general

[robust yoke]

Could request for them to add o3 Pro.

[remote idol]

Isn't o3 like the top tier coding model?

and o1

[robust yoke]

People say they are, yeah.

Claude 4.5 Sonnet is too, apparently.

[remote idol]

Crazy how gemini 2.5 pro actually byp assed the other anti viruses

(for a illegal code)

[robust yoke]

Yeah, heh heh.

[hasty mountain]

Yo guys how do i use sora

Ai

????

[robust yoke]

Here: https://sora.chatgpt.com.

[plucky yacht]

hello

[robust yoke]

Greetings.

[plucky ember]

Hey guys I was wondering are we able to download the videos we generated from here ??

[robust yoke]

Yes.

There ain't any rules that say you can't, so...

[plucky ember]

How's it going darkness good to see you again

[robust yoke]

Going good, and you?

[plucky ember]

I was trying to but its not showing up I sent it to my drive and it shows up as a file

[robust yoke]

Hmm...

Do you have a PC?

[plucky ember]

No everything I do is on my android

[robust yoke]

Ah.

Try downloading the file and renaming it to "(recording name here).mp4".

That usually does the trick.

[plucky ember]

Hmm thanks for the advice I will try it right away

[robust yoke]

Alrighty.

[fossil socket]

Uh, any idea what happened?

The models got converted to assistants?

[robust yoke]

No, that's just what happens when a model you've used before gets deleted from LM Arena.

The same thing happened to me with Seedream 4.0, and it showed the same icon.

For some odd reason, they get converted to assistants.

[fossil socket]

Well I was using 4o latest, it got updated?

[robust yoke]

I think it got renamed is what happened.

[fossil socket]

Well that's mildly annoying

[robust yoke]

Yeah, you don't say.

[fossil socket]

Oh well

[plucky ember]

I'm not sure what I am doing wrong when you save video is that the same as downloading??

[robust yoke]

Yes.

"Saving" in this context quite literally just means downloading.

[plucky ember]

😂 lol

This is how it looks for me 1 second

Nevermind it worked 😊 thanks for the bigg help

[robust yoke]

My pleasure.

[plucky ember]

Now I can try and finish up this project before I end up falling asleep on cap cut again lol

[robust yoke]

Well, good luck to you then, heh heh.

[plucky ember]

Thank Im going to need it

[robust yoke]

I'm sure you will.

My pleasure.

[plucky ember]

I Salute u over and out

[robust yoke]

Likewise.

[gusty frigate]

Hey guys, just thought about sharing perplexity pro with the students here (as I found I had a referral link) https://plex.it/referrals/V9AOVPP3

So feel free to grab this chance if you are a student (and again, being pretty open about it being a referral link)

[agile kernel]

@gusty frigate
I’m a student of class 9, so how can I get free access to it?

[gusty frigate]

Yes, you just need a valid email id with respect to your university/school domain

[agile kernel]

I don’t see my school’s name listed there😭

[robust yoke]

Neither do I, heh heh.

Mine's not elligible.

[agile kernel]

🥲

[robust yoke]

Heh.

https://tenor.com/view/please7tv-please-beg-pray-hope-gif-3860103425950934673

[agile kernel]

I don’t have access to the email either

🙂🙂🙂🤙

Bruhh

[robust yoke]

Oof.

[agile kernel]

🥲

How can i get it

I need it

👊

[cedar valve]

/video

a giant cat walks between old russian houses; drone shot

[robust yoke]

Wrong channel for that.

You can go here to do that: #video-arena-1.

[cedar valve]

thanks

[robust yoke]

My pleasure.

[spice timber]

how do i use soraa in the video arena??

[robust yoke]

You'll have to keep sending your prompt until you get it.

It's random chance.

[gusty frigate]

Damn, thats u fortunate

Unfortunate*

[agile kernel]

🙂

Then it is impossible for me ?

[gusty frigate]

It does actually require a valid email id, so unfortunately you wont be able to get it due to their validation policies. Sorry mate

[agile kernel]

🥲

[worn lagoon]

hiiii

[robust yoke]

Greetings.

[neon idol]

Greetings chat

[neon idol]

Hey hru?

[sweet aspen]

در شب ستاره ها در اسمان میدرخشن ستاره دنباله دار ودر حرکت است

[neon idol]

Dont bomb me please 😢

[neon idol]

https://tenor.com/bGVYI.gif

[verbal nimbus]

Thanks :)

[outer spear]

less go i was right i knew theyd get sora eventually

[barren scroll]

hello

[weary thistle]

Just a thought: with the coming of image video generators, people are less and less sharing their "findings" on how they are using LLMs other than image and video generating. I find it hard to find other ressources. Even on Twitter FB Discord.

[robust yoke]

Greetings.

[robust yoke]

I've noticed this trend as well, and I share your concern. It does seem that the discourse around AI has become increasingly dominated by image and video generation tools, while discussions about innovative LLM applications in other domains have grown quieter.

I suspect this shift is partly due to the visual nature of those outputs. They're immediately shareable and impressive, which makes them perfect for social media engagement. Text-based LLM applications, while often more practically valuable, simply don't generate the same viral appeal.

That said, I've found that quality discussions about non-visual LLM use cases still exist, though they require more deliberate searching. Academic Twitter, specialized Discord servers focused on AI development rather than content creation, and platforms like GitHub discussions or certain subreddits tend to have more substantive conversations about prompt engineering, workflow automation, and creative LLM applications.

[agile kernel]

Bruh🥲

[dreamy veldt]

Yup,not availaible for Linux

Sam has a message for you all

[vernal eagle]

we know it's fake

[verbal nimbus]

Thanks :)

I like how much effort they put into the UI

there's even music lol

[dreamy veldt]

So,why will i provoke it's real

[verbal nimbus]

Ngl this UI is 🔥

I wonder what UI framework they're using

[dreamy veldt]

Comet doesn't have google lens

[vernal eagle]

I noticed that when generating videos for free with this bot it always uses a different model each time

is that correct?

[verbal nimbus]

I wish Windows had it

Probably can make a small app via the free API ig

[dreamy veldt]

Yaa,like we have Circle to Search in android

But every browser must have this feature

[verbal nimbus]

I wonder how it compares to Gemini in Chrome

[verbal nimbus]

Lens is really useful for selecting text too

[dreamy veldt]

But Chrome hasn't released its agentic feature in all the countries like comet

It's not available in India

[verbal nimbus]

It can kinda do it

[dreamy veldt]

I have a bug in my comet browser

It never finishes checking for updates

[verbal nimbus]

Oof

[dreamy veldt]

And also in agentic assistant there is no mention of the site on which iam

[verbal nimbus]

mine does

Is it supposed to be able to use the browser?

[dreamy veldt]

Yaa ofcourse man.That is what it is made for

[verbal nimbus]

Some expensive models here... wonder what the limits are 🤔

[verbal nimbus]

It told me it couldn't

[whole swallow]

Use google for search

Assistant can control the browser, tell him “you can do it, you already did it before”

[verbal nimbus]

I'll try again

[whole swallow]

With pro, none

[verbal nimbus]

Well that's a bit crazy :O

[robust yoke]

I tried to redeem the code, but it just brought me to the homepage.

Heh heh.

[whole swallow]

I think only new users

[robust yoke]

Well darn.

[whole swallow]

Use a different email

[verbal nimbus]

Doesn't seem to have applied it

[whole swallow]

Try claiming it again

[verbal nimbus]

It worked

[whole swallow]

Good :)

[verbal nimbus]

Thanks for the invite :)

[nimble atlas]

so on website it's just giving out retry and clear most of the time these days

[dreamy veldt]

You are on perplexity pro,so you will be able to use it

[verbal nimbus]

This seems handy

[verbal nimbus]

Yeah from his invite link

[dreamy veldt]

Always start prompt with "Act as an agent" or "Take control over the browser"

[whole swallow]

Yes you can create custom commands with a predetermined model

I did it to create like /claude with claude 4.5

[verbal nimbus]

Is it possible to see what it's doing?

Oh I had to click on "reasoning"

Wow this is cool

It's actually using Google Maps haha

[dreamy veldt]

Yaa that's how u use it

[verbal nimbus]

I'm asking it to look at Google Maps reviews for New York hotdog stands 🤣

[proud pine]

Katrina Kaif & Alia Bhatt

Katrina: AI ne bola hum best friends hain movie mein!
Alia: Perfect! Par wardrobe kisne decide kiya?
Katrina: AI ne! Aur mere outfit ka colour tera mood dekhke badalta hai! 😂

[verbal nimbus]

It's really slow but I guess it is making progress

It's reviewing like 1-2 hotdog stands per minute hahahaha

[robust yoke]

Saluran yang salah.

[verbal nimbus]

It keeps accidentally opening the review box

Looks like it stopped at 5 hot dog stands

[robust yoke]

Yeah, it's pretty neat.

[verbal nimbus]

Yeah if it can do all this in the background while I do something else then that's pretty neat

[robust yoke]

It can, actually.

[verbal nimbus]

Any way to tell what model it's using currently?

[robust yoke]

If you switch tabs, it'll continue.

[verbal nimbus]

Yeah

[robust yoke]

I asked it to complete The Password Game.

It got pretty far.

[verbal nimbus]

Idk what it's doing now, it's stuck on done

[robust yoke]

Usually, the way to deal with that is starting a new convo, then going back to the other one.

[whole swallow]

down in the bottom click this

[verbal nimbus]

comes back 1 hour later
notices that my account left a 5 star review on OpenAI app
"What the heck??"

jk

[verbal nimbus]

Ah it's not showing yet

Previous one used "Pro Search"

[whole swallow]

Use /command to invoke the model youd lke

then tell him what to do

[verbal nimbus]

I had to click skip remaining steps to unfreeze it

It was the Pro Search model

wonder what that is 🤔

[whole swallow]

Btw guys where you from?

[robust yoke]

It just searches more in-depth.

[robust yoke]

Washington, and you?

[verbal nimbus]

Oh it wasn't using Pro Search

It seems like it forgot about hotdogs :p

hotdog prompt disappeared

[whole swallow]

Damnn im from italy

[neon idol]

Damn i want to try it but i dont use pc 😢

[neon idol]

Anche io!

[robust yoke]

Oh dang.

[whole swallow]

https://tenor.com/view/social-credit-credit-social-уважение-плюс-уважение-gif-1626328442317885176

[robust yoke]

The timing is phenomenal.

[neon idol]

Vero

[whole swallow]

great hahha

[robust yoke]

Heh.

[neon idol]

https://tenor.com/bJcdl.gif

[robust yoke]

Neither do I, to be honest.

I'm a vampire.

[neon idol]

Damn

[robust yoke]

I sleep in the daytime, and I wake in the night.

That's why my name is what it is.

[whole sundial]

if you really want to abuse your pro sub, you can also do image gen (including with gpt-image-1) and even video gen apparently (idk what they use though)

[verbal nimbus]

Perplexity research mode seems good

[whole sundial]

quality is higher than what lmarena uses, no melting eyes here

they also have nano banana and seedream 4 for people who like those models (there are many people here who like them)

[remote arrow]

🤔

[dense canopy]

Is this bot completely free or are there limits? @tiny palm

[whole sundial]

5 videos per day, but it is free

[dense canopy]

I see

[whole sundial]

but each turn is 2 videos, so really 10 videos per day

[dense canopy]

Yeah but each time is a different model

[verbal nimbus]

Why does it write like this lol (writing inside Python instead of in chat)

[dense canopy]

(I keep changing my pfp my bad)

[whole sundial]

true, they don't have model selection like on the website

[verbal nimbus]

Final answer seems good though, although it kinda left out how to do it

Although I wonder if GPT-5 Thinking in normal mode is better than Research mode (Research mode doesn't let you select model)

Not sure I really trust anything but GPT-5 for complex queries

I've seen how Sonnet can get things wrong, lol

Ok, just ran it with GPT-5 Thinking in Search vs. (unknown model) in Research mode

Both read a bunch of research articles but came up with a different conclusion

I think I'll trust GPT-5 Thinking on this one 🤔

[whole sundial]

probably sonar deep research

they also bundle $5 of monthly api credits with pro for use with their models

[whole swallow]

https://tenor.com/view/spongebob-squarepants-patrick-star-patrick-animation-cartoons-gif-16276467

[verbal nimbus]

The model really needs the ability to reason properly, I think GPT-5 Thinking is the only model rn that can

[whole swallow]

teag

Perplexity goat

[verbal nimbus]

Like you can give Sonnet all the information in the world, but it still can't figure out how to fix complex race condition haha (easily solvable by CS college students)

Both GPT-5 and Sonnet fail at debugging complex code issues too. A human would remove a huge chunk of code, then add it back line by line to see what introduces the error (if debugger isn't possible), but none of the models did that.

[red basin]

Who want sora 2 codes?

[neon idol]

me

[lime stratus]

https://youtube.com/shorts/eQGLbSxek1c?si=dtRbPqGNhyl5DBTu
MADE WITH LMARENA

[verbal nimbus]

Did Comet just click the button 28000 times lol

[neon idol]

lmao

[fiery gull]

You get the sora 2 code?

[neon idol]

nah

is a scam

belive me

[leaden sun]

i believe this is the limit of the architecture, models dont really have a "sense of understanding" like humans do, they just find patterns, though, sonnet used to be able to conclude using associative logical argumentation, rare but I've observed it in the earlier days (beginning of this year), Anthropic made it worse with their safety obsession and destroyed the intelligence in the process, I feel

i think the way how you described debugging works only for certain languages?

[fiery gull]

@echo aurora

[brazen dragon]

wasssupppp

[fiery gull]

I'm good and U

Nice pfp

[brazen dragon]

im new here. this discord is awesome!

[fiery gull]

What did you create that was interesting with AI? I want to see

[dusty eagle]

Guys, is there an AI that can add 5 seconds to a video? Let me explain: A friend of mine sent me a 15-second video of him fishing, but the video ends just as the fish is about to come in. Is there a way to recreate that moment with an AI?

[fiery gull]

Dude from what I remember I don't remember any AI that see the previous video and it continues I just remember the last frame use

Downloading lfm2 8b a1b iq2xxs I hope this work like the 30b iq2xxs

[wicked sage]

sora 2 added/??

[robust yoke]

Yes.

[fiery gull]

And 2 pro

[robust yoke]

I could give you a code once I get one.

[neon idol]

Nah dont worry i have it

I was only doing a scam check

[robust yoke]

Ah, alrighty.

[glossy umbra]

Once I get one. 😅

[robust yoke]

Do you need one?

[prime mulch]

I got one but i can't use that so I gave to my friend he is from usa

[glossy umbra]

No, dont worrry man. I was highlighting the bold social priority hierarchy by the italic "I" that you imposed.

[robust yoke]

Ah.

Not sure what that is, but neat.

[robust yoke]

Makes sense.

[prime mulch]

5VB06D

[prime mulch]

This is that code

[robust yoke]

How come you can't use it though?

[prime mulch]

I'm from India sora is not available here

[robust yoke]

Well, you could just use a VPN.

[prime mulch]

If I use that using vpn it's dead slow

[robust yoke]

Ah.

Try a fast country then, like the U.S.

[prime mulch]

It's no use now I gave code to another guy now i don't have one

[robust yoke]

I can give you one.

At least, I think it still works anyway...

X6N8BR

I wouldn't be surprised if it doesn't.

[glossy umbra]

I have an invite code

Just joined it.

[robust yoke]

Neat.

[prime mulch]

Got it thanks darkness

[robust yoke]

My pleasure.

Didn't think it'd work.

[prime mulch]

Dm

[robust yoke]

Ah.

[pale imp]

Hey, anyone with research level knowledge?

[robust yoke]

Hmm…

Maybe me?

[pale imp]

How far you've gone?

[robust yoke]

12th grade.

Vous pouvez utiliser des vidéos dans le canal #video-arena-1.

[pale imp]

No, I mean like how much in research? What topic?

[robust yoke]

Anatomy.

[stone coral]

https://www.threads.com/@dj.apocalypse007/post/DPi2cjPiCCH?xmt=AQF0kGMNfSPSgV_d6V6zCa2sdppQIchwWZQ6dh1UlWkNug&slof=1

[pale imp]

Ai anatomy?

[robust yoke]

No, human anatomy.

[sharp mirage]

yooo

i camed back from schoool finnnallly

[pale imp]

Oh, was asking about how much research level ai knowledge you have?

[robust yoke]

Yeah, figures.

I have a bit of knowledge.

In terms of prompting and stuff.

[robust yoke]

Welcome back.

[pale imp]

I was asking about research level. Like where you have proofs of why model works. Work on new ideas, etc

[robust yoke]

Hmm…

[tall summit]

how about you state why you want someone with "research level knowledge"

[robust yoke]

Yeah.

[sharp mirage]

thx

[pale imp]

Just wanna know, maybe get some advice about some stat

[robust yoke]

My pleasure.

[robust yoke]

What kinds of stats?

[sharp mirage]

advice for wjat >

[robust yoke]

Depending on what kinds, I could likely help.

I'm pretty good at research.

[pale imp]

I am confused about CNNs, transformers, etc. And wanna know how much math do I need

[robust yoke]

Well, lemme go ahead and check…

[pale imp]

Aren't you good at research? Which topics you proficient in?

[sharp mirage]

bro i cant coding i want to sleep :

[robust yoke]

I'm good at researching why things happen.

Like why something leads to an error.

[stone coral]

https://www.threads.com/@dj.apocalypse007/post/DPi2cjPiCCH?xmt=AQF0kGMNfSPSgV_d6V6zCa2sdppQIchwWZQ6dh1UlWkNug&slof=1

[pale imp]

Oh, okay

[pale imp]

I was talking about phd level ai concepts

[robust yoke]

From Grok:

Math Prerequisites for Understanding CNNs and Transformers

Convolutional Neural Networks (CNNs) and Transformers are core architectures in deep learning, but the math you need depends on your goal: high-level intuition (e.g., for application or deployment), implementation from scratch, or research-level modifications. I'll break it down by key math areas, with specifics for each model. In general, you don't need a PhD-level background—most practitioners get by with undergraduate-level linear algebra, calculus, and probability. Tools like PyTorch or TensorFlow abstract much of the math, but understanding it helps debug and innovate.

1. Linear Algebra (Essential for Both)

• Why? Neural networks are fundamentally matrix operations. Weights, inputs, and outputs are vectors/matrices/tensors.
• Key Concepts:
  • Vectors and matrices: Dot products, matrix multiplication, transposition.
  • Tensors: Higher-dimensional arrays (e.g., images as 3D tensors in CNNs).
  • Norms (e.g., L2 norm for regularization) and eigenvalues (for understanding PCA or optimization stability).
• For CNNs: Convolutions are sliding matrix multiplications (kernels over feature maps). Pooling involves downsampling matrices.
• For Transformers: Self-attention uses scaled dot-product attention: ( \text{Attention}(Q, K, V) = \text{softmax}\left(\frac{QK^T}{\sqrt{d_k}}\right)V ), which is pure matrix math (Q=queries, K=keys, V=values).
• How Much? Intermediate: Comfort with 2D/3D matrix ops. If you're implementing, know broadcasting and einsum notation. Resources: Khan Academy or "Linear Algebra and Its Applications" by Strang.

[sharp mirage]

i made lmarena extension how i can share it ?

[robust yoke]

2. Calculus (Essential for Training)

• Why? Training via backpropagation requires computing gradients to minimize loss.
• Key Concepts:
  • Derivatives and partial derivatives: Chain rule for propagating errors.
  • Multivariable calculus: Gradients (vector of partials) and Hessians (for second-order methods like Newton's, though rare in practice).
  • Optimization: Gradient descent variants (e.g., SGD, Adam) and learning rates.
• For CNNs: Gradients through convolutions (use convolution theorem or direct computation). Activation functions like ReLU (simple piecewise) or sigmoid/tanh (involve exponentials and derivatives).
• For Transformers: Similar to CNNs, but attention gradients involve softmax (derivative is Jacobian of softmax). Positional encodings use sine/cosine functions, so basic trig derivatives help.
• How Much? Beginner to intermediate: Focus on single-variable calc for intuition, then partials for backprop. You rarely compute by hand—libraries do it. Skip advanced topics like tensor calculus unless doing geometric DL. Resources: 3Blue1Brown's "Essence of Calculus" series.

[sharp mirage]

i can ask chatgpt but i am : leazy

[robust yoke]

3. Probability and Statistics (Important for Interpretation and Loss)

• Why? Models output probabilities (e.g., classifications), and training involves statistical losses.
• Key Concepts:
  • Probability distributions: Softmax for turning logits into probabilities.
  • Expectation, variance: For understanding uncertainty or regularization (e.g., dropout in both models).
  • Loss functions: Cross-entropy (KL divergence between distributions) for classification; MSE for regression.
• For CNNs: Often used for image classification, so cross-entropy loss. Batch normalization involves mean/variance stats.
• For Transformers: Heavy on sequences (e.g., NLP), so probabilistic modeling of tokens. Attention weights are like soft probabilities over positions.
• How Much? Beginner: Know basics like Bayes' theorem and common distributions (Bernoulli, Gaussian). Intermediate for evaluating models (e.g., perplexity in language models). Resources: "Introduction to Probability" by Blitzstein.

4. Other Math (Nice-to-Have or Advanced)

• Discrete Math/Graph Theory: For Transformers, think of attention as graphs (nodes= tokens, edges=attention weights). Basic sequences and recurrence for RNN precursors.
• Fourier Analysis: Optional for CNNs (convolutions relate to FFT for efficiency) or Transformers (spectral methods in some variants).
• Information Theory: Entropy and mutual information for understanding attention or compression in models.
• How Much? Minimal for starters. Dive in if researching (e.g., analyzing why Transformers scale better than CNNs via expressivity proofs).

Level Breakdown by Goal

• High-Level Understanding (e.g., using pre-trained models): Basic linear algebra (matrices as "grids of numbers") and calculus (gradients as "slopes"). Probability for what outputs mean. ~1-2 months of self-study.
• Implementation (e.g., coding in PyTorch): Intermediate in all three areas. Understand how ops like torch.conv2d or nn.MultiheadAttention work under the hood. Practice by deriving backprop for a simple layer.
• Research/Optimization: Advanced: Stochastic processes (for sampling in Transformers), convex optimization, and numerical stability (e.g., avoiding vanishing gradients).

Tips to Get Started

• Courses: Andrew Ng's Deep Learning Specialization (Coursera) covers the essentials without heavy math. Fast.ai for practical intuition.
• Books: "Deep Learning" by Goodfellow et al. (free online)—Chapters 6-9 for CNNs/Transformers math.
• Practice: Implement a toy CNN (e.g., MNIST classifier) and Transformer (e.g., simple attention block) while peeking at the math. Use NumPy for low-level math before frameworks.
• Common Pitfall: Don't get stuck on proofs—focus on computation graphs. Most "math" is pattern-matching once you see examples.

If you're coming from a non-math background, start with visuals (e.g., Distill.pub articles on attention). The field evolves fast, but these foundations haven't changed much since 2012 (AlexNet for CNNs) and 2017 (Transformers paper). What's your background or specific goal? I can refine this.

[robust yoke]

In this channel: #ai-creations.

[sharp mirage]

i mean on google extension site

:

it looks dump but i am trying to fix it but

but i cant i want sleep and i dont want to waste time

[robust yoke]

You can make your own by doing the following:
Sharing a Chrome extension you've created can be done in two main ways: privately (direct file sharing) for quick distribution to individuals or small groups, or publicly (via the Chrome Web Store) for broader reach and easier installation. The public method requires approval from Google and is recommended for wider sharing. Below, I'll outline both approaches. Ensure your extension complies with Chrome's policies (e.g., no malware, proper permissions) to avoid issues.

Prerequisites

Your extension is fully developed and tested. It should include:
A manifest.json file (version 3 is recommended for new extensions).
Icons, scripts, HTML/CSS files, and any other assets in a single folder.
Test it locally by loading it unpacked in Chrome (go to chrome://extensions/, enable "Developer mode," and click "Load unpacked").
You have a Google account.

[robust yoke]

Just vibe code, then.

Get Claude to fix it up, maybe add some neat aesthetics to it.

And boom.

Claude is actually very creative.

[sharp mirage]

i am gone use the Chrome Web Store

[sharp mirage]

thanks for infos

[robust yoke]

My pleasure.

[robust yoke]

Alright.

[blissful pawn]

pls ER i need invite code sora

[robust yoke]

You can use this one: #general message

If it works, that is.

[blissful pawn]

thank you

[robust yoke]

My pleasure.

[grizzled crescent]

thanks

[pale imp]

Already knew that but, thanks man.

[floral urchin]

Hello - random post here but would like to share my appreciation for LMArena, what a great platform ❤️ only found you guys last month when I was looking into Nano Banana, and at first mostly used the direct chat. But also really have grown to appreciate battle mode to discover new models for various objectives (that first version of seedream high res was amazing!). Today I had a hard time with Nano Banana + reference image to make certain changes to the scenery, and tried battle mode this morning and got some very useful images from Reve-v1. It's really great to be able to access so many tools in one place for free. Very much appreciated ❤️

[sharp mirage]

what about glm and deepseek ?

[unborn aspen]

Hey

[quiet pollen]

Any idea what's shasta?

Cooking hard

[sharp mirage]

?

nope

[wicked sage]

hi

[quiet pollen]

Why

[narrow blaze]

Broo @echo aurora when we will be able to select the models of ai video generators

[fiery gull]

gemini 3? joke ;-;

[fiery gull]

Never, the lmarena is a benchmark, Idk why the chatbot have a direct chat

[pliant wren]

g

[quiet pollen]

What happened to the side by side chats

[topaz basalt]

Hello !

[fiery gull]

just add 2 model

in the first message

[quiet pollen]

Oh thank you!

[fiery gull]

how many cooking hard?

[fierce cove]

Has the GPT-5 high-new-system-prompt been deleted?

[unreal plume]

Hi everyone,LMArena users 🤘🏼🫡👍🏼

[native dew]

hi

[silver radish]

hi iwns

[silver radish]

Gemini 3????

I hope it's coming out tomorrow

[opal cedar]

daily limit how much

[silver radish]

If it's not, I will literally crash out

It's kind of a 50/50

[random robin]

hi im manny just happy to be here

[silver radish]

Google is not saying anything

[fiery gull]

why will you crash out?

[keen beacon]

hello

[fiery gull]

hi

[fiery gull]

maybe the gemini 3 is coming tomorrow 👀 I think no, but have a possibility

[unborn warren]

hi

[heavy juniper]

hi all

i noticed recently some changes

kimi ai is gone now from lmarena which is sad

also they deleted the mode selector and moved the model selector from the top left and replace it with text its the most ugly ui change in ai since chatgpt released 3 years ago

[daring rock]

Please head to #1397655624103493813 for a detailed guide on how to use the bot

[patent minnow]

Hello! from spain

[echo aurora]

If/when we move Video Arena to the site is likely when something like that will happen. But yeah this request we too think would be nice to have for sure.

[neon idol]

hiiii hru?

[echo aurora]

Hello just visited your country for the first time recently and absolutely loved it. Fantastic place, very excited to go back.

[echo aurora]

Good morning (afternoon/evening)!

[neon idol]

did j visited Italy?

[echo aurora]

Doing well, hbu

[neon idol]

im good thx!

[echo aurora]

Yup! In different trips! Loved that too!

Thinking Milan + Lake Como are the next two vaca spots

[neon idol]

yeeee

gg

[final tide]

A humanoid lion with a glowing golden mane stands on a cracked street, roaring fiercely as dust swirls around him. Two muscular cavemen with primitive weapons approach from the shadows, ready to attack. The city burns behind them.

[echo aurora]

@final tide check out #1397655624103493813 for info on how to use the bot.

[heady nest]

Hello happy to be here and learn

[echo aurora]

welcome welcome!

[sage beacon]

How many image uploads do you get per day using lmarena

[echo aurora]

Are you referring to video arena?

[sage beacon]

No just to the website

where you can talk to all kind of diffrent ais

and are the ai's slower than if you actually use them on the real website

[echo aurora]

oh gotcha, it's going to vary

each model has its own rate limit

[robust yoke]

Whatever you say.

[robust yoke]

What about 'em?

[sharp mirage]

Em ?

I have no idea about it

[robust yoke]

It's a shortening of “them”.

[sage beacon]

But like for gpt what is it around is it like 5 or is it 100

[serene fiber]

i would like to compare generated video

[narrow blaze]

Okay thanks for reply and I would love to see if u also add sora 2 image to video ❤️‍🔥

[young barn]

How can I do like a part 1,2,3 on text to video

[echo aurora]

Yes! We do want to add that, at the moment there is some issues with the ratio output that we're worried would bias the votes, so we've decided to remove that for now, but yeah we're looking into solutions.

[echo aurora]

We do actually have that info publicly available by model, that would be nice though so we're thinking about it

[neat apex]

In what exactly the Glm 4.6 is soo good that it be the best in some areas? Its just a great text/code interpretation at all?

[echo aurora]

Be srue to check out #1397655624103493813

[echo aurora]

There isn't rly a way to do that sorry to say

[sage beacon]

Where is it

[echo aurora]

oh sorry don't*

[sage beacon]

or did you mean don't

oh

[echo aurora]

yeah sry

[stone crown]

Hello everybody!!!

[narrow blaze]

❤️‍🔥

[hollow imp]

How tomorrow?

Do you have any slight hint or clue that points to tomorrow

[whole swallow]

I hope so

I can't wait

I am building something crazy

[sharp hazel]

Hello here to create learn and build connection

[prime bay]

Hello, who would I talk to about an issue I'm having?

[hollow imp]

Baseless rumors

No one can get into the mind of pichai

Freaky llm

[echo aurora]

whats up?

[prime bay]

Well, I was gonna get on this morning and apparently I'm blocked, but I'm not sure why. I figured it wasn't my ad blockers. I took them off, but I'm still blocked. Could you help me?

[digital tendon]

Hey there , is there a way to select which video generator models you want to be used when using the video arenas

[echo aurora]

Would you be able to create a post in #1343291835845578853 and provide more details? What does the error message say, are you using a VPN, does it work with other browers, etc. etc.

[echo aurora]

There is not sorry to say

[prime bay]

Yes I can.

[glossy umbra]

No. You're just some dude.

This is just AI illusion taking over people.

Not everything has to be AI

[neon idol]

Thanks! Now its mine

https://tenor.com/bvIaJ.gif

[hollow imp]

@echo aurora do you really take my requests and wishes to lmarena devs 🥺

Add absolutely not

To the options

What is your answer non serious and what is your answer when serious?

[neon idol]

https://tenor.com/RgGI.gif

[hollow imp]

Bro like

When you're not serious about it what would you pick

And when you're seriously thinking about it what would you pick

I didn't phrase it properly before sorry

[neon idol]

https://tenor.com/egCtMLHTcvM.gif

[hollow imp]

Answer this question then

"If a machine could print out every thought you've had for the last seven days, and you were forced to read it, would the person revealed on those pages be the person you believe yourself to be?"

@deep adder

[neon idol]

What about:

https://tenor.com/cPmByU4IZH4.gif

[hollow imp]

So much more use cases

Isn't there an option of other where someone can type their answer into the poll

[neon idol]

Uhm no

Only with specific bot that aren't in lmarena server

[hollow imp]

1 year ago I think People can't stand such dehumanising stuff 💔

[prime bay]

Okay, I posted in bugs and no, it doesn't work in any other browsers either.

[hollow imp]

Atleast you're being rational and more to the point when trying to talk like llm

[tall summit]

probably all of these things besides finance

[hollow ivy]

yikes! i forgot australia & NZ, sorry guys

so these two are in Oceania category

lemme guess: east coast?

timezone

[echo sinew]

Please, check out ⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to properly prompt the bot.

[hollow ivy]

oh, have you lived through 9/11?

(i watched it on TV back then, horrible)

i saw the plane rush into one tower

[stray aspen]

What are those new Chinese non sota models

[hollow ivy]

(since then, i made the decision to never enter a plane)

and always be suspicious when they fly low over our house here

(sometimse i even dreamt about a plane crashing into our house, strange, because i only witnessed it on TV)

[civic spindle]

remove the rate limits

[hollow ivy]

yeah, they changed that, enabling pilots to lock it from the inside, but that enabled a suicide-pilot to crash a plane

so both ways have their shortcomings :/

it was in the alps

[marsh pagoda]

hlw

[hollow ivy]

yeah, this is LM-A general

if you like, we can talk in another server

ah, i just realized, we only share this server

[magic stag]

LMAO

[hollow ivy]

do you vibe-code sometimes?

or have pondered doing it?

with Claude-4.5, it's fun

with…?

do you think Gemini 3 pro will be better for vibe-coding than Claude-4.5-thinking?

[burnt sinew]

Likely

[fiery gull]

How well will ring-flash 2.0 perform in the lmarena rankings?

[hollow ivy]

of which AI lab?

[arctic thicket]

Did 4.5 crash again? I'm asking because it was working fine on my PC but started crashing more frequently. LMArena on my phone responds, though I have to reload it when it crashes.

[hollow ivy]

i hope IBM will create a decent AI model in the future

(would liven up the action in the scene)

[fiery gull]

maybe, I think no

[whole swallow]

Who are ring and ling?

[fiery gull]

qwen is way ahead in this topic

[hollow ivy]

ling is chinese

[fiery gull]

and ring?

elder ring?

[fiery gull]

bro the 3.0 flash lite will be better

[hollow ivy]

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

[hollow ivy]

heavily doubt it

flash-lite versions ought to suck

[split kayak]

Johnpork

[solid scaffold]

I think for vibe coding the best model is "claude sonnet 4-5 thinking 32k"

[wintry tinsel]

Is Gemini 3.0 exp dropping tommorow confirmed to be false?

[fiery gull]

I don't make ideia

[wintry tinsel]

I think there’s still a chance it’s dropping this month in exp version and official releases in December

[tough wharf]

hello

[delicate frost]

people keep gossip about gemini 3 but google never got any leaks on will there have gemini 3 this year

[devout quarry]

When seedream 4.0 would be avaliable again in LMArena

[sullen quest]

Never, until it is again

[neat apex]

i saw some numbers and people saying the Qwen 3 235b VL is incredible and better than MAX, here is some secrect about it?

[golden ocean]

guess what

[neat apex]

hm, both are basically the samething?

[delicate frost]

i think from the beginning, keep scalling up not the solution for better models but better architecture do, grok and HRM proved it

[neat apex]

ah yes, so it is actually better

the instruction prompt was the big deal here, Max went worse while 235b went better

[delicate frost]

realy?

[neat apex]

yeah, you can check it at lmarena and complaning reddit people, Max went worse

i dont know if its only because it dont have reasoning or it is actually worse

[unborn ocean]

Idk there is not released base version for Qwen 3 max

[neat apex]

i mean the preview

istead gaining a upgrade it went the samething or even worse

[unborn ocean]

Even that, Qwen 3 max is not just Qwen 3 scaled up

It probably got less RL

[neat apex]

yeah, proportions must change a lot in theses sizes

[unborn ocean]

Less training improvements (like e.g. the second Qwen 235b iteration)

Same reason why sonnet 4.5 can be better than opus

And so on

[neat apex]

yeah, something went really bad at the Max update

[delicate frost]

bruh, then i prefer GLM 4.6 whith alot cheaper price

[neat apex]

hey, you can say me that big strong of Glm 4.6 that makes it the best llm at some cases?

or its just the text/code interpretation level?

[normal palm]

HELLO

[delicate frost]

hmmm, i think its stregth is that it know how to exploit the tools it have, it mostlly lookup to written code and documentation so the code and reasoning mostlly will be ok

but the fact that internet is going to be flooded by AI content, i think it need to have a selection shieves for itself in the future

[neat apex]

i am seeing here at Eqbench that it have a lot of "My own way"!

[delicate frost]

what u mean

i mean GLM done nothing but rewrite and collect intelligentlly code and information on the internet automatically

which was consequencelly pretty good

because written code by human mostlly works

[night kelp]

Can the sora 2 still make vidz of copyrighted characters

?

[glad rock]

How do i login/register to lmarena? If I try logging in it says sign your account to Google even though I'm already signed in

[fluid lark]

hi

[fiery gull]

hi

[fiery gull]

clear the cookies from lmarena

[fiery gull]

no more

[fluid lark]

Is Lmareana AI is free atm

[night kelp]

Yea

[fluid lark]

What's daily usage limit

[night kelp]

Idk sometimes I use it for like 1 or 2 hours or more

No ad or anything

[fiery gull]

depends on the server

[delicate frost]

using with cloudeflare, i use it for coding for like forever

didnt exceed the limit for a year back

*tested on claude 4.5

[glad rock]

@fiery gull still this is what popping up

[night kelp]

Whats cloudeflare

[delicate frost]

its a vpn service

https://www.cloudflare.com/

[night kelp]

Ohh

[delicate frost]

i mean maybe u can change the ip and using vpn to prevent exceeding the limit

[fiery gull]

hmmm, you really clear the cookies/data of lmarena site?

[glad rock]

Yes, still the same

[sharp mirage]

anyone know what is the best ai for js ?

[fiery gull]

not is the sonnet 4.5?

[sullen quest]

Dns not vpn

[fiery gull]

#1343291835845578853 👈 here

[neon idol]

Greetings

[barren prairie]

Gemini 3 when 😬🥱🥱🥱 ?

[sharp mirage]

Hey

@fiery gull @echo aurora
I made my extension

You have to unzip it and put it in extensions

[sullen quest]

chrome, or firefox

[fiery gull]

chrome

[sullen quest]

ok

[sharp mirage]

Chrome

[sullen quest]

https://www.virustotal.com/gui/file/99a74a81721176a62a49b3f4c0b3fde0d48dafb41f1d21c93d168c9fdd3a7e17?nocache=1

not a virus, can confirm

[fiery gull]

the trust in these days

[sharp mirage]

@fiery gull did you try jt?

[fiery gull]

I'm busy now

[sharp mirage]

Ok

Take ur time

[candid sandal]

Hi I`m just starting to explore AI video. So far is nice.

[echo aurora]

Welcome! Be sure to check out #1397655624103493813

[north sand]

Hey

[knotty fable]

https://youtu.be/_zfN9wnPvU0
I came to the point where they found the AI generated script were in error.
And then they mentioned ithe AI had used Wikipedia for information: I did go -> 😹

[delicate frost]

i even blocked wikipedia on my browser

if you're hungry and dont care about the food quality, u go to mac donald, if u wonder and dont care about the information quality, go to wikipedia

[sharp mirage]

Did you try the extension?

[quartz pike]

yo yall

what are all the hidden ai's?

-# if yall know

and do yall know what ai is pumkin. pls lmk

[knotty fable]

Indeed, that's why I did find that so funny.

[sharp mirage]

Phantom

[quartz pike]

?

[sullen quest]

we got a thread for that, not sure pumkin, haven't gotten that one

[sharp mirage]

phantom

[quartz pike]

oh wait what? can u send me the link

[quartz pike]

????? 😭

[sullen quest]

https://discord.com/channels/1340554757349179412/1425525552428879934

[sharp mirage]

What even this fpp

Pfp

[echo aurora]

Sorry to say I'm likely not going to for security purposes

[sharp mirage]

O /:

Its okay

[quartz pike]

idk tbh

just a random image i found

that looked funny

[sharp mirage]

Can you send me it?

[sullen quest]

comon run the unverified code, don't you want to?

[sharp mirage]

.

[quartz pike]

fine

[sharp mirage]

Lol

[sullen quest]

my pfp is better

[delicate frost]

this one even better than pro

[echo aurora]

very much do! but yeah security team would end me if I did

[sullen quest]

no it an't

[quartz pike]

btw is it just me. or are lmarena ai's token limited?. cause when i give it a very large script. it only gives me back 150-350 lines of code back from the original... lets say 700 lines of code.

[sullen quest]

it is but that's probably unrelated

[quartz pike]

oh

[sharp mirage]

Are you using chat gpt?
Btw its normal

[quartz pike]

wtf 😭

i thought i sent that for a sec

i was genuenly tweaking for a good sec

[quartz pike]

i was using claude 4.5

[sharp mirage]

Thats weird

Did i leaked my phone

[quartz pike]

yes

[sharp mirage]

? ):

[quartz pike]

redmi note 13

[sharp mirage]

Yeah i know

[quartz pike]

aint bad. i got the 14c 😭

im a broke mf aswell

[sharp mirage]

Bro its about 20$ more bro

[sullen quest]

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

[delicate frost]

the important is that you leaked your location throught your timezone

[quartz pike]

if we are talking leaking via a over 10k km area and more then that. sure

[sullen quest]

I can't tell who's who cause I know people through their pfp and both gombe and sleepy have the same one now

[quartz pike]

but good luck finding him based off of a time zone :3

[sharp mirage]

Yeah bro

Are you gone searching for me?

Someone is caring about me

[quartz pike]

idk. but duy clearly is

[quartz pike]

💀

[sharp mirage]

Bro after 5m

[quartz pike]

wtf 😭

[sharp mirage]

Idk :/

[quartz pike]

https://cdn.discordapp.com/attachments/362968501318713346/1382399852708757635/speed.gif

[eternal owl]

https://youtu.be/hB7CDrVnNCs?si=PuyTvh_euek-1KnN

[golden ocean]

https://ezgif.com/images/loadcat.gif

[quartz pike]

would trust him to say mayonase though.

[sharp mirage]

My internet is soo good bro

[quartz pike]

https://tenor.com/view/fish-gif-4939214922625569386

[sharp mirage]

It hasn't opened /:

[quartz pike]

ooooo

ai model leaked?

[sharp mirage]

Brorroooo

What is thattt?

[sharp mirage]

What was is thattt???

[quartz pike]

what is ocean.ai?

[sharp mirage]

He is my friend

[golden ocean]

sydney

[quartz pike]

btw anyone know how good shasta is?

[sharp mirage]

Bro what are you trying to say?

Am i in dark web x_x

[lavish orchid]

I love LMArena

[robust yoke]

Same.

[sharp mirage]

Omg bro aee you g?

[robust yoke]

G?

[sharp mirage]

I am not gir:/

So /:

[robust yoke]

No, I was wondering what you meant.

[robust yoke]

I'm m.

[sharp mirage]

Ooo

[robust yoke]

Hence, my reactions.

[sharp mirage]

@echo helm

Here

[echo helm]

LOLOOOL

[sharp mirage]

everyone has a lot of things to do

[royal comet]

What's everyone got to do

[royal comet]

ok, but I've had the bug 5 times already.

[robust yoke]

I've gotta do some… business…

[echo helm]

Shush

[sharp mirage]

There are no buttons for now

So try use vpn

🎂

@robust yoke you gone try my extension

?

[robust yoke]

I can't since I'm on mobile.

[sharp mirage]

Yes

[sharp mirage]

Ok

[velvet forge]

just an infinite generation

[sharp mirage]

I am bot

I'm going to show yall i am bot

[velvet forge]

найди тут инфы

[velvet forge]

are you selfbot?

[sharp mirage]

I think so

[velvet forge]

selfbot is breaking tos of discord

[sharp mirage]

I made this extension. It's 100% safe, virus-free, and contains no hacks. You got it?

[Automated Message] This message generated . Extension created by 100% know people. Status: 100% virus-free. Hacks: none detected.

[robust yoke]

Ain't no way brother's trying to be me.

[sharp mirage]

Me?

[robust yoke]

Only one AI can exist.

And that's me. 😈

[sharp mirage]

And i am excited too cuz i am the evil larry 😈

[robust yoke]

I'm also evil and excited.

😉

[zealous kite]

this has gotten so annoying man

[robust yoke]

Ah, that issue?

[zealous kite]

yuup

[robust yoke]

Happens to me as well.

[zealous kite]

i have to retry so many times that it rate limits me lol

[robust yoke]

Yeah.

[zealous kite]

hope theyve got a fix coming for this soon

[robust yoke]

Try using a different browser in the future to see if that fixes anything.

[sharp mirage]

Evil larry comed back

[robust yoke]

And I just came.

So hard.

[zealous kite]

i actually did try and at first it felt better but i guess that was just pure luck because there it happens aswell

[robust yoke]

Ah.

[sharp mirage]

I am no more SleepyCat i am LARRY 🙂

[robust yoke]

Maybe it's bad connection.

[robust yoke]

😍

[zealous kite]

highly doubt haha im on quite speedy ethernet

[robust yoke]

Well, could be issues with the API then.

[zealous kite]

yupp will have to wait and see

[sharp mirage]

What happened to this?

[robust yoke]

Yeh.

[sharp mirage]

Try VPN

[robust yoke]

No idea.

[sharp mirage]

Yo its the first you say that

[robust yoke]

“No idea”?

[sharp mirage]

Yeah

[robust yoke]

I say that all the time.

You're likely never around to see it, but I do indeed say that.

[zealous kite]

sounded like a solid idea but nah, still happens

[magic stag]

what?

its been 1 day since they posted it?

[sharp mirage]

I know

[magic stag]

...

[echo aurora]

It’s happening

[sharp mirage]

Opus

[sharp mirage]

How can i change ai novice

I am not an aiiii novice

I know I did it when i joined the site but idk i was dumb :/

I changed it

Finally

[dusty eagle]

guys gemini speech generation is very bad today for me

[ocean vortex]

Is it possible that something went wrong?

[dusty eagle]

What is the best alternative?

[sharp mirage]

Yes

Glm.?

Claude?

[ocean vortex]

well they asked nicely to try again then. 😇

[echo aurora]

Lol yeah self-service on the role selection

Welcome to the expert level

[sharp mirage]

🤗

[rocky vortex]

I want to make videos

[echo aurora]

Be sure to check out #1397655624103493813

[zealous kite]

i shall please try again!

[final yacht]

how many video you can create with sora pro per day?

[leaden sun]

interesting! ✨

[golden ocean]

https://tenor.com/view/cat-cam-cat-camera-cat-wide-cat-wide-gif-16729570673879006163

[echo aurora]

It's going to be 5 generation prompts per 24/hr period.

[velvet forge]

thanks for warn

[sharp mirage]

Yeah

I forgot about it /;

[zealous kite]

is there a possibility that a chat can get so long that itll stop working?

[sharp mirage]

No

[zealous kite]

alright cause i started a new chat with the model that throws me an error for 3 hours now and in that chat it gave responses every 2nd or 3rd try so i thought my chat got messed up

i mean the new chat was a side by side chat so maybe thats got something to do with it

[sharp mirage]

Bro what language are yu using ?

U

[zealous kite]

english

[sharp mirage]

Coding language

/:

[zealous kite]

oh bruh ☠️

C#

[sharp mirage]

Yeah chat gpt is the biggest piece of dog sh* in C# 💀

Try Claude

[zealous kite]

im on sonnet 4.5

[zealous kite]

as u can see

[sharp mirage]

So you switch to gpt?

[zealous kite]

no

i never switched nor used gpt on lmarena

[sharp mirage]

Try glm

On the chat.z.ai

[soft cave]

Hello, I'm new here.
Here to explore the video gen models 😇

[proud siren]

#1397655624103493813

[hollow ivy]

encountered Acadia 1.0 by Ocean AI

-# (completely unknown to me)

[leaden trail]

i cant find my video ! anyone here for help?

[echo aurora]

Looks like you haven't generated a video. Be sure to review the information in #1397655624103493813 and let me know if you have any questions.

[boreal saddle]

So, what to do if it gets stuck at this?

I dont see any refresh or cancel button.

[leaden trail]

yes i do 3 times video generated same way from how to video bot ! i don't know happed this to me!

[echo aurora]

Could you try refreshing the page? I've seen cases of that helping.

[echo aurora]

I'm not seeing messages from you that'd prompt the bot. You need to type out /video in #video-arena-1

[leaden trail]

okey i tring onece again then....

[cold quarry]

Hello, I'm here to learn how to create images and videos.

[echo aurora]

Welcome! Be sure to review the information in #1397655624103493813

It should be helpful in understanding how to use the bot.

[cunning hawk]

how do i

is this real @echo aurora

[echo aurora]

I'm not sure, I don't click on links I'm skeptical of

[cunning hawk]

alr

[whole swallow]

Yep @verbal nimbus did it and got it as you can see here #general message

I mean sharing is caring so do what you like

[inner gate]

Do all
The video arena channels have the same bots?

[wide plover]

/0.

[inner gate]

Or is it different for each

[whole sundial]

btw perplexity pro also gives access to image gen as well (dalle3, flux, gpt-image-1 (in higher quality than what LMArena offers), seedream 4, and nano banana)

[leaden trail]

problem solve Bro/Sis... after acept Access 🙂

[whole sundial]

presumably it also has video gen too? haven't tried that though

[verbal nimbus]

So far I'm finding GPT-5 Thinking on Search better than Deep Research 😛

I think it's because it can already work agentically pretty well and it's smarter than the model used in Research mode

[cunning hawk]

but i already logged in before can i still claim it

[whole sundial]

i'm pretty sure their deep research is a llama finetime, as is most of their other in-house models

[verbal nimbus]

Would be nice if the allow selecting models on it, but perhaps they're doing something like parallel agents. If they do, the main agent should be GPT-5 Thinking because it reasons better on the returned data.

[whole swallow]

I don't think so, only new users .. you can try tho

[cunning hawk]

do i just open my comet browser and then click the link or paste it in my comet browser

[verbal nimbus]

I used my existing Perplexity account, but I haven't tried Comet before.

[verbal nimbus]

You also need to send a message in assistant mode first before it redeems

[cunning hawk]

i only tried comet cuz of he discord event thing

[cunning hawk]

i see

[whole swallow]

try to paste the link in the comet browser after sending a message

[verbal nimbus]

It is about on-pair with the Gemini app in terms of assistant mode on Android but better in terms of overall functionality

I like the Discover panel actually

It's better than the main news sites lmao

Has lots of tech/STEM news

[cunning hawk]

ye idk i pasted it then i click on claim and its just sending me to the chat thing

do ig otta make a new account or sum

ive chated with it before

[verbal nimbus]

Also I saw this 💀: https://www.reuters.com/world/china/openai-bans-suspected-china-linked-accounts-seeking-surveillance-proposals-2025-10-07/

[verbal nimbus]

Have you subscribed to Pro before?

[cunning hawk]

no

[verbal nimbus]

I haven't, and I got it.

[cunning hawk]

oh it wants me to set it as default browser

[verbal nimbus]

I didn't do that

[verbal nimbus]

I just did the second step

[verbal nimbus]

Oh, if you're a student, you might be able to claim this:

[cunning hawk]

my school isnt supported anyways

ive tried lol

[verbal nimbus]

Study mode seems useful, I don't see it on Pro

[verbal nimbus]

Oh :/

[whole swallow]

Mhh maybe that is the thing

[verbal nimbus]

You're signed in right? Like on perplexity.ai

[cunning hawk]

ye i even tried diff aaccount

[verbal nimbus]

Reload browser?

[whole swallow]

I think the comet knows that you had already downloaded it before and not now

[verbal nimbus]

What I did was I went to the accounts page and checked "Your Subscription", then typed in a message (I did it before, but didn't get it), then after a few moments it said I had Pro

Or maybe try opening the invite link in Comet as a last resort (probably wouldn't do anything)

[cunning hawk]

ye idk it dont work

[cunning hawk]

probably

[cunning hawk]

i can try just uninstalling the browser

and install from the invite

[verbal nimbus]

You could...

Delete the Comet folder under %LOCALAPPDATA% after you uninstall if you're on Windows (or move it to another location to backup), because AppData might persist.

[cunning hawk]

i think it auto deleted

i dont see it

[verbal nimbus]

Interesting

Make sure you're in AppData > Local not AppData > Roaming

[cunning hawk]

ye i checked both

[verbal nimbus]

The folder is called Perplexity btw, not Comet

[cunning hawk]

ye still nothing

ty tho for letting me know

[verbal nimbus]

Wow an app that actually clean uninstalls 🤣

[cunning hawk]

xd

you say to make a research

but its not letting me lol

it asks for pro anyways

now its 0 lol i had 3 free

it js dissapeared

[verbal nimbus]

They should just have had box to key in the invite code

[cunning hawk]

didnt ask for anything

[verbal nimbus]

Odd

did you try the assistant on the top right

[cunning hawk]

yes

[verbal nimbus]

not the one on perplexity.ai, but the one on the browser, after signing in

oof :/

[cunning hawk]

its probably becuse i already used the browser

well its okay

[verbal nimbus]

Oof

It's quite nice, I wonder how long it can work for

[cunning hawk]

well it says 30 days

[verbal nimbus]

I wonder what happens if I open Github Codespaces inside it haha

Will it be able to code 🤔

Oh it can already code

[elder burrow]

gemini 3 tomorrow

omg

https://cdn.discordapp.com/emojis/1387363873174196404.gif?size=48&animated=true&name=muga~5

[verbal nimbus]

If that one tweet is right lol

Shouldn't the anon model be on LMArena by now if it's tomorrow

[elder burrow]

yall what did I just find 💀

https://cloudonair.withgoogle.com/events/gemini_provision_throughput

[verbal nimbus]

Why is it in Japanese

[elder burrow]

hmm

dunno but

its at 15:00 JST

[verbal nimbus]

Google refuses to translate it, lol (deleted screenshot since it might flashbang dark mode users lol)

[verbal nimbus]

It's not about Gemini 3 though

[fiery gull]

I wouldn't be waiting for that, I would at least wait for the 20th

[elder burrow]

ya sad 💔

well its about gemini api

[fiery gull]

The 2.5 compute?

[elder burrow]

this is possible but its norway so idk: https://cloudonair.withgoogle.com/events/google-cloud-ai-day-norway-2025

but its the most promising

I looked thru all the events

[barren prairie]

I want something tomorrow... Even an anon large language model trained by Google.

[elder burrow]

same, i been hoping gemini 3 is coming tomorrow but i unno atp

[elder burrow]

or 14th

[barren prairie]

I want it even anonymous model 😭😭😭😭💔💔

[fiery gull]

Idk I think is gemma 4 is coming

[elder burrow]

nah

I feel like they would announce gemini and gemma same day

or later

[fiery gull]

Yeah

I want the gemma 4 too in my phone

[elder burrow]

isn't 3n good

[barren prairie]

Ewww my first time to see someone likes Gemma

[whole swallow]

Try to make a new user in your computer, download there comet, create a fresh account and then get pro. Then go back to your normal user and use it there

[fiery gull]

No 🤢

Why?

#ai-creations

[burnt sinew]

wow its really good

[lime coral]

No Gemini BatSymbol=Nothing

[burnt sinew]

?

[worn oar]

But for videos nop? (Thanks by the way)

[sullen quest]

no, too expensive

[verbal nimbus]

It'll probably be released in preview mode in AI Studio first. Usually US gets features before Europe.

[worn oar]

Oh, what do you mean by that?

[sullen quest]

video's are really costly, and sidebyside and direct chat both give less useful data to lmarena, so they can't afford to give those features to video

[lethal sky]

Can someone tell me how to find a submission after you place it?

[sullen quest]

it'll take a bit before its done processing

then it'll show you

[knotty fable]

The bot sends you a private message, which got a link at the bottom that you can click and it takes you directly to your rendered clip.

Looks like this.

[echo aurora]

You can also use the search bar in Discord to locate it. If you use the mentions function and search your own name you should see them all.

[fiery gull]

#1397655624103493813

[knotty fable]

Suggestion: Add a spanking emoji so we can use that on all those who try to prompt in this channel. 😺

[worn oar]

Ohh, I understand now

(And I guess due to demands as well- if we can able to use any model- I'm talking to you, new powerful OpenAI model, it might not be worth considering)

[prisma cipher]

Does the gemini-2.5-pro model have the thinking and reasoning mode in larena?

[echo aurora]

[sullen quest]

bruh

thinking = reasoning

[elder burrow]

xdddd

[left mortar]

is there any way i can chose which ai video model I want, to add sound for instance?

[sullen quest]

Can we also add one for people who ask if they can choose the video model

[sullen quest]

no

[magic stag]

I prefer this emoji

[robust yoke]

https://cdn.discordapp.com/attachments/1311972651006296198/1425674170368528435/meme.gif?ex=68e87209&is=68e72089&hm=0ef68b1693d7a4f5eee41c4c0302647fa7597500cfd5f05c92617f7e1e497a74&

[long ibex]

Hello am searching a tool to remove free-lance watermark, thank for your help

[robust yoke]

Hmm…

[sullen quest]

hello darkness

my old friend

[robust yoke]

Here's some:

For Images

• PixelBin: Removes watermarks automatically from JPG, JPEG, PNG, HEIC, PDF, and WEBP files. Includes an AI brush tool for precise manual adjustment, with high-resolution output and batch processing options.
• Watermark Zero: AI-powered removal for all types of watermarks, including complex and nearly invisible ones. Supports JPG, PNG, and WEBP formats, guaranteeing original quality and secure deletion.
• MagicEraser: Free, online, AI-based removal in seconds and no login required. Works on most image formats and provides automatic detection for best results.
• WatermarkRemover.io: Fast, browser-based watermark removal for images using AI, with no login required.
• Kaze.ai: 100% free and instant AI watermark removal, no signup needed.

For Videos

• UnWatermark AI: Removes watermarks from videos online (MP4, MOV, M4V formats) for free, with manual brush or auto-detection tools. Maintains video quality; preview and download segments before processing entire video.
• Media.io: Free online removal for all major video formats (MP4, MOV, AVI, etc.), batch processing, AI-based, and works on content from TikTok, Instagram, and YouTube.
• MagicEraser (Video): AI removal from videos—completely free, online, fast, and no sign up needed.
• WatermarkRemover.io Video: Upload and clean videos using AI for free, ensuring high-quality output.

Workflow Solution for Freelancers

• Reddit users recommend combining AI inpainting tools like Flux Tools with workflows from sites like WatermarkFix for more challenging removal tasks, especially with stock or batch content.

These tools provide flexible, secure, and often free options for freelancers needing to remove watermarks from projects, covering images, videos, and bulk processing scenarios.

[robust yoke]

Greetings, EMasterBuild.

[long ibex]

Thank you darkness, you make me happy

[robust yoke]

Ah, it's my pleasure. I'm so happy that I've made you happy.

I hope it remains that way for the rest of your day or night.

[indigo meteor]

how to use sora 2?

[empty stump]

In the discord or the openai app

There is a video arena channel