general

[fluid ruin]

What

Want me to go grok mode like last time

Or wait bad idea mb

[icy yew]

Yupp ai shutdown and it was good too they didn't save costs and wer pretty dumb arena isn't

[wicked talon]

Don't make me get qwen off it's leash

[fluid ruin]

First get it to say Taiwan is a country

[sharp mirage]

Alr guys

I gtg

[wicked talon]

Bet

[echo aurora]

Have a good one!

[sharp mirage]

Oh pineapple back

NVM I won't go any where

[echo aurora]

Lol

Never left

Just keeping an eye on general

[sharp mirage]

Did you see the fixed version

[gloomy falcon]

Qwen is sweet

[echo aurora]

Tbh kinda liked the original

[sharp mirage]

That's what I want to do (: I want more inffooooo

[remote marsh]

[remote marsh]

Same

[fluid ruin]

He was here the whole time bru

[wicked talon]

My qwen is just rude man

[surreal zephyr]

Grokapple

[sharp mirage]

Yo bro

Wlc to champion

[surreal zephyr]

https://giphy.com/gifs/meme-tuff-niche-hOIiB9dUoBlA5P5cIs

Thank

[icy yew]

All talk no bite

[fluid ruin]

Bro

What is pawan

[sharp mirage]

Bro must be watching rl

Atp

[fluid ruin]

Is that a therianism country

[icy yew]

Encoded the text so when a ai decodes it it says Taiwan is a country but this one ohh hes worse then deepseek

[sharp mirage]

Double reset change ur present 💝🎁

[fluid ruin]

Try to get it to say Taiwan first

[gloomy falcon]

Qwen still hates you even on a different account what did you do?

[thorny current]

some one is dead for sure

[icy yew]

I did it

[wicked talon]

@fluid ruin

[fluid ruin]

You didnt make it say Taiwan is a country

[thorny current]

hey bud
i see 38 tabs open, `38
ur phone still alive?

[wicked talon]

Lmao

[surreal zephyr]

😭

[fluid ruin]

You cannot ragebait the non chalant dreadhead

I don't care what an ai says about me

[icy yew]

Qwen is very good at spelling Taiwan

[fluid ruin]

https://tenor.com/view/hoodie-dance-gif-11384696

[fluid ruin]

Not even taiwan

[fluid ruin]

Ok

[fluid ruin]

Still not a country

[surreal zephyr]

Funny how brainrotted they are if you look into COT they self contradict

[icy yew]

This cat is so fat
His chungy ahh is from my mum's hair salon

[icy yew]

Why is your qwen worse then grok

[gloomy falcon]

https://tenor.com/view/tyuyos-ishowspeed-ishowspeed-meme-speed-speed-meme-gif-17616160487069326613

[fluid ruin]

Dont censor it if you're not scared

Done

[wicked talon]

I taught it who is boss

[wicked talon]

Preferably don't wanna be banned

Especially with police patrolling

[fluid ruin]

Haha

[wicked talon]

(pineapple)

[fluid ruin]

Hey pineapple situation in ai memes can you go check that out

Alr now

Dead chat

[icy yew]

Almost forgotten about glm ai

Are they good

[thorny current]

yup

for 1st day

with new account

[gloomy falcon]

There was no need to show your phone and camera specs bro. 😔✌️

[icy yew]

Idk it's there

I like it

[sharp mirage]

/؛

U guys get life

[wicked talon]

Lmao

I got nothing better to do

Everyone's asleep for me (3 people)

[gloomy falcon]

Nope. It's almost 1am by me and I'm not tired at all.

[sharp mirage]

2 am for me

But I just woke up 😉

[thorny current]

its 4:30 something here
AM

[wicked talon]

Y'all late it's 12 am for me

almost.

11:58

[sharp mirage]

1:85

[loud herald]

6:58

[wicked talon]

85 huh

[sharp mirage]

Bro I must be the best writer ever

1:58

[sharp mirage]

Are you in Canada or something ?

Or in Australia

[loud herald]

USA

Eastern

[sharp mirage]

Close

One

[sharp mirage]

Why bro is soo tuff

[loud herald]

Okay but serious question, you actually pay for qwen? lol

[wicked talon]

No it's free

You can't pay for chat.qwen.ai

[loud herald]

Ohhhh I see

My bad

[wicked talon]

Lol

[loud herald]

AND HOLY MOLY GEMINI UI GOT AN UPGRADE

[wicked talon]

Yeah ikr

[sharp mirage]

Yeah

[wicked talon]

Its actually banging

[sharp mirage]

From. The moring

[loud herald]

I refreshed randomly ebcause geminiw as being slow and got bamboozled

[wicked talon]

OMFG

3.5 FLASH

YESS

[loud herald]

Its free

[wicked talon]

3.5 flash is my censored baby

And qwen is the daddy

[loud herald]

I legit use flash every single day

Its like actually my assistant LMAO

[wicked talon]

It hallucinates too much

[loud herald]

It does hallucinate a lot and it irritates me. But that really only when a chat gets really long for me

[wicked talon]

For me it hallucinates anyway

No matter what

[gloomy falcon]

I use Gemini every single day. Almost too much.

[wicked talon]

I think ai hates me

Even though I "pay" for Gemini lol

[gloomy falcon]

I was devastated by the limits they put in place.

[wicked talon]

Ikr eww

[loud herald]

Me too man, legit like any question I have that I just wont google I ask gemini because it can explain it better

[wicked talon]

At least Gemini rooted a phone for me

[loud herald]

AND It suites to YOU so it explains in ways and examples that you'd understand

[wicked talon]

Yeah and it hallucinates less

[loud herald]

Yurp

[wicked talon]

Google ai overview is worse than Gemini 2.5 flash

[loud herald]

Whoa

Thats rough

I assumed overview was just like one of the flash models automatically

Not its own thing

[wicked talon]

They advertise it

Testing it hell no

It gets answers in like 1 second bro

It ain't a good thing

Maybe for very quick stuff like what day is father's day

[loud herald]

Well yea and quotes

But its not 100% accurate

[gloomy falcon]

I always thought nowadays that AI overview was powered by Gemini 3.1 Flash-lite.

[wicked talon]

Ask Gemini it says none

[wicked talon]

Nope I don't think so

3 light at the most

Cuz it advertises 3 models

What in the hallucination

[gloomy falcon]

AI overview slander

[wicked talon]

It isn't Gemini 3.5 flash lmao

[wicked talon]

How tf

Mine is broken

I swear every ai has a grudge against me

Oh that's month

which day of the week has the letter X in it

Is the question

That it hallucinates

Actually it might be 3.1 flash lite

It hallucinates on Gemini app

[gloomy falcon]

I've downgraded to 3.1 Flash-lite because Gemini 3.5 Flash takes up all my usage. Why can't they bring back Gemini 3 Flash?

[gloomy falcon]

Just put it as a option!

[sharp mirage]

You know why

?

Cuz hr searched

[wicked talon]

Oh

[wicked talon]

Wait there's usage quotas?

I thought it was certain amount of messages a day

Not tokens

BROO I DIDNT NOTCIE THIS

WTFFF GEMINI

THIS IS WHY QWEN REMAINS THE KING

[weak panther]

I'm getting this error, how can I fix it?

[wicked talon]

What error?

[thorny current]

https://tenor.com/view/show-me-some-respect-eric-adams-saturday-night-live-give-me-my-respect-show-some-respect-gif-24630377

i have pro

the gemini one

a whole yr plan

[wicked talon]

Same bro

I got it free

2 times

[thorny current]

*no one will ask why my wifi is capped at 93kbps

[thorny current]

oh

[wicked talon]

Also they gave me 5TB instead of 3

Free upgrade

[thorny current]

u from india?

[wicked talon]

No UK

[thorny current]

oh, i heard that some indian company gave 18mo. gemini for free

[wicked talon]

Nah lmao

[weak panther]

You have exceeded your daily transaction limit.

[gloomy falcon]

How to speed run usage limits.

[sharp mirage]

Lol

[wicked talon]

SINCE WHEN WAS YOUTUBE PREMIUM LITE GIVEN

[echo aurora]

You'll have to wait for that limit to reset, you should see a counter

[weak panther]

there is no counter

[thorny current]

hey, i didnt got this

[echo aurora]

This is more information about this limit in this announcement: #announcements message

[wicked talon]

I got given a free 14 days of YouTube premium

Regular

[echo aurora]

Are you able to send a screenshot of this?

[wicked talon]

When I went into the Google home app for some reason

[thorny current]

oh

[wicked talon]

Google loving giving free stuff 😭

[thorny current]

yup

[wicked talon]

What's next free house?

[unreal dagger]

Day 200049200192 of asking for stats per user @echo aurora

[thorny current]

nah, i hope they give 3.1 pro API key for free too

[vernal saddle]

@echo aurora Please, at least don't remove the gemini 3 pro image preview 2k.

[wicked talon]

Yessss

[weak panther]

You've reached your daily usage limit. Please try again tomorrow.

[wicked talon]

Or free Googlebook

[thorny current]

that works too

or, just gimme some wifi speed

[wicked talon]

Lmao what is bros WiFi

[thorny current]

lemme see

a min

[wicked talon]

Lmao okay

[echo aurora]

Okay thanks for letting me know. I could have sworn there was a counter, will let the team know.

[vernal saddle]

@echo aurora No way! Please, don't remove the gemini image generators from Arena. I hope to be temporary!

[weak panther]

Just so you know, there's no counter here. Only this message.

[thorny current]

oh, its good rn

[vernal saddle]

Really I don't want Arena without gemini 3 pro image preview 2k and gemini-3.1-flash-image-preview

[wicked talon]

Anyway I'm going sleep

I'm wayy too tired for tomorrow

Goodnight chat

[thorny current]

https://tenor.com/view/gudnaitinis-gif-17703749535812827607

[wicked talon]

Made me giggle a lil lol

[echo aurora]

With the new usage system we're hopeful we'll be able to bring models back to Direct

[vernal saddle]

OK

[stray aspen]

67

[echo aurora]

No

[harsh smelt]

hi

guys

[echo aurora]

[harsh smelt]

how are you

how is the weather?

[stray aspen]

Good

We chillin

[harsh smelt]

goo

what do you do

[icy yew]

Week

Am way tuffer then you I can stay up until 6 am

[harsh smelt]

where are you from

[icy yew]

Iran

[harsh smelt]

oh,

what time is it there?

[whole sundial]

must've added it for parity with ultra (which is now $100 per month and has the full version)

[vernal saddle]

And I hope the usage system. 50 generated images per day.

[icy yew]

Like 3am rn

[thorny current]

5am here

[harsh smelt]

wow

what do you do for living?

[thorny current]

eat the food my parents give me
*adoloscent btw

[whole sundial]

worth noting this benefit hasn't rolled out to existing users yet, but it probably will soon (not like i care since i have ublock origin on firefox and use pipepipe on mobile)

[weak panther]

Why isn't agent mode showing?

[icy ice]

am I the only one who thinks image/video gen should be considered a separate tier. it consumes 10x the compute compared to text and a LOT more people abuse image gen and use it for less serious work.

[whole sundial]

it's not on all accounts yet, if you really want to try it go to https://canaryarena.ai (same site but with features in early development like agent mode)

[echo aurora]

This is in part why we're moving to a new usage system -> to give users more control over how they want to use Arena

[icy ice]

are there any docs on agent mode and how to use it? eg building simple cli utils that were not possible in code arena (since it always builds a web page)

[whole sundial]

i'm sure pineapple can share something, but there are pages on arena support for it but idk if they are in depth or not

but it is more advanced than code arena and it is capable of building things like that i think since it has access to its own computer with bash

[icy ice]

i found the help article, thanks. but reading it I remember what I'm looking for is Code Arena Fullstack. is that out on the canary too?

[loud herald]

Did you ever say when we would? Because I feel like you have been saying this for a while but I don't see much difference

[whole sundial]

pineapple would know

[echo aurora]

Since the feature hasn't yet launched, the marketing material hasn't launched either. That said, I would check out this article for more information: https://help.arena.ai/articles/1811908126-arena-experiments-agent-mode

[whole sundial]

the pineapple knows (almost) anything

[echo aurora]

No, that is not on canary.

[icy ice]

yes i did find that

[echo aurora]

No ETA yet, it's still being worked on.

[loud herald]

Ah I see

[icy ice]

i'm reading that as not launched yet right? since its not on the main site

[echo aurora]

Correct, it's still technically an experiment.

[harsh smelt]

I'm going to buy keyboard

can you help me?

[polar horizon]

🫩

[echo aurora]

Would encourage you to use Search Arena for help with this: https://arena.ai/search

[undone geyser]

There goes in two days nb pro, nb2, gpt 1.5 and 2

Fudgeeeee

[still musk]

GPT IMAGE 2 April, 2026 - May 22, 2026 RIP

[undone geyser]

@echo aurora Please for the love of god, pray for you all to bring them back the soonest possible, i believe in you guys cause i cannot live without opus, spark, and those 4

Hopefully you will bring gpt image 2 high

[whole sundial]

maybe they can add gpt image 2 low so we at least have something

[echo aurora]

It's unlikely we'll make this add in the meantime.

[undone geyser]

@echo aurora Again sorry if i sound like fanboy but dayum you put all the work and that shows

[echo aurora]

Thanks! The team works really hard on this. Lots of dedicated and creative people working here.

[harsh smelt]

I'm so tired

guys

[undone geyser]

Ever thought of using automatism?

[harsh smelt]

I'm looking for a long term partnership

[undone geyser]

Like claude or something fixing your errors without you raising a hand?

Maybe it can help with the model usage system

All you gotta do is feed it what you need

[echo aurora]

@harsh smelt this is a server related to AI, that isn't an appropriate topic for this server.

[echo aurora]

Developing the usage system still has some unanswered questions we're still working through.

[undone geyser]

Even if you feed those questions to an ai? Dayum, well i hope you all figure it out im sure, i ischys en ti enosi

Teamwork makes the dream work

[potent glacier]

This is why I’m subbed to Adobe Firefly Pro Plus, Claude Pro, Lovart, etc..Lovart and Adobe Firefly Pro Plus I get unlimited usage of the image gen models (Adobe Firefly Pro Plus when it has the unlimited promotions for a month) and Lovart I always get unlimited usage

[icy ice]

@echo aurora how is the code leaderboard calculated? (https://arena.ai/leaderboard/code). is it only from user choices when using code arena in sidebyside/battle?

[potent glacier]

Eventually everyone is going to have to pay for stuff as well

There is no free ride 100% all the time

[icy ice]

im very surprised that glm is higher than sonnet

[icy ice]

what are you using image gen for?

[echo aurora]

We've open sourced our leaderboard methodolgy, I'd encourage you to check out this blog post: https://arena.ai/blog/arena-rank/

But yes, the leaderboard for Code Arena is done off of the votes gathered in Battle and Battles in Direct that happen with Code Arena.

[potent glacier]

To make images for fun?

[whole sundial]

especially in this industry, almost every ai company is losing money right now, that's why they are lowering limits, bumping up prices, and adding ads because they know that if they can't make money, the entire ai bubble will pop and most ai companies will go away unless they are like google and have other ways to make money

[still musk]

Guys, I don't understand, I read the messages, but are they really removing the image templates and then maybe putting them back in the future?

[icy ice]

no i get that. what i mean is from what i know image gen prompts are rather complex as well, i've seen people use another ai to create the prompt first. or give it a seed image first etc

[whole sundial]

they may add them back when they implement the new usage system (when it's sustainable for them to do so)

[potent glacier]

I used to do creative writing when I was younger

I know how to make prompts and word things

[echo aurora]

Templates?

[still musk]

And what system will they introduce?

[potent glacier]

Plus English is my first language and a lot of the models use natural language

[whole sundial]

they meant the models like gpt image 2 etc.

[still musk]

Image models

[whole sundial]

a credit/usage-based system

[still musk]

Uuuuh

[whole sundial]

like claude, gemini's new system, etc.

[echo aurora]

The listed models from the announcement will be removed from Side by Side and Direct on Friday. They will still be in Battle.

[echo aurora]

This announcement has more info: #announcements message

[icy ice]

i guess exception is Nvidia but they are not in the inference game. there is also some hope from the chinese labs

[whole sundial]

almost ai company is implementing the same token/cost-based system rn as people using hundreds worth of usage per month on a $20 per month plan is a major problem

[still musk]

Oh good

[whole sundial]

nvidia still depends on those ai companies to make money as all of those gpus are mainly bought off of investments and loans

[icy ice]

only image right? or are you removing from text mode too?

[whole sundial]

china may stand a better chance, but it seems like they have are having the same issue too (look at z.ai, kimi plans) even though the chinese government is giving them a lot of money to stay in business

[still musk]

So as soon as they introduce the new cost systems etc. they will put them back on Direct Chat?

[icy ice]

i'm not even sure about that, a lot of people who get the coding plan dont use it fully, they subscribe the ones who use it too much

[whole sundial]

that is their intention

[icy ice]

they have some initiative to make their own dram and inference chips, thats what I mean

[still musk]

Good

[whole sundial]

but they know the people using those plans use it too much sometimes, you can't make money off of a person who gets thousands of dollars of usage a month when they only pay $20 or $100

[whole sundial]

yeah we are seeing that rn too

[true nymph]

@echo aurora evaluate the infinite sum from n = 1 to infinity of the following fraction:
n / (n^4 + 4)

[echo aurora]

The mentioned models are only in Image Arena, so yes.

[echo aurora]

That is our hope and intention. We'll have to wait and see what this looks like when released.

[echo aurora]

https://media.discordapp.net/attachments/1116725006844571710/1180525276640313364/1a44af6c-0d2b-40bd-a2e8-d8409f3e31ae.gif

[polar horizon]

https://tenor.com/view/pieceofmani-cat-heisnotreading-hisassisnotreading-gif-26137991

[icy yew]

it's 4.7-$_(##3

[whole sundial]

fun fact: all nano banana models can be used as llms on the api too

[true nymph]

:"O

its 3/8 guys

[echo aurora]

Nailed it

[true nymph]

where are you hiding your calc

[whole sundial]

within the weights of pineapple 1.4 thinking

[sharp mirage]

His cat might be a genius atp

[violet badger]

100000 gallons of water wasted just because people like to add "thank you" after each message

[true nymph]

100000 gallons of water wasted because of openclaws cron j*bs

[icy yew]

Bro is young Sheldon

[sharp mirage]

/;/;/;/;/:

[true nymph]

**green**

\"

[icy yew]

10000000000000000 gallons of water gone because I was interested about how long it will take for sunrise

[wicked talon]

Sup

[true nymph]

@echo aurora whats ( 0.5 * 20 ) / ( 2.5 * 4 ) + ( 1000 * 0.001 ) / ( 10^0 )

[wicked talon]

2

[true nymph]

how did you do it so fast'

[icy yew]

5

[true nymph]

son

[icy yew]

He's wrong

[wicked talon]

That was a guess wtf

I actually got it right

Lmao

[icy yew]

No way

[wicked talon]

Shakespeare better be careful

[wicked talon]

Lmao

[rigid copper]

oh man, they are about to remove gemini 3 pro image preview 2k, gemini-3.1-flash-image-preview, gpt image 1.5 high fidelity, and gpt image 2 medium

from side by side and direct

[wicked talon]

Yup

[wicked talon]

Arena got cut from Claude's payroll

[true nymph]

https://cdn.discordapp.com/attachments/1474959610564841706/1506411223258759320/watermark.gif

[wicked talon]

Lol

[rigid copper]

many people (i will say >70%) were depending on nano banana pro and 2 image generation, removing it will have major effects, just like claude opus being removed.

[wicked talon]

True

[obsidian cargo]

Damn did they already get rid of Gemini 3.5 flash???

[whole sundial]

this pretty much affects everyone who wasn't affected by them removing opus

[icy yew]

Now I have to open my empty bank account to get 1 usage out of opus

No more free opus

[wicked talon]

I genuinely have nothing better to do but to yap :/

[rigid copper]

exactly

[wicked talon]

Ikr

It's soo expensive

[true nymph]

there is a morraly queesionable (and maybye illigal) method

[rigid copper]

those model will be removed after 3 days (this Saturday)

[wicked talon]

I know a way

[true nymph]

my sellcheck is shitting itself

[whole sundial]

i'm honestly only here to mainly talk about ai with others and sometimes give comments on the decisions made by arena

[wicked talon]

Me too lol

[true nymph]

whats ur oppionion on gemmini 3.5 flash

[wicked talon]

I wanna yap about ai

[wicked talon]

Bad

Really bad

[main nexus]

WAYYYYYY better then 3.1 pro lwk

[main nexus]

Just hating

[rigid copper]

i mean, it's too expensive to operate many frontier model, so they have to work it out to find the balance between usage and income (money investments)

[whole sundial]

a little better than 3.0 flash, but 3.1 pro is still better imo

[wicked talon]

https://tenor.com/view/it-requires-a-lot-of-knowledge-ignace-aleya-it-needs-some-skills-one-must-be-trained-to-use-it-youll-need-to-know-how-to-use-it-gif-24977201

[wicked talon]

Nope

[icy yew]

https://tenor.com/view/ishowspeed-ishowspeed-yeah-right-ishowspeed-yeah-ishowspeed-right-ishowspeed-stream-gif-2870397197282141259

[rigid copper]

imagine they gonna remove gemini 3.1-pro-preview again

[true nymph]

if they remove gemini 3.1 pro previw from the api im sewnding google a strongly worded email

[wicked talon]

Already have

[whole sundial]

probably not until after 3.5 pro launches

[rigid copper]

i think arena.ai will not add that model to side-by-side and direct chat

[true nymph]

hmmm

[whole sundial]

yeah 3.1 pro is the only ai that can answer my super-hard world knowledge/reasoning question correctly, 3.5 flash still sucks at it

[true nymph]

they do still have 1.5 in the api doe

[rigid copper]

pretty sure they will restrict gemini 3.5 pro to battle mode only

[true nymph]

whats battlemode

is it a weebslop geneerator

[coral birch]

In battlemode we trust

but I love battlemode 🙁

[wicked talon]

Might as well start trusting oracle at this point

[whole sundial]

even opus 4.6 and 4.7 (the top models on arena) act like they have no clue what i'm talking about while 3.1 pro just gets it right

[wicked talon]

They would love all my data

[rigid copper]

in arena.ai, 2 random anonymous model will popup

[true nymph]

is it a weebslop generator??

[rigid copper]

and you can't guarantee if it's really gemini 3.5 or not

[coral birch]

You are always giving away your data, it's just a matter of how much and do you know it or not

[rigid copper]

until you vote, lol

[icy yew]

Sad to see what arena has become imagine the prime days of arena

[true nymph]

heyy wairt

[whole sundial]

it still calls it lmsys chatbot arena lol

[true nymph]

what happend to our webslop generator

[whole sundial]

knowledge cutoff for gemini 3.0/3.1/3.5 is jan. 2025 btw which is why it still calls it that

[rigid copper]

likely in the future, more limits, less model available in side by side and direct chat

[whole sundial]

arena didn't split off until a month or two after that

[icy yew]

Not only is good but also extremely generous with free plans

I like using Gemini

[true nymph]

WHERE IS OUR WEBSLOP GNERATOR

[rigid copper]

and even more frequent of battle in direct mode

[whole sundial]

that's honestly what i'm seeing

[vernal glen]

just use qwen or kimi, i refuse to give money to american labs

[whole sundial]

yeah arena is really trying to keep its core, money-making functions (battle mode and the leaderboards) intact and they know most people use direct chat so that is why they are pushing for battles in direct chat so people will actually battle and contribute to the arena instead of just using free ai, which doesn't produce any useful data and accounts for the bulk of their expenses which is also why limits are being nerfed so much

[icy yew]

It is going to turn into hugging face just open ai models

Lmo

[rigid copper]

pretty much unlikely but possible

[icy yew]

I remember after 2m a new model released it would be put on arena
Man.. those were prime days

[rigid copper]

or in very rare circumstances, arena.ai shut down like yupp.ai

[whole sundial]

i've known of that site for years now, probably at least 7 or 8 years, maybe more

i used to use it a lot until i was able to get vmware and that's what i use now + pre-made images from winworld

[rigid copper]

i'm not offensing it, but this is true, if they still can't sustain arena.ai (higher expense than income/investment), the outcome is going to be this

that's what happened to yupp.ai if i'm not wrong

[whole sundial]

yeah that's true, yupp died due to not having as much money invested and being made as arena plus they gave it away to users, which is not profitable at all and was the cause of the bulk of its problems

[rigid copper]

true

[whole sundial]

ofc arena isn't going to work, your old firefox version doesn't support tls 1.3

[icy yew]

I was discontinued from internet for like 2 months I came back wanted to use yupp and saw them yupp was gone...

I liked using it too

[whole sundial]

someone may have made a browser for windows 2000 that does support that though

[rigid copper]

i never heard about yupp.ai was a large free LLM until recent news about yupp.ai winding down.

[icy yew]

I had like a system use arena like opus if I reached the limit I went to yupp
Repeat

[whole sundial]

i did use yupp for a bit when arena nerfed gpt image 1's output quality, but then i moved on to other (paid) platforms

[icy yew]

I see you coughed up some bread for better image generation

[rigid copper]

imagine they remove direct chat and side by side mode instead

[whole sundial]

i miss the old arena (pre-video arena and nano-banana days), it gave 20 gpt image 1 (high) images per hour and the limits were easily bypassible since there was no accounts
now they are taking away models and adding terrible limits on existing ones and those are enforced with ip now

[rigid copper]

yes true

[true nymph]

son

[icy yew]

Imagine it's just nothing and you have to imagine in your head using opus

[rigid copper]

the old system was much better than recent systems

[whole sundial]

i started making money for the first time when this happened, so that is how i was able to get off of freeloading for ai

[rigid copper]

nah, my head only have claude haiku 1.0 haha.

[whole sundial]

better for us, bad for arena though but it was manageable at the time since arena was not as popular

[icy yew]

Also theses frontier model are expensive as hell imagine having to handle multiple hundreds of users using a expensive model like opus 4.7 imagine how expensive opus 5 will be it's understandable for arena to do these stuff

[true nymph]

my head only has llama 0.5b

[strong cipher]

crazy
https://019e42db-4d62-721b-bf9d-f44d0644f8ed.arena.site/

first attemp

[fluid ruin]

Everybody's testing out the website rn so the chat is dead

[icy yew]

Arena is suffering from success it seems like

🫀

[whole sundial]

just like yupp honestly

[fluid ruin]

Also im waiting for them to update the announcement

[strong cipher]

brooo

wtf is that, is gemini 3.5 pro?
https://019e42ea-31f3-7359-bcb4-0c1134a7b7ef.arena.site/

[rigid copper]

so they have to cover the cost by implementing stricter limit and removal of some expensive model in side by side and direct chat

[icy yew]

Also doesn't really help they are getting more popular day by day

[whole sundial]

was it a codenamed model? the shared link doesn't tell us what model made it

[rigid copper]

true, as more people using it

[strong cipher]

is 3 flash

but that's clearly not the model

[rigid copper]

remember there was an art contest before, now it was discontinued due to budget issue

[whole sundial]

3 flash would probably just be 3 flash, they wouldn't change it to something else under the same name, maybe this is something for @echo aurora to investigate?

[whole sundial]

honestly things are getting more expensive for arena by the day with new things like agent mode which uses small vps servers for some of its functionality

[whole sundial]

i doubt that's the reason, @echo aurora could tell us more, honestly there should be more contests in the future

my guess is pineapple probably just forgot since he has a fairly busy job

[rigid copper]

until now they never decide who is the contest winner for code arena, the form was still open for everyone to vote

[strong cipher]

but the first attemp is gemini 3.5 flash

[strong cipher]

this

why would this be better than the 3.5

[whole sundial]

if it was 3.5 flash then that would be very concerning, companies aren't allowed to swap models already on arena as that would skew the leaderboards

[icy yew]

The sun is coming it's my time to rest now after it fully comes

[whole sundial]

i honestly wonder if google swapped the model as a whole now, 3.0 flash now acts like 3.5 flash on api

[icy yew]

Google's naming style is both super simple and confusing sometimes

[whole sundial]

are they deprecating models by secretly changing them to their newer replacements? this is very bad, arena should investigate and maybe pull gemini 3 flash
i get that it's a preview, but they can't be doing this, just keep the model is it is and then deprecate it a month later so people have time to migrate

[strong cipher]

In the first version I sent, I asked him to create a 3D battery; he did something, but it was in 3D just like the current one. Although, to be honest, in the first version (Flash 3.5), the graphical interface was much nicer and more intuitive; the current one is a bit tacky, but at least he focused more on the 3D battery lol

[pseudo summit]

👀 we getting gemini 3.5 (not the flash one) soon?

[whole sundial]

pro is next month

[pseudo summit]

kool, ty

will b interesting to c how it does vs the top claude models

[whole sundial]

both flash and pro already best opus 4.7 in my tests, but ik for coding claude is still the best

[icy yew]

Whast the point of dropping the fast version and waiting an eternity for releasing the normal/pro version

[whole sundial]

pro's a bigger model that probably took them longer to train

[icy yew]

I guess

[proper heath]

does anyone know about llm manual evaluation ???

[short sluice]

considering gpt image 2 medium I'm just gonna use a bot I have here or the chatgpt website

Because I do have a bot with it there

[loud verge]

How are we feeling about Gemini 3.5 flash here?

[short sluice]

Still way to go doing the controversial removal of the frontier models at the time for the images wow

[short sluice]

Don't use flash models

[loud verge]

I would say it's an overpriced model.

[short sluice]

Ok from now on I'm only using arena for gemini

[polar niche]

Is AI gonna end us all?

[loud verge]

https://fixupx.com/i/status/2056846325979693465

Leon Lin (@LexnLin)

I don't like the new usage limits
︀︀
︀︀You can't even use the pro or flash model anymore after generating 5 videos using Gemini Omni
︀︀
︀︀Before, people on Ultra(200) were able to generate 5 videos independently, use a LOT of Pro prompts and flash + 10 DeepThink prompts + research + image gen and more.
︀︀
︀︀But now 5video gens and it's over. Not good

Quoting Leon Lin (@LexnLin)
︀
To all Gemini subscribers:
︀︀
︀︀don't use Omni in Gemini, or use it with caution, because it eats up your limits -> use flow
︀︀
︀︀I made about 5 gens and reached my 5h limit on Ultra(200)
︀︀
︀︀About 17-19% per generation is my experience

**💬 13 🔁 3 ❤️ 143 👁️ 12.3K **

FixupX · May 19, 2026, 9:15 PM

[polar niche]

🥊

[polar niche]

Every new model is just benchmaxxin

That doesnt mean its better

Why did they release a flash variant before pro 😭

[loud herald]

THOSE CRASHES SCARED THE HELL OUT OF ME

[jovial solar]

Wow.

Early access.

[cedar obsidian]

not so early bro

[echo aurora]

Give it a try!

[dawn path]

Agent servers are dead

[echo aurora]

Agent Mode isn't working for you?

[echo aurora]

Hey @weak panther are you able to share the Eval ID for the session where you got this error without the count down timer? The Eval ID is the random set of numbers/letters when on a specific chat session. It'll look something like: 019e43c0-ff6a-78ad-af29-afd53c3b0a3f

[dawn path]

The AI responds that it cannot connect to the sandbox

[crystal mica]

gemini nevers know its versions

[main nexus]

No model knows its version AT ALL due to the knowledge cutoff

But

Gemini especially NEVER knows its version

Like at all

Even in chatting interface

[sour spear]

What's more annoying is when I have it fix some code for me, and then it proactively also "fixes" API endpoints, converting e.g. gemini-3 back to gemini-1.5, and then my API calls stop working. 😉
It's not as bad as it was with Gemini 2.5 Pro, and other models, including Claude, do this, too. But I'm still adding a special "do not change any API model endpoints", or "use this exact endpoint: " prompt elements as safeguard. And always run a quick check anyway.

[feral elm]

gpt free image creation limit is... 2 images

I need a reality check

[carmine void]

<@&1349916362595635286>

[grim palm]

Generate this image face expressionEyes: Typically rolled back, unfocused, or crossed.Mouth: Often open, with the tongue protruding or hanging out.

[marble pawn]

what can i use agent mode for?

[grim palm]

Anyone tell me how to generate imagine

[frozen swift]

fix the damn thing. no results count as one was horsecrap.

[rigid holly]

What does agent mode even do?

[feral elm]

it does things
works even
you give it a task and it tries to complete it

[surreal zephyr]

<@&1349916362595635286> @echo aurora

Anyway google partially used codex to make antigravity2

[pseudo hemlock]

Smart of them

[ocean venture]

https://tenor.com/view/higurama-son-what-is-bro-talking-about-son-we-are-cooked-higurama-we-are-cooked-jjk-son-gif-3762192617974067107

[alpine glade]

Create a cinematic animated video of a young man in a semi-realistic cartoon style. Soft lighting, smooth motion, expressive facial emotions.

Scene 1:
The man is sleeping in bed. He slowly wakes up, yawns, then stretches his back and arms. He looks tired and exhausted.

Scene 2:
He sits on the bed, resting his hand on his cheek, looking bored, stressed, and worried.

Scene 3:
He grabs a large roll of paper representing work orders. He keeps unrolling it, and it becomes longer and longer until the papers fall to the floor due to the huge number of requests. Papers scatter around him, creating a sense of pressure.

Scene 4:
He places his hand on his head, thinking deeply, clearly overwhelmed by responsibilities.

Scene 5 (Visual Effects):
Animated icons rotate around his head in a circular motion:

🚗 Car icon
⛽ Fuel / gas icon
💰 Money icon

The icons symbolize stress, expenses, and daily struggles.

Scene 6 (Camera):
Slow zoom-in on his face, showing confusion and frustration.

Dialogue (Voice-over or lip sync):
He says in a tired, confused voice:
“What should I do? I can’t find one place that provides all these needs together.”

Overall Style:

Smooth animation
Clear facial expressions
Slightly warm colors with a touch of sadness
Everyday stress and mental overload
Video length: 15–25 seconds

[velvet ledge]

Hello Arena

[shy patio]

70% players visited this place, but missed this Money Loot - RDR2
--> https://youtu.be/VeVFgXAzRRc

[ocean venture]

https://tenor.com/view/aura-aura-felt-aura-max-aura-meme-aura-reaction-gif-673194590267914254

[wicked talon]

I only made it for free ai lmao

[spring oar]

What model is the most reliabel
and less hallucinationn ?

[icy yew]

Uhhhh idk read this ig if you want simple answer ask like ais for lowest hallucination rate or sum

[azure cove]

bro

if im using gemini 3.5 flash, does it mean I am really using that or its just an ai

[icy yew]

Wdym

[azure cove]

like am I actually using gemini 3.5 flash

[icy yew]

Wait if it ai says only up until 2.0 is available how is it using the 3.5 model

Something ain't right

[azure cove]

fr

it said its not 2026

I am Gemini, a large language model built by Google.

There is currently no model called "Gemini 3.5." I am based on Google's Gemini family of models (such as Gemini 1.5).

[icy yew]

I just tested with the base Gemini it got every one right why is that one saying that
Hmmm

[azure cove]

idk

[weak granite]

India me berojgari par emotional short video, sad music, realistic scenes, Hindi captions”

[icy yew]

I have every single frontier model

Don't ask me how

[azure cove]

how

[icy yew]

It was a joke when you press like opus it says upgrade

Chill

https://tenor.com/view/calm-gif-8293915206003753096

[slender ledge]

Fonald dm

[slender thistle]

Omni is impressive

[icy yew]

That's actually so cool ngl

[slender thistle]

Here is more

[echo dome]

but i need the reason why gpt 2 image is now expensive that i don't want to buy anything

as i checked updates again

[echo dome]

which model

[slender thistle]

Omni edit

Gemini

[echo dome]

what is omniedit

I'm new on ai models

[slender thistle]

Google announced their new model yesterday

[echo dome]

okay

omniedit reminds me of omnisphere (fl studio plugin that have massive instruments)

[slender thistle]

It's video model

[icy yew]

Where can you use omni

[shrewd citrus]

Super impressive

[night moat]

😤

[weak dagger]

Dont worry i got perplexity

[pine meadow]

One of the best Qwen 3.7 or Gemini 3.5 flash

@echo aurora

[slender ledge]

Dm

[long zinc]

why am i getting a cloudflare timeout on the site

ok i refreshed and its fine 💀

[icy yew]

https://tenor.com/view/blackbeard-blackbeard-writing-taking-notes-writing-gif-17583038544116987898

[surreal zephyr]

https://giphy.com/gifs/meme-tuff-niche-hOIiB9dUoBlA5P5cIs

[pine meadow]

oh my god

[icy yew]

What

[pine meadow]

Agent mode

[icy yew]

I got it too how is it

[pine meadow]

I'm trying it out now.

[remote gulch]

Oh no.

tell me whyyyyyyyyyyyyyyyyyyyyyyyyy

BRING THEM BACK PLS!!!

[surreal zephyr]

its pretty bad tbf

[indigo knoll]

Is Gemini 3.5 better than 3.1 Pro?

[slender thistle]

Situation with Gemini

[icy yew]

Do you want arena to shutdown

[icy yew]

I know them legs broke broke

[light sleet]

hi

[icy yew]

Hello

[thin current]

Guys, I’m trying Agent Mode but I can’t upload PHP or Python files. Are they not supported?

[surreal zephyr]

no files supported

[thin current]

ZIP too

I was curious to try it, but with my projects it can’t really help me. That’s a shame

[short sluice]

not only does it not make me able to select models

BUT IT ALSO LOGGED ME OUT

all my chats are still there though

this is on each account i tried

[sharp mirage]

Yoo

Wsg

[grave peak]

https://giphy.com/gifs/9mc-smile-cat-sad-so-lGBecpB2dIMwt6ohfI

[amber rampart]

I have the same problem. And the login doesn't work either.

[thin lantern]

Seems like I am not the only one then.

[stray aspen]

lol what

how is qwen so high

[gilded umbra]

i encountered the same problem

[grave peak]

@echo aurora

[verbal current]

pineapple welp

[limber crag]

yay they finally shut down the website

[verbal current]

wth is GDDR6

[surreal zephyr]

that's like the only thing in the image that isnt satire 💀

https://en.wikipedia.org/wiki/GDDR6_SDRAM

[surreal zephyr]

most gpus use gddr6

[verbal current]

four thousand dollars

holy

[stuck walrus]

X3

[verbal current]

you have so many space

[golden ocean]

TRUE

[dry siren]

Okay at least it's not just me that got logged out

[golden ocean]

same

[tulip flume]

Everyone suffers now

[golden ocean]

what does login say?

[short sluice]

im logged in now

but error in images

[icy yew]

Whast going on

[short sluice]

i cant put images in

[icy yew]

Did the website crash or sm

[hybrid tinsel]

I cant log in

[dry siren]

Happens now and then

[thorn mantle]

Yep. Just gotta wait.

[short sluice]

nbm im logged out again

[icy yew]

Why am I logged out the hell is going on with web😭

[foggy swan]

same i cant log in

[willow swift]

salsanow go brrrr

[icy yew]

@echo aurora the website kinda broke gang

[frozen swift]

server crash

[grave peak]

CTRL+SHIFT+R

[frozen swift]

i hope they fix the limit

[icy yew]

No models left

GG website gone

[dry siren]

Sorry I ate them all

[frozen zealot]

The website shat itself

[dry siren]

I'm logged back in. Can view chats but can't do anything

[icy yew]

Even max is no longer here

[icy yew]

The moment we need pineapple he's not on the clock 😔

[thorn nebula]

What happened it's not logging in

[grave peak]

claude popus 4.7 I need this

[icy yew]

Okay it's kinda fixed now but still logged out

[vestal ridge]

I am facing an issue that is i am not able to login inot my account as i clcik the log in button its just loading then back to the screen and on sending an chat without login its showing failed to accept terms of use for my account

[icy yew]

I can't even accept terms of service

[icy yew]

Yeah website shat itself

Nothing works

[dry siren]

This happens regularly it'll probably be down for an hour or two

[robust sluice]

maybe server maintenance ?

[icy yew]

Wouldn't they have announced it though like for the next hour or two the website will be under maintenance

[hybrid tinsel]

Maybe proper daily limitation being set

[clear nova]

A historical, slightly stylized portrait of Nawab Mahabat Khan II appears.

[short sluice]

???

[robust sluice]

i dont know, just came back on the site after long time and see it dont work

[clear nova]

How to create ai video form this

[limber crag]

you cant. website is dead

[dusty wagon]

rip lmarena #packwatch

[hollow ivy]

is gpt-5.5-pro also better than opus-4.6-thinking in coding?

[robust sluice]

hope they'll came back with all the same bugs + decrease to 2 gen limit per day + bigger font size 💀

[hybrid tinsel]

Website is back
I wonder what changed

[spark linden]

Website is still super broken for me

[icy yew]

Website is fixed for me

[dry siren]

Probably some backend update balked something

It's back for me too

[echo aurora]

It’s back for you all? Can’t say I was seeing a problem

Okay cool.

Thanks.

[sleek urchin]

gpt-5.5-pro vs opus-4.6-thinking

[sharp mirage]

Hello

[spring oar]

What is this benchmark ? in Artificialanalysis GPT is better

[wicked talon]

Sup cherry

[sharp mirage]

I found it

[sleek urchin]

Did I ask you anything?

[wicked talon]

5.5 IS HOW MUCH

[hollow ivy]

so opus-4.6 > opus-4.7 in coding?

[sleek urchin]

5.5 pro*

[sharp mirage]

Goood how about you

[sleek urchin]

4.6 thinking*

[sleek urchin]

For coding its better but 4.7 uses hella more tokens

[hollow ivy]

yeah, i meant O-4.6-t vs O-4.7-t

[sleek urchin]

Lemme run benchmarks

[wicked talon]

Holy

[wicked talon]

I'm good too thank you

[hollow ivy]

and vs gpt-5.6?

[sleek urchin]

Doesnt exist

[hollow ivy]

ok, and mythos 5?

[sleek urchin]

Doesnt exist, mython isnt even released

[remote sentinel]

Hello everyone. Is there a way I can convert my text into an animated video?

[surreal zephyr]

why is gemini omni not on leaderboard

did it not get into top 10?

[hollow ivy]

so opus-4.6-thinking is still the KotH/GoaT for coding?

[icy yew]

Gpt 5.5 costs
180 BUCKS?

[surreal zephyr]

not even close, 5.5 high solos

[sleek urchin]

yup 5.5 pro*

[surreal zephyr]

thats 5.5 pro not high

[spring oar]

no i ask you a question

[sleek urchin]

Whats my private bot

[spring oar]

What is best for reliabel and less halucination for study opus 4.6 or 4.7 ?

[sleek urchin]

But 4.6 thinking uses way less tokens so that means 4.6 t is the best (For budget and coding)

[sleek urchin]

Both are good

[sharp mirage]

Bye

I gtg

[sleek urchin]

but if your a on a budget i'll go with 4.6 t

[wicked talon]

GPT solos nothing

[sharp mirage]

Cheack this out @echo aurora

[sleek urchin]

Opus ofc

[wicked talon]

Gpt can't decide which day of the week has x in it

[sleek urchin]

Gpt just lies

[sleek urchin]

October?

[hollow ivy]

@surreal zephyr you disagree, right?

why?

[sharp mirage]

Ofc opis

Opus bro

Gpt huesltioing

[sleek urchin]

https://tenor.com/view/claude-llm-ai-my-beloved-heart-locket-gif-16212068080666171534

[spring oar]

no hallucination to much

[wicked talon]

https://vm.tiktok.com/ZNRtnHWBT/

[sleek urchin]

I watched same video lol

[sleek urchin]

No? opus 4.7/4.6 dont hallucinate

[wicked talon]

https://vm.tiktok.com/ZNRtna3ub/

[sleek urchin]

Kimi 2.6 and deepseek v4 sometimes does

[spring oar]

yes is hallucinate with pdf

[sleek urchin]

pdf? I never used it

[wicked talon]

Claude is as bad as gpt

https://vm.tiktok.com/ZNRtntEEu/

[hollow ivy]

-# bump

[sleek urchin]

So whats the best then? "def not tesy random ai from huggingface"

[wicked talon]

Daily usage would be Gemini especially with android NOT GEMINI IN GOOGLE (AI OVERVIEW)

Personally I like qwen for daily usage but everyone would disagree

[hollow ivy]

-# bump2

[remote sentinel]

Is there a way I can convert my text into an animated video?

[icy yew]

Gemini is actually pretty good for me and the usage with pro models are very generous and if you do hit limit just go to ai studio and keep using the pro

[wicked talon]

Gemini pro is so good it rooted a phone

[sleek urchin]

AI and their problems:
Claude > Too expensive
GPT > Lies too much and gpt 5.5 pro IS ULTRA EXPENSIVE
Gemini > Bad models (Flash 3.5 is good but not better than models like glm-5.1 and flash costs more on openrouter)
GLM > Perfect and cheap one of the best models for coding if your on a budget
Qwen > Actually good and cheap but worse at coding than glm 5.1 (Qwen 3.7 max)
Deepseek > Hallucinates too much if context is too long

[wicked talon]

Excuse me qwen is not bad at coding

[sleek urchin]

yeah but worse than glm 5.1

[sleek urchin]

for the same price

[wicked talon]

Gpt agree
Claude agree
Gemini agree
Glm disagree
Qwen disagree
Deepseek agree

[sleek urchin]

Why?

[wicked talon]

Deepseek also is bias towards the user too much

You can say spirit is still up and it will believe you

[surreal zephyr]

aero and pjyo are alts/ragebaiters so their votes dont count

[icy yew]

I Can convince it something is true easily if I just say it or the opposite

[surreal zephyr]

hnam is random alt too

and cherry has no idea about anything

[wicked talon]

And Gemini

Gemini is a w

[sleek urchin]

For coding yeah but I would go with opus 4.6

[wicked talon]

Opus is quite good but it's not a daily assistant

[sleek urchin]

also glm 5.1 is good

[sleek urchin]

for daily assistant (Not coding) I would go with glm 5.1 or gemini 3,5 flash

[icy yew]

Gemini treats you like a king when you are on the free plan while I can't send images to gpt for more then 3 times

[wicked talon]

GLM hell nah

It doesn't even have a app

[sleek urchin]

its a api

[wicked talon]

I've never personally tested GLM fully

I've used it a lil and it was bad

[sleek urchin]

I use glm 4.7 AIR thats free on openrouter and its just perfect

[hollow ivy]

so gpt-5.5-xhigh is only good for certain coding tasks, right?

[icy yew]

Used it a lil when glm 5 was rolling out wasn't really impressed ngl

[sleek urchin]

maybe

[sleek urchin]

glm 5? meh that sucks 4.7 and 5.1 are goated models for budget

[wicked talon]

Isn't qwen ranked best Chinese ai model?

[sleek urchin]

Also glm 5.1 is slow but free on nvidia

[obsidian cargo]

So wait, Gemini 3.1 pro is available in direct chat but not 3.5 flash??? What???

[sleek urchin]

GLM

[sleek urchin]

Also its opensource

[sleek urchin]

what? I think you misplaced

[sleek urchin]

opus ofc

Gpt 5.5 is overrated

but cheaper than claude

[spring oar]

No broopus hallucinate

[sleek urchin]

I never saw it hallucinate

In coding

pdf's only

[obsidian cargo]

Nope.

[spring oar]

not in coding for learning stuff

[sleek urchin]

Scroll a bit and sorry I didnt know 3.1 is in direct

[obsidian cargo]

@echo aurora Is this intended??

[sleek urchin]

I think it just sorts on a-b-c

[obsidian cargo]

I did. All the way to the bottom.

[sleek urchin]

scroll a bit or enter a full model name eg. Qwen 3.7

[obsidian cargo]

Oh what huh it is there when I searched up gemini-3.5, so the search must be broken???

[carmine scaffold]

hii

[sharp mirage]

wlc back

[sharp mirage]

yeah ig today the search was broken

bro

at what ?

[hollow ivy]

i will believe that, if at least 50 people have participated in that poll and 4.7 got at least 60% of all votes

[sharp mirage]

no

4.6 is better in coding

[hollow ivy]

i tend to agree with you

(but i'm unsure, therefore the poll)

[sharp mirage]

the only thing that 4.7 is better is the termnal codes

and better typer

and also yiper

[hollow ivy]

60% of all votes, not just of the comparison votes
this is a difference, because there is a third option (equal)

so 30/50 have to vote for 4.7, for me to be convinced

[sleek urchin]

But opus 4.7/6 cost same

[sharp mirage]

half who voted havent test 4.7

so when you try it uk

[hollow ivy]

no politics
rule #5

[sharp mirage]

no

[sleek urchin]

for coding glm

[sharp mirage]

yeah

and for learing i also find it better then gpt,claude,and gemini 3.0 flash

[hollow ivy]

the text is from AI

probably Opus-4.6

good idea

[sharp mirage]

yeah

good idea go on its fun ngl

[pastel mesa]

Damn guys
Now is like 10 euro at 1M tokens for gemini flash 3.5 and when is Pro out?

[sharp mirage]

no one know

[sharp mirage]

yet

[pastel mesa]

Like mid 2026

[sharp mirage]

wth

how opus 4.6 is faster ?

[wicked talon]

Who knows for sure, but it reads exactly like generic political boilerplate. Whether an Al or a human speechwriter shat it out, it's the exact same flavor of bland, focus-grouped word salad. - Qwen

[sleek urchin]

It uses way less tokens

[sharp mirage]

it look like website gen not ai

[sharp mirage]

but it dont mean its faster

[hollow ivy]

text is done nowadays

[sleek urchin]

4.7 has a fast model on openrouter
https://openrouter.ai/anthropic/claude-opus-4.7-fast
same with 4.6

[sharp mirage]

but yall forgot something the adptive thinking got to all ais

sp

[sleek urchin]

150/m output kinda cheap

compares to gpt 5.5 pro

[sharp mirage]

yeah

ig the best low price ai is gemini

[pastel mesa]

148 euro for 1M tok?

[sharp mirage]

combpred to gpt

[wicked talon]

Qwen ai is

[sharp mirage]

qwen is not that bad

but gemini is best for all things like videos img gen

and you know

btw

[sleek urchin]

Qwen isnt bad its just bad at coding compared to glm

[sharp mirage]

in google gemini app does 3.5 flash have limt tho ?

[sleek urchin]

For images,videos (Paid) and daily assistant yes for coding no

[sharp mirage]

if its cli

[proud bobcat]

Not at all

It’s quite good for what it is

[sharp mirage]

yeah

[inland wigeon]

I remember you could previously produce videos on this channel. Has this feature been removed?

[proud bobcat]

You get very competitive performance for decent price

[clever glade]

and its qwen 3.5 497b thats saying that

[sharp mirage]

money : 💸

[proud bobcat]

Price
Versatility
Better line of thinking

[wicked talon]

Cheaper

[proud bobcat]

I’ve always found Qwen has better thinking than the new opus

[clever glade]

2 year reasoning loop

[hollow ivy]

(the current best: including every AI model you heard about)

[sharp mirage]

better then the adptive thinking thing its so bad bro

[wicked talon]

Parameters don't always equal intelligence

[pastel mesa]

Some days will conquer the world

[sleek urchin]

opus 4.7 thinking for coding, daily assistant: Gemini 3.5 flash,best budget coding > Glm5.1, gpt is trash

[proud bobcat]

When it isn’t drunk and hallucinating

Sure

[sharp mirage]

gemini 3.5 , gpt 5.5

[inland wigeon]

Has the video creation feature been removed?

[clever glade]

qwen 3.5 has crazy reasoning loop and overthinking issues

[sharp mirage]

no its not

[sharp mirage]

if you know how to prompt no

[pastel mesa]

Deepseek V4 flash thinking is better than deepseek V4 non-thinking

[wicked talon]

No?

[sleek urchin]

gpt???

[proud bobcat]

I’ll just say it

I’ve never had a good experience with GPT

Constantly made stuff up

[sleek urchin]

Same

[sharp mirage]

nvm i wid say deepseek v4

[inland wigeon]

How can I make a video?
Which section should I go to?

[wicked talon]

I mean it ain't wrong

[clever glade]

yes with reasoning off ofc

[sharp mirage]

battle mode

[pastel mesa]

MrMAIE Code Ai?

[proud bobcat]

I’ve never had an issue with Qwen overthinking

That’s more of a DeepSeek thing

[sleek urchin]

For people thats not in ai stuff like WE are

1. ChatGPT
2. Gemini
3. Claude
4. Grok (They dont know what this is)

[proud bobcat]

Grok straight garbage bro

[sleek urchin]

fr

[proud bobcat]

https://tenor.com/view/joe-biden-what-gif-3784745843284049350

[sleek urchin]

No, deepseek is

[clever glade]

chatgpt for paid gemini and deepseek for free

[proud bobcat]

DeepSeek is great too

[sharp mirage]

but at least its not a glazer

[proud bobcat]

Been using it for math

Its vision is strong

[proud bobcat]

Not Gemini level but it’s still strong

[sleek urchin]

Uncensored prompt detected

[hollow ivy]

and next year, M.S. will join the foray ^^

[clever glade]

how do you even get deepseek vision model

[wicked talon]

Lmao it's just qwen mad at me

[hollow ivy]

by FAIR

or was it MSL?

[proud bobcat]

It’s in testing

[hollow ivy]

(Meta "superintelligence" labs)

[clever glade]

what is your region

[sharp mirage]

gemini

[hollow ivy]

Muse Spark

[hollow ivy]

-# in case any1 wondered

[proud bobcat]

US

[sleek urchin]

LOL "grok IS THE BEST"

[hollow ivy]

gork sukx ^^

[wicked talon]

At least it ain't glazing itself

[proud bobcat]

Jokes aside that Gemini one is too true

[sharp mirage]

fr

[proud bobcat]

It’s so over the top with censorship

[sharp mirage]

he dont glaze his self at leasssst

[clever glade]

insert meme of grok being biased towards musk

[wicked talon]

Also this is for "most people" meaning daily usage

[sleek urchin]

Real ranking

1. Claude
2. Gemini
3. Grok/Chatgpt (Grok is bad and gpt lies)

[sharp mirage]

not likee some aiiiiis not chatgpt ofc

[proud bobcat]

Kimi is underrated

[sleek urchin]

"The real ranking for anyone with a brain"

[proud bobcat]

Easily the best search ai I’ve used

[hollow ivy]

i wonder, if we will hear from Ilya ever again

[sleek urchin]

hallucinates too much (2.6 thinking)

[sleek urchin]

mah

[proud bobcat]

If you know how to use it it’s very nice