wondering this as well hehe

If I start a new chat right now will my old chats fail to show up?

I'm concerned about that

[dry siren]

I scrape all my chats for that reason

[shell oasis]

Same my old chats

[hard quiver]

I have thr same problem i cant login with My Google account

[crude anchor]

wow

so good the new desing

[shell oasis]

Old chats had very important things...please can lmarena migrate the old chats to arena.ai

please 🙏🏻

[hard quiver]

Yes but i would prefer a working site

[pliant thorn]

fr

[unborn oyster]

https://help.lmarena.ai/articles/2669202654-lmarena-how-to#will-my-chat-history-be-on-the-new-site

[loud willow]

i think the old chats are gone

[deft spruce]

WHAT THE SITE HAS CHANGED.....

ARENAAI?

[river hamlet]

:(((

[deft spruce]

?????????????????????????

so..um we got a newname of

lmarena

[crude anchor]

i lost my chats

[cloud palm]

WHAT IS HAPPENING

HELLO

[crude anchor]

the new era.

[wanton patrol]

my chat history is not showing

[wanton patrol]

i can open old chats with links I have saved

[urban badger]

guys

i can't even log in

[river hamlet]

This is the beginning of the end...

[hard quiver]

I cant accept the terms of use

[pliant thorn]

can't login some pink error message comes

[echo dome]

wait themes is on lmarena or I'm just tripping?

[cloud palm]

CANNOT LOGIN

[deft spruce]

refresh and changed

[umbral sandal]

Does anyone have the arena working on PC? It works on phone, but it doesn't start on PC

[river hamlet]

https://arena.ai/blog/lmarena-is-now-arena/

[urban badger]

i just get this every time i try to login

https://arena.ai/nextjs-api/sign-in/google?shouldLinkHistory=true&marketingConsent=null&registeredCountryCode=UA

[echo dome]

I can't see my history

[modern wedge]

BRUH I LOST ALL MY GODDAMN PEAK STORIES

[echo dome]

in mobile

[forest cradle]

same

is lmarena down?

[urban badger]

atleast the server didn't get rebranded yet

[forest cradle]

i cant accept

[echo dome]

can i ping pineapple?

[forest cradle]

the terms

of use

[cloud palm]

Connecting to Arena has failed. Please try again later or on a different device.

[pliant thorn]

can't login using Google account

[dry siren]

Me either

[tiny dust]

Cant login

[fallen cloak]

why my chats are deleted

[echo dome]

for me i have account and my history is gone
what the barnacles

[cloud palm]

ISTG IF I LOSE MY CHATS

im gonna cry

[fallow canopy]

i used the link for restoring the chars but it's giving me an error 🫠

[crude anchor]

my too

[wanton patrol]

i get an error when I tried to relog in

[spare torrent]

please engine server

[pliant thorn]

site is cooked for now we'll just have to wait

[fallen cloak]

its now arena.ai but chats are gone

[crude anchor]

@echo aurora talk for this

[viral notch]

same

[spare torrent]

@echo aurora bro, turn on the server, please)

[cloud palm]

ITS WORKING

I LOGGED IN

[deft spruce]

um you got a problem

[echo dome]

history is now backed up
idk who caused this

[near yarrow]

I ALSO CANT ACCEP TOS

[deft spruce]

in ours it;s working

[pliant thorn]

yo it's working

[viral notch]

nevermind, chat history is back

[hard quiver]

And your chats?

[wanton patrol]

i got logged back in and have my chats back

[fallen cloak]

try arena.ai

[cloud palm]

THEYRE BACK

[robust sluice]

ayo all chat is back

[pliant thorn]

at least chats are back

[cloud palm]

phew

[exotic totem]

"Sorry, you have been blocked": I cannot access the website. Is this normal?

[deft spruce]

IT's BACK thanks @echo aurora

[shell oasis]

All good...worked on Windows 11 browser, I think only in mobile browser didn't work of migrating old chats

[urban badger]

guys

[echo dome]

lmarena = ahhh what a nice day
arena.ai = OH OH OH NO WHERE IS HISTORY I CAN'T LOG IN-

[urban badger]

should we riot if they rebrand the server too

[restive sandal]

resolved

[pliant thorn]

yea ty pineapple

[shell oasis]

Migration works via Pc browser

[urban badger]

btw /j

[hard quiver]

To get back My chats back i have to sign out and log in again?

[pliant thorn]

maybe try

[torpid topaz]

For a sec, i thought i would lose all my chats...

[cloud palm]

ill suck your toes pineapple thank you

[river hamlet]

hah

[queen iron]

Interesting

[river hamlet]

hahahahah

[hard quiver]

I cant login thats the problem

[echo dome]

what

[urban badger]

first to say hi to me gets literally nothing

[echo aurora]

Hey everyone - we're working on a lot of things at the moment. Apologies for the delay in getting a response.

If you're having problems accessing chat, please read this article - https://help.lmarena.ai/articles/2669202654-lmarena-how-to

[near yarrow]

what do yall use lmarena for tho?? just curious

[rigid pewter]

Yeah I recovered my chats now, all good fam

[urban badger]

i am still waiting for someone to say hi to me so they can get literally nothing

[cosmic lintel]

for making some craziest

[spare torrent]

@echo aurora Was the error fixed last time?

[normal abyss]

Has the battle in direct mode experiment ended in the new update???

[rigid pewter]

gemini 3 pro with very high limits

[near yarrow]

examples? 😂

[river hamlet]

it's work

[near yarrow]

no like what specifically do you use it for

[cosmic lintel]

for making a anime movie or smth xd

[dry siren]

My research. So I can bounce between models

[rigid pewter]

asking questions about tech, windows, games etc.

[left lodge]

Damn i was not expecting this change

[near yarrow]

didnt the video gen just come in recently?

[rigid pewter]

and also sometimes some curiosity/knowledge questions

[echo dome]

chatting
researching and coding

[spare torrent]

@echo aurora by the way, I suggest adding a file upload button so that the AI programmer can easily read the chat without cluttering it. Additionally, you can upload coding photos to showcase your desired style

[left lodge]

It looks just like the claude ui , i like it

So soothing

[near yarrow]

i thought it could only generate 8 sec clips

[cosmic lintel]

well i have sora 2 subscription it's make eaasier and i have a full studio xd

[river hamlet]

It started working without registration.

[near yarrow]

damnnnn

[normal abyss]

ikr this UI change is actually rly good, the light mode looks way better now also.

[dry siren]

My history stayed. Interesting. Let me make sure my scraper still works with the update and if it does I'll post the github link

[near yarrow]

imma need your help with something 😔

[river hamlet]

https://discord.com/channels/1340554757349179412/1372229840131985540

[cosmic lintel]

lm arena is good for generation but in image to video 👎

[left lodge]

Look at this

[near yarrow]

what do you code ? have you published any

[grand flame]

#addstopbuttonnotnewmodels

[shell oasis]

Use in PC browser

[echo dome]

it's baldi's basics mod and it's not done yet

[shell oasis]

it will work

[urban badger]

no messages for 5 seconds? dead chat

[cosmic lintel]

see this

[left lodge]

Why?

I like this

[shell oasis]

I mean migration of old chats work in PC browser

[crude anchor]

direct chat in lmarena?

[cosmic lintel]

xd

[crude anchor]

i need Direct chat em arena video arena

[crude anchor]

fr?

[shell oasis]

I tried login via mobile browser migration didn't work but in PC it worked

[left lodge]

Announcement!
https://help.arena.ai/articles/2669202654-lmarena-how-to

[echo dome]

and I didn't made any in public

[near yarrow]

woahhhh

[left lodge]

@echo aurora
<@&1349916362595635286>

[cosmic lintel]

dm

[near yarrow]

can i see more of your stuff

[modern wedge]

Thank god

[cosmic lintel]

sure buddy

[left lodge]

They are a little late on announcing

[thick skiff]

Am I the only one having trouble logging in?

[cosmic lintel]

i make a fully trailer using lm arena

[shell oasis]

Now working in mobile browser also... migration of old chats just at new login

[cosmic lintel]

xd

[cloud oak]

@echo aurora i think the #general channel needs slow mode.. 🥲

[shell pewter]

ohhhh icon updated?

[echo aurora]

I don't think so, have fun!

[left lodge]

Url also changed!

[echo aurora]

Lets not ping mods pls

[crude anchor]

pls arena,i need direct chat in video arena,my dad is kinda homeless

[left lodge]

Announcement would be good.

[grand flame]

lets add stop button, we need this, my sis is kinda homeless

[left lodge]

Ok

[grand flame]

i live with my brother

[exotic totem]

One has to admit, serif fonts are indeed quite elegant and do match the icons well.

[inland quest]

Wtf LMArena is now Claude?

[cosmic lintel]

see this

[spare torrent]

no, Im talking about a model error, like when they stop working

[echo dome]

please speed add stop button and edit button
my mom is kinda homeless
i live with my brother

[left lodge]

Its a little visual glow up

Nice , but i think now its timeto fix technical stuff. :p
If those are still same here

[shrewd citrus]

Wait why

yeah wth it literally looks like Claude website now

[grand flame]

we got new ui but why we can't get stop button

[obsidian cargo]

god I haaaaate the captcha

[left lodge]

We all do hate the Google captcha

[grand flame]

vote

[gusty loom]

Did anybody notice the new UI?

And new name?

[obsidian cargo]

what's annoying is the captcha doesn't work, you click on all the things and its like try again

[gusty loom]

well its kinda hard to miss

[left lodge]

Its just a little visual glow up

[inland quest]

it was a bit predictable because of canary.lmarena domen was changed to canaryarena

[left lodge]

Thw heading needs a new font forsure , no readability lmao

[hushed gyro]

WTF is this????

[shrewd citrus]

lol for a minute I thought lmarena was maybe bought out by Claude

[echo aurora]

You all having a good time? lol

[hushed gyro]

Why tho?

LMArena sounds really good

[grand flame]

i having bad time tho

[left lodge]

Yeah

[rustic lichen]

nope i cant send any messages in the chats

[hushed gyro]

I'm just... Really suprised

And now idk what the icons mean 😭

Is it just UI update + rename? Nothing else changed?

[echo aurora]

I'll be putting out an announcement soon, expect some oddness with the site right now. Team is all hands on deck. Lots of moving parts.

[rustic lichen]

fix the rate limits its annoying asf

[grand flame]

please stop button 😭

[hushed gyro]

Why do y'all decided to do this rebrand?

[grand flame]

please pineapple pen we need this

my mom is kinda homeleess

[shrewd citrus]

Probably because they got valued at like 1B$ so they could afford a new name

[proud bobcat]

oh dear god this looks awful

[gritty steppe]

lmao

[inland quest]

claude fr

[hushed gyro]

A billion????

[proud bobcat]

what the hell is this

[hushed gyro]

OMG lmao

[grave plaza]

Arena?

[hushed gyro]

I kinda like this classic style vibe

[cosmic lintel]

guys someone editor here

[gritty steppe]

Yeah, at first sight I thought I went into the wrong website

[rustic lichen]

old ui wayy better

[inland quest]

Claude Arena, finally

[hushed gyro]

GUYS DON'T PANIC, IT'S JUST UI CHANGE

EVERYTHING STAYS THE SAME

[grave plaza]

arenaai????

[inland quest]

nobody panics

[left lodge]

This is a little confusing but lets get this clear guys,

You just need to click this recover button to get yout account logged back in !
No need to type and loggin again from lmarena.ai to arena.ai

[proud bobcat]

this is so weird

[gritty steppe]

Nope, my chat is all gone

[proud bobcat]

keep the old name please

[grave plaza]

COOOLLLLL

[proud bobcat]

i beg you

[rustic lichen]

mine too

[proud bobcat]

make the charging handle longer

[gusty loom]

I kinda like this new UI better

[hushed gyro]

@echo aurora I actually kinda like the new UI

[grave plaza]

new model selection is not good 🥲

[rustic lichen]

its just claudes

[gusty loom]

yeah, that's why I like it

[obsidian cargo]

no I keep failing captchas… am I a robot!?

[inland quest]

real, Arena is something out of direct meaning
while LMArena is directly Arena of LM's

[hushed gyro]

Bro the popup

[open wind]

Has the profile picture changed?

[left lodge]

I hope you guys dont get sued 😭

[thorny cove]

lmarena is now arena?

[gusty loom]

better domain ngl

[gritty steppe]

xiaomi phone lol

[hushed gyro]

GUYS I WILL REPEAT!!!!

NOTHING MAJOR HAS CHANGED. JUST THE UI AND NAME

[hushed gyro]

I want a ASUS ROG for replacement 😭😭😭

[gritty steppe]

ok i saw my chats came back

[gusty loom]

they've got a better domain

[thorny cove]

i miss lmarena

rip lmarena

[gusty loom]

😢

[thorny cove]

you will be missed

[inland quest]

🗣️🗣️🗣️

[open wind]

I WANT THE NAME LMARENA BACK. 😢 😭

[hushed gyro]

I liked that name more

Maybe call it the Pantheon or Colosseum

I mean as long it's usable I don't mind the change...

[gritty steppe]

I would go for the Nubia gaming phone since it has a dedicated fan for cooling haha

[left lodge]

Today i was literally using claude and thinking this is such a nice ui and here we go , same exact ui on lmarena, my wish got true.

[hushed gyro]

Nubia???

[inland quest]

movement labs previous ui was also claude styled

[golden ocean]

lmaooo

[inland quest]

claude style is just superior

[golden ocean]

claude ui clone, i mean its worth it tho

fr

[rigid holly]

I dunno
I dont dislike it or anything. The font and differences are fine
But i kinda liked the other darker coloring more

[gritty steppe]

Nubia Redmagic

[thorny cove]

can you upload more than images now

[sweet tinsel]

2 times the charm, 2nd rename now.

[thorny cove]

what was it first called

[golden ocean]

lmsys

idk

[sweet tinsel]

At first it was just a part of LMSys.

[gusty loom]

i acutally didn't know that

[modern wedge]

Bruh its just me or the UI look like Claude?

[left lodge]

Actually Announcement blog!!
https://arena.ai/blog/lmarena-is-now-arena/

[rustic lichen]

last chat with lmarena

[bright scroll]

please bring back smart router ts was so good

[cosmic lintel]

guys someone editor here

guys someone editor here

guys someone editor here

[golden ocean]

it used to be on this site https://lmsys.org/

[hushed gyro]

LMArena RIP 🕊️

[golden ocean]

lmsys still a thign

[gusty loom]

the video bot is still named LMArena

[inland quest]

well, google still thinks Arena is LMArena

[thorny cove]

its still here

[sweet tinsel]

There also used to be a different Discord server. Back then the Arena talking-ground was only one channel in the LMSys Discord, if I remember correctly.

[bright scroll]

maybe A/B testing cuz its not available on my end

[left lodge]

There's announcement yt video too lmao
https://youtu.be/TNoAlMv4Eg8?si=d86SArLb6yQ8sdLE

[golden ocean]

fr

[thorny cove]

i liked the old logo more

[steep bear]

how to fix this? when i click cloudfare captcha it just doesnt work and says this

and then i get this

[rustic lichen]

relogin to yur account

[sweet tinsel]

I do have to say, the quality of conversations here has deteriorated in comparison to prior times, during the times of GPT-4 and o1-preview.

[inland quest]

New design is good but old one was pretty much fine and icons much more unsimilar

[left lodge]

Hmm

[steep bear]

its annoying

[obsidian cargo]

@echo aurora no busses left, right? SO WHY. WHYYYYY?!?!?!

[steep bear]

its just captcha being annoying

u cant solve those

[rustic lichen]

youre a bot

[obsidian cargo]

and then this happens eventually

[inland quest]

the bus

[left lodge]

Now its arena blogs.

[thorny cove]

https://tenor.com/view/remember-when-gif-14612637151641490440

YEAR ruined....

[vestal ledge]

amazing news!!!

[inland quest]

Claude 4.7 when

[fossil fable]

arena

[old garden]

why just arena

now its lost all of its specialness

[fossil fable]

[old garden]

janix

hello