#general
1 messages · Page 372 of 1
bro are you sure 3.5 come today ?
Claude opus 4.6 >>>> gpt 5.5 pro xhigh
100%
It aint coming out this year
This ragebait sucks
You suck at ragebaiting btw
Opus 4.7 better mate
No, opus 4.7 seems like a downgrade honestly
what about this???
Frr. That(gpt 5.5 x-High)was another experience when I used it in the Arena. Its mapping abilities are perfect although it goes unnoticed with all other topics as well. It can very effectively collect the correct data and apply it.
No, its actually worse your wasting way more tokens on 4.7 and get a litle bit more coding experience. For me opus 4.6 is the best (Opus 4.6 and Opus 4.7 have same pricing)
Idk, if you believe if this was veo 4, then idk, it's not out yet, and besides, veo 3.1 can generate only 8 seconds, not 16, besides, this could be seedance or kling again
you think new gemini can beat opus ?
Never, google invested 20b in claude
and ?
gemini can be better than opus ?
20b is nothing to google lol
AVG techtoker
True
even mythos loses to 5.5 miserably
5.5 Lies all
🙏
1/40 or their monthly revenue btw
nah it just tells you when you are stupid
They all lose to siri
I just hope this new gemini model at least is good, i still have ultra and wanna give gemini one more chance, after 2.5 they started going downhill
2.5 was goated
At last year's Google I/O, they didn't release a new Gemini
That's last year
Didnt they release 3.1?
Wsg all
no
Yeah its was gemini 3.1 pro
no bro
Guys do u know other free platforms where i can generate at least 5 vids?
Bro its there
kling maybe
or seeddream (I dont really know)
Siri da goat
Ragebait -9/10
Siri might be the goat
ROBLOX LOL
The calcuator clears
Brainrotted
Bro here AI not roblox
gpt'ed*
I just showed his profile
Kling-ed
ik the rules
Ik
No, kling is actually smart
As a video gen
Roblox got that grown ah man community, It's an entertaining watch 67/100
dudes literally talking about roblox inhere of all the things
bro
But i also hate roblox it just like broswer games
Yooo wsg
I'm 16 gang, not grown
"Ground up man (Maybe) btw"
Hey
bro
Im good wbu
I got like 50 dollaroos on one of those
Speed stickers in "massive" ram shortage
broom broom
yes
Speed might be the ram shorage
Can i make you one if you want
Yooo wsg watermelon
i honestly dont understand how vince (banana) went from the most UNKNOWN person here to one of the most known person here in just 6 days 😭
this is what produce does to u
Who is that
makes u famous 😈
Random
good wbu
Bananafarm
I might have to go by peach around these parts then...
alot of spots are available in produce
We dont want any kne more
It kinda is
better than
opus
4.7
Yeah
5.5 xhigh mogs tuff
But i dont think better then 4.6
Siri slams
Opus 4.6 mogs almost all models in the world
no way
I gotta summon @surreal zephyr for this one
😭
Btw did you know first ai was released in 2018 (QuickDraw)
Okay dude, let's chill on the ragebait
How aero gold ?? Is he a specter
Generative ai*
See, opus is way better
the only produce that betrayed us:
Jalapeño (now "Teleste")
Yeahhh
who is that bro 😭 💔
And spud bro
Thats my custom bot
slam yo bot
Nah, its runnin on my server (Not vps but my homemade one)
My grandma mogs gpt 5.67x plus ultra into oblivion...
e
e
5.5 is better, if people just saying opus 4.7 better just because of html, you are already from 2000years
e
e
e
e
Negative
opus 4.7 is not even close to mimo v2.5 pro
The first conversational AI was ELIZA, released in 1966. SHRDLU was the first machine able to linguistically comprehend commands (1970)
already gave it my all, not going to use opus anymore
Mimo v2.5 isn't THAT good son
Do you think, Mimo (2.5 pro) is the best coding AI of arena?
either you trust me or this fake benchmark
literally
Sonion this is a public chatroom
Incorrect
Thanks for correcting my mistake!
correct. opus gets mogged even by grok bro
its THREE tung
Triple T not whatever this is
these people just waiting to see which ai can genrate better UI, but when comes to debugging, software engineering, Gpt 5.5 is beast. Alrrady patched a 5 years old software, claude failed
Don't you ever disrespect the big T, this ain't no game
okay
Honestly I rarely saw in any ai community that done any good testing, all they do just blah blah blah
<@&1349916362595635286> #rules
It is okay to have discussion related to new policy or laws as long as it’s related to AI.
BRO
@surreal zephyr
can u tell me
why
aero always
pings mods
for the randomest things ever
like
rules
there are prob hundreds of triple t messages
don’t question it bro
still
aero has a different aura about them
bro could probably ping trump himself and he would respond
"produce" should have been against the rules too then?
why is it tolerated by pineapple
lol
because its not related to AI
ooh cool avatar icon, the most creative one I saw
why is it not banned yet 💀
lol I had the same profile pic for the last 5 years
Another triple t
this one is FROM HAPPY HORSE
triple t is against the laws
🔥
@main nexus u cookee
u gonna be arrested
Best if you have unlimited budget?
where is mistral
best of all models which can be accessed in battle mode of arena.ai
I'M BACK
only 10 options are possible in polls, i had to choose the best
Ayy
- not ai
Not ai
Word "Not" isnt ai
AI related
Not ai
should i have crammed it in option #10?
Not, ai, related
Pineapple is typin]
nah it should’ve been at the top 😭
Roblox IS ai related
This rule is bit of a grey area as there are going to be discussion around policy and laws that will effect this field, and having discussions about that is perfectly okay. Overall though, we'll step in when things start to turn into a political conversation, although sometimes seeing where that line is sometimes isn't too clear.
Hey claude this is an amazing project we created!!
Thinking
-# The user likes the project we made, I should come back with a good friendly response,
-# But oh wait I found a small bug here.
Deletes all files
-# Oopsies, I'll just replace it with a starter template
Claude: Yeah user! Its an amazing project we just made, I also fixed some bugs for you.
although i value Mistral, it is not frontier, for me
Gang, no it's not?
LMAO 😭
Roblox uses AI to sell users face/id data
Never happend to me
@surreal zephyr
Tripple T is a ai generated charcter chud
"I am too poor to afford claude code"
triple t is real
It wasnt? lemme search
Pineapple came early today
😭
Ragebait pro max
dont listen to him he got da new ragebait
why u useless pinging mods dude
Avoid political and religious content. #rules
💀
Second time I was wrong.
@sleek urchin avoid unnecessary minimodding
Yeah, it kinda worked 5/10 he could do better
ISTG
IFRNBGW
Did you really use ai to answer that??
No like
Dude
What
Cus search is not ai related
hes fr one of those kids that think they'll get something special for reporting useless stuff
causing drama and unnecessary spam pings
Yes it is?
fr
false pinging mods, prolly trolling
he wants a full AI chat
Pineapple typing
fun fact: gpt 5.6, gemini 3.5, and mythos 5 dropping really soon
🤑
We can move on from this conversation, people here are going to ping mods when they think it's appropriate, there isn't really any value to be had having some disagreement if they were right or not to do so. We're here to get pinged, it's okay.
ok but tell him to stop replying with rules I mean I'm sure we all read the rules here
hi
Yeah..
Speed sticker isnt related to ai, you didnt reply to anyones message so your message is useless.
yo
HI Aerop
Dude can you shut up
Are you his bf
you are free to create your own poll :)
Was a reference btw
Just noticed
ok if u want a full AI focused chat I'd recommend you going to OpenAI, Anthropic or Dev Mode discord server
dev mode discord? what's that?
do you have its "code"?
Thanks
Pineapple is typin
This isn't very helpful, I recommend blocking/muting other users here instead of continuing to engage with them.
Let's refocuss the conversation back to AI please.
Understood.
@echo aurora Pineapple i just used agent mode can you explain me what task it petforms ?
I heard there was a dinosaur
A brutal late jurassic lord
That lived upon this land well before us
It was 30 feet long and fifteen high
With razor teeth and hunter's eyes
The once extant king triumphant
Allosaurus
Allosaurus, allosaurus, allosaurus, allosaurus, allosaurus, allosaurus
New gemini model Soon ?
7
8
1
Yes
Really good question! I'd check out this help center as it give a good overview of how Agent Mode is different: https://help.arena.ai/articles/1811908126-arena-experiments-agent-mode
2 for video and I forgot the rest
We don't have that info displayed at the moment. The new daily usage system is still being worked out. When it's implimented, it'll include clear visibility into that.
damn! Veo4 spotted!
Hi guys, I have a problem: I use a lot of different AI agents for programming, and I use 3/4 different MCP servers, so when I change a configuration on an MCP server or add one, I have to do it for all of them (I use 2 PCs, so I have to reconfigure everything twice on different OSes and sometimes it doesn't work). Has anyone found a solution?
what I recommend is downloading a virus so you wouldn’t need to use your computer anymore which would fix the problem
Nobody noticed my hallelujah parody :(
fr?
The Video Arena is currently accessible through: https://arena.ai/video. More information on how to use Video Arena can be found in this article
sota yes but nowhere close 5.5
wow a prompt
<@&1349916362595635286>
u grabbing my ip?
yeh, link based ip tracing
tried it last month, and it works
i think u should say 'c' instead of 'p'
oh
probably being integrated soon to arena
really
works with forward too
<@&1349916362595635286> another
dude whats this new attack
thanks mod
whoever did it
spicy fried chicken, fade or idk
so many mods
this is really great
Its hard to encounter skadi and jacksparrow
yesterday or the day before yesterday a guy came speaking abuses in russian
again
got his screenshots
it has those screenshots i grabbed from the russian guy
idk if they let prohub in
even with normal user permissions it could be set to just ping mods when detected, 0 risk
u gotta ask pineapple
pineapple says he's unsure
bru
tell him
tell him to test it
first
on bodyguard demo
if hes unsure
who the hell is this typing
prohub will be able to block this ™
bro
someone said someone said
he is back
yes
prohub save us 🙏
we may intercept this new threat 🙏
yeh
<@&1349916362595635286> theres this another bot ig
he keeps typing and its so annoying
because of the large name
wait
whats his username tho
plot twist
call_49844
😭
yo dont say that u will get instant ban
prohub detected a blocked word 😡
get banned
wait no it was
🚨 COMMUNITYGUARD FLAGGED A REPORTED MESSAGE
User @surreal zephyr
Sry whats the problem?
opus
There's this bot
sending youtube ip logger links
and deleting them when mod is tagged
banana caught his username
call_49844
his username
he keeps changing name to weird names too
Okay will keep an eye out
HE'S BACK
its him
@echo aurora
Thanks
YESSSS
got screenshot
FINALLY
people nowadays just come with random names and spam stuff
such as diddypilot like whats that name bro 😭 🙏
prohub wouldve prevent this ™
Microsoft diddypilot
Yeah those are great
thank you gpt image 2 for these epic images
Ima make some for u too
gpt v2 best
honestly
theres no need of new image models now
img 2 cooked
where apple lol
too hard
img2 needs no watermark
Ayyyy
yes
Me every day
Tuff
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
Oh wow
Good for Anthropic
idk how you come from make small but popular small and open model experiments to joining the only major US lab to not release an open model ever in its history
anthropic really, really hates the idea of open weights models to the point that it went after Chinese companies distilling from their models and giving them away despite anthropic itself using copyrighted data they did not have the rights to to train their models
anthropic sucks thats why
they have ZERO idea how to make efficient models
Gato-13 reacted 🔥 🔥
Yo
yeah if elon was right with the opus parameter size (5T), which he might be due to some xai employees being formerly from anthropic, then that would make either around the same size or bigger than gemini 3.1 pro, which has far more world knowledge than opus, gpt, or any other model closed or open
and yet gemini 3.1 pro is half the price
Does anyone have Gemini omni videos
no, google gave them 20b and they hired collosus 1 from xai
grok 4.20, a benchmaxxed 500b model
grok 4.3 is the same size but different base i think
next grok will be 1.5t parameters
grok 3 was 3t
we getting some new gemini 🤔 ?
Yes
yes, 3.5 flash later today
nice
i believe nano banana 2 pro will be called
tbh i saw rumours of 3.5 pro as well
gemini 3.1 pro is still much smarter but i can tell just by using it that grok 3 was massive
i mean we had nano banana and nano banana pro
Does anyone have access to Gemini omni
now it has to be nano banana 2 and nano banana 2 pro
yeah it definitely knew a lot of stuff but it could've been much better while keeping the same size if they spent more time training it
now i wonder what will happen with their future 6t and 10t models
champion only chat ahh
New gemini modele can be the sota !
7
12
1
Yes
they are still 20x better than claude limits lmao
and claude models suck compared to 5.5
unless they drop mythos at price lower than 5.5 they are cooked hard
skill issue
5.5 medium is higher intelligence than opus 4.7 max and still uses 20x less usage
btw same guy who deleted his 100s of messages in arena
completely unemployed
😭
I just checked from:
bro has only 30 messages in the whole server now what
and most recent are today's (which is 10 mins ago)
all previous are like months ag9
ago
probably because I disproved him on him saying 5.5 is the best model out there
and hes always wrong
idk how people find him trustworthy
most of the stuff he says are wrong
😭 💔
I mean gpt 5.5 is better in most tasks than opus
There's no way Craig said that lol
lol
Opus couldn't do my homework but gpt 5.5 managed to
all of his messages are gone
oh yes he did
yeah thats average craig's opinion
check in search
that is stupid lol, i've used nbpro/nb2 and gpt image 2 and i can confidently say that they have nothing to do with each other
"Original message was deleted"
So its ragebait
he deleted every of his old messages so no one can disprove him
Its just ragebait
fr
he always switches to the newest frontier model that releases lol
to glaze it
@surreal zephyr dw he likes gpt 5.5 the best
Gemini
(also one of his deleted messages)
No way 😭
What model
how did u even manage to have the time
to delete hundreds of messages
because u definitely did have hundreds here
😭
unemployment?
there are programs you can use to remove them
hm
I doubt he would even use
them
What the hell
Where all your messages craig
Chatgpt plus is probably enough
You've been here like 2 years
deleted because of some "weird people"
messaging hom
him
ig
idk how he noticed the "weird people" now
bro thought to take action now instead of doing it before 😔
well
Tuff 67 phonk
Tuffimelon
67 tufflord from ohio
I guess he's not truly american
Even when his bio says otherwise
if you change your opinion everytime to just cause drama then thats ragebait
lol
Not tuff
code arena: frontend*
yeah +/- 5% so the first one says nothing
second one is purely about ui not about how good code is
so skill issue
Where can I use gemini 3.5
You got a subscription
pro
Uptime rollercoaster
i don't even think people without a sub can generate a video
they cant
💔
bad thing about gemini video models
fully paywalled
(in gemini)
i believe this is gemini omni (internally known as veo 4)
someone saw gemini's new web UI
i have that too
nah
The spud is.not released yet
at least the spud base model is much better than whatever they used for gpt image 2
gpt image v2 and gpt 5.5 & gpt 5.6 are distills of spud
no its 6o image gen on quantization
no, if it was it would be much more intelligent
spud is literally confirmed to be 6o omni model
with voice and video in and voice and image out
sonnt-4.6-search said this
crazy hallucinations
hi

what is the ellsworth model a new secret İmage model?
Hello - there is currently qwen3.7-plus-preview & qwen3.7-max-preview in Direct and Side by Side
You'll have to scroll for them
I couldn't say 
gpt 5.5 pro, prob
any big company might drop the nuke, google, claude, openAI
5.5 pro xhigh
I think there are 3 minutes left until the Google I/O kickoff. I think it’ll be something like Nano Banana 2.5 or 3.0 I think this time it’ll be based on the Gemini 3.5 Pro architecture ellsworth is creates a very good images
You think this week?
gpt 5.6, gemini 3.5, and mythos 5 are coming
whos gonna win
5.6
ofc gpt 5.6
mythos nor gem 3.5 arent even beating 5.5
Im NOT watching Google I/O
what is Google I/O like what was that thing? 
I think Gemini 3.5 will be released first, followed by Frontier, and then GPT 5.6. I think 5.6 will be better.
Mythos, prob. not, i read it on some legit place, they said that mythos wont be out for public use, companies and brands will use it for their security purposes
no they said they will release mythos once they manage to make it less expensive and less unreliable
I think Gemini Omni can generate both images and videos thats what I gathered My English isn’t very good, and it seems Gemini will also be available on Android 17
tuff
Pineapple please anwser
google io is live fyi
This new mode is similar to Opus 4.7, and it looks like the API price will be low.
so far no new image model...or...idk
antigravity-cli Antigravity-cli and 2.0 look good.
everybody watching that google I/O huh??
yep
so far no image model announced
if not on google then c u @october with new gpt image 2.5 lmao
3.5 pro coming too?
Kimi such an underrated ai tbh imo
composer is kimi
comp2 is kimi 2.5, comp 2.5 is kimi 2.6
idk about that one
not very good 3.5 flash qwen 3.6 dense 27 b close to and and fine tuned qwen 3.6 is Gemini is close to 3.5, and honestly, I was expecting a result around 58 on SWE-Bench-Pro. Of course, real-world experience will be different.
when the gemini 3.5 out?
bruh what
ai bubble gona pop like dot com bubble trust
(dont mind the fact it skyrocket right after uhhh)
bubble so big here it totally didnt raise back up
And on Google AIO, they've only talked about the Gemini 3.5 Pro so far, mostly flash.?
3.5 pro coming next motnh
crazy
Gemini is so good, but they just added a rate limit, and it’s not very generous
next month huh
The jump is crazy. It went from no limit with a very strict rate limit
i dont see the option of api for normal models in arena, have they not added the feature yet?
text*
how
I tested Gemini Omni with a video to see if it’s better than Seedance 2, and it’s still not even close
thats super bad
Veo is the only model that can’t get cartoon voices right
where is that from
antigravity
So it's not very good, I didn't expect this from Google's new video production model, and SeeDance 2.0 is still better.
At least Veo is unrestricted
water is not wet...unless he means the scenery xd
europe
In my opinion, Seedance 2.0 is currently too limited and doesn't produce very good results. Happy Horse 1.0 is currently average; it's neither too limited nor too good at producing results.
hell yeah
if new image model, lmk
its so fast
phew so fast, even though the live on Discord Google hasn't finished yet lel
it seems it had update for Antigravity
3.5
so fast @echo aurora 

PINEAPPLE ANWSER THESE QUESTIONS!!!:
How did you get qwen 3.7 api and gemini 3.5 flash???
Pleaseee
Because they came out
What advantage does it even have over 3.1 pro
Idk
They didnt
WHO KEEPS DOING THAT
Anyone remember when everyone was excited for the Apple keynote? Guess that has changed quite a bit. 😉
gemini 3.5 pro when
next month
somewhat cheaper
was gemini 3 flash actually gemini 3.5?
What about qualitywise
next month
it's smarter
dang wait gemini 3.5 came out already loll?
depend on what you are using it for, but should be mostly equal for now
3.5 flash is currently best Google model
is 3.5flash same intelligence 3.1pro
Does it have bigger world knowledge
doesnt pro release first where is that?
why don't you try it yourself, it really vary base on what you use it for
It uses its knowledge better and produces better code
release next month
next month
I'm really longing for a creativity/ideation model
Something that can output interesting ideas in lists
my bad i meant to call it Geminerf
Nice yall renamed it to frontend arena :)
what website did you make that?
Arena
wait arena can make like 3d stuff?
I'm currently comparing the Gemini 3.5 flash with the Kimi k2.6 thinking and Mimo 2.5 pro. I'll post the results here.
Yeah
Its not 3d model models tho
Just text models generating 3d stuff
Did someone manage to run antigravity 2.0 on mac
Great
r.i.p antigravity btw, soft depreciated to heck
just tell it 'make a three.js minecraft house in HTML'
3.5 flash is pretty impressive
SAME
Makes me even more excited for pro
Hard decision
3.5 flash will probably be lobotomized and drop to sonnet levels
whybro
yeah they slowly do gradual declines, hard refresh on their next release, then repeat
its not like 1 big nerf in 1 week, its many small nerfs over time, reducing its power
it probbaly makes the newer model look like a massive upgrade, since they ruin the old model
3.5 flash is probably a non nerfed gemini 3.1 pro lmao
Quantitative
12x faster non nerfed 3.1 pro? Son 😭
is it worth switching to claude from gemini now?
Brotato chips
Bro what's this
such a cool antigravity update
I love spending more time fixing errors than actually developing in antigravity
they hit their rate limits again on antigravity which is so bad
it was good for a while, damn google
Google trying not to ruin their products
this is ok. I was hoping 3.5 flash to be better than medium 5.5
when they compare 3.5 flash with GPT 5.5 .. is it medium 5.5 or xhigh 5.5 ?
Just shut it down already
but they just launched Antigravity 2.0?
hows 3.5 flash
arena is probbaly the last place that has decent free ai access now
medium 5.5 > opus high/deepseek max so ...
wait when?
now
opus sucks
Delete it
depends if it actually updated the models to include opus 4.7
mogs
why is claude taking so much business away from OpenAI if 5.5 medium > Opus ?
For price, it is a great model. Google is going to integrate it with all the surfaces like search, workspace etc. Very bullish on google
lets go 
Anyone watching the !googleio2026
Soon will have to use our credit card just to say hello to an AI 🤣
poor claude opus 4.7 🙁
its not? a lot of people moved away from claude
2 weeks old news : Claude revenue surpassed OpenAI. And there revenue growth is much higher.
...yeah cuz claude is exit scamming rn by raising prices?
while openai is luring more users
if not xai and google, anthropic wouldve bankrupted already
3.5 flash doesn't accept file upload 🫠
it's personal preference at this point, I really like the way opus 4.6 specifically answers
just tried. working for me on gemini.google.com
at this point, I am convinced that all these 3 companies are going to be sucessful in AI : google, claude, openai. Just like cloud business where aws, azure and gcp, all are successful
no way did they actually break the login entirely???
"Gemini 3.5 Pro is coming next month".
who cares who wins lol
as long as I get to use SOTA models I'm happy
Hallucinations are crazy
That happens with almost every model
lol antigravity 2 doesnt even work
Sonion
Who even uses antigravity lmao
i did put 2 accounts on antigravity so i wouldnt suprised if im banned 🤣
https://gemini.google.com/share/4e5beaa06f20 gemini omni kinda good
they dropped 2nd one apparently
matters a lot for two reasons (atleast for me) :
- if you are looking for opportunities to invest
- you want to find an ecosystem that you want to stick with in long term
but one generation takes 25% of your TOTAL limits
still small artificats that makes it look a bit unnatural
you invest in companies with an implied volatility that can only be expressed in scientific notation? 😭
aka watermark
Make a tuff 67 tung sahur simulator
To be honest, hallucinating identity is common but inventing a model that didn't ever exist. It's more of an edge case.
Fair
Who is hyped up for gemini 3.5 pro?
it will be solid for 3 months
Hope the day comes where they release an AI model so good at and specifically made for creative ideation that literally with one prompt it comes up with something new so well thought and interesting that it becomes sensational to the extent of being added to knowyourmeme or something or becomes global trend. Usually stuff like this takes people months or years to stumble upon or come up with while that hypothetical AI would make it in seconds. Or it could even come up with new inventions on the spot that people never thought about and immediately learn from AI how to do it and end up doing it on the spot, whereas without such AI it would take people decades for example.
Obviously traditional AIs of today can't ever do this because they train on existing data and use probability, so AI learns from us not the other way around, that's why they would have to make a great leap forward into nextgen which will be AGI that will use completely different technique instead of training on data. Kinda like how no matter how faster you try to make an HDD, an SSD (new technology) will always be faster no matter what.
gemini omni is not omnimodel btw its just veo4
it has awful prompt adherence
it gets mogged even by happyhorse
WTFFF
GEMINI LIED IN A MADE UP EMAIL I MADE
Did Tinker used to have Aura
5
7
3
Nooo
In the email i mentioned "I think im talking to a robot"
Gemini's email: "You are talking to a human being, we review all reports manually and take them seriously"
whats the free quota for gemini 3.5 in ai studio
a free credit limit for u to test gemini 3.5 flash
yeah but whawts the limit
@sleek urchin that’s wild isn’t it
Yeah.. I am sorry 😕
All good mane
Gemini 3.5 Flash - Claude Sonnet 4.6 (you can only use one, for the rest of time).
8
10
1
Gemini 3.5 Flash
guys in antigravity 2.0 gemini 3.5 flash created a 3k+ lines of code in 15 second ??
what is this speed
i hope every companies will bring such speed to their model
it doesn't even took a minute
first time im slower than the ai lol
It's such a pity that I can't log into Antigravity from my country
Oh seriously ?
yeah
Maybe there's free vpn that you can use
but honestly the model speed is just TOO MUCH satisfying
you basically ask it anything and you get the full output in a minute maximum its just crazy
Seriously, 3k lines of code in a measly 15 seconds?
it took like 1minute and 10 second
cause i took alot of time to click on the approve thing
gpt 5.5 low
or 5.5 instant
tpu i8
5.5 instant is not good at complex tasks I think
yeah but its unbelievable i never thought of such speed
much smarter than gemini still
5.3 spark has 2500 tps
Yeah but i want it on all model you know i just hope every companies will have this
for some reason i never got access to 5.3 spark
What are the limits on Antigravity? Should I chat there?
idk why
What do you mean about the limits exactly ?
for example, a limit on messages (requests)
Oh i don't know for free tier, i have the gemini subscription, it work with it, but i got it for free for a whole year just like many people
im tested Mimo V2.5 pro, Deepseek v4 flash thinking, Gemini 3.5 Flash here promt "" and this is windows 11 similator
how to send video
?
its only in the 200$ sub
Yeah that's why then
but like why the normal gpt 5.5 for example is not as fast
i mean it legit are 20x slower than this gemini 3.5 flash for example
If I'm not mistaken, you can't do that here
its not that slow if the api isnt being overcooked
ok ı uploaded resault to media fire .mp4 https://www.mediafire.com/file/dcfei3l5hi1gt7h/test-resault.mp4/file
yeah but nothing that you can compare to spark or to gemini 3.5 flash
in term of speed its another world
Guy's quick review for gemini 3.5? I am a little bus so won't be able to see much
the output speed is unbelievable, it legit do anything even the hardest task in minutes maximum
but for the intelligence its not the best
yes because spark is 2k-2.5k tps while 3.5 is 850 tps
and 5.5 is 120 ish tps tops
How would you define hardest tasks?
but thats still miles behind taalas's 22k tps
honestly i don't think its doing any difference for the user when its as fast as gemini 3.5 then its more than enough but i used alot 5.5 normal it just take way too long
most of the time its frustating how slow it is
Also what other things were announced by Google in terms of ai?
for me
have u seen taalas' 22k tps?
No i don't know what that is
like 3k lines of code in 1 minute max
how it go that fast lol
hardwired silicon, taalas hc1 chip
try it out its insane
but its only 16gb of memory and its running acient model rn
i just want all model to have such speed now i just wait for this, cause even if its not really smarter, its just 1000x more satisfying for us
Damn
btw its 1000x more efficient than h200
the only downside is you cannot change the model after the gpu is made
honestly that's just overkill what would be cool is all model with same speed that have 3.5 flash and im happy forever
Saar where is kpopus 4.7
iirc next one they will make gpt oss 120b or qwen 3.0
bc next chip hc2 is slower but upgradeable
Who use gpt oss
noone
Advertising
Soon we will work at the speed of our imagination literally
dude never heard about taalas 😭
the only thing that will slow you down is if it have to download something for you it'll depend on your hardware speed
and connection
human imagination is below 50tps
nowhere near 22k tps of hc1
I heard of but it’s not related to arena so it’s considered advertising
OMG IM ADVERTISING
this speed up the work by so much its unbelievable
yea
I want old arena back
models are getting much faster
like you think of something, you do the prompt and yeah that's it its already done
its literally you that are slowing it done
that's just crazy
you take more time to think of it than the time it'll take to do it
for the ai
this is 5.3 spark
spark is literally realtime, especially at computer use (but it has no vision so it has to use accesibility api)
Like i said i can't wait for it to be on all model, maybe not as fast as spark but if its almost as fast as flash 3.5 it'll change everything
i cant express how bad antigravity still is, 50% of responses 🤦♂️
even 2x slower than 3.5 flash i take it honestly
I want a model that can watch video realtime
Thats it
Watch video + work while watching
i mean what i could do in hours with other model will literally take 2 minute with those one