#general
1 messages Β· Page 171 of 1
how do you know it will be free? π
general release will be in dec 100%. they go every 6 month a .5 update.
everyone expecting it will be free cause 2.5 was
Because Google isn't openai
2.5 is free?
API will ofcourse be paid but if its on ai studio then people happy
but what if 2.5 was only free, to make and train the killer 3 model they intend to charge a fortune for?
sort of
And it will definitely be on aistudio
free in exchange for your data on aistudio
google will block free usage on ai studio
@magic stag bro vanished as soon as we told him to test gpt 5.1
Need us forget that 2.0 and 2.5 were just as big during the time as 3.0 is now
Gemini 3 is available via an API and was also available via Al Studio, which is a developer tool (in fact to use it you need to attest you're a developer). We've now seen reports of non-developers trying to use Gemini 3 in Al Studio and ask it factual questions.
We never intended this to be a consumer tool or model, or to be used this way. To
Yes they will make it free
In terms of they use your data for payment
Free as in free use
they should do this π―
yeah, what u wanna see?
source
Gemini 3 was not available on aistudio don't know what that is about
They had a ab test
one of the checkpoints were
yeah A/B
For a possible Gemini 3.0
is thjs confirmed or speculation?
But that's old
hopefully its paid only.
Think about it like this
pretty sure it was only available as an a/b test
There competition is based on the fact that they allow us to have access to their models
ya
And heavily like openai
can you do svg of pelican riding bike
only because they've always been close or worse than the latest openai model
but this time is different.. if they didn't nerf it anyway π
they weren't free originally. they only became free when they realized they couldn't compete with openai, not having a far superior model. they thought they were at first, but people mocked it
that feels nerfed
That's true
I remember that
But keep in mind
Once they gained a bit on the lead
They stayed free
They could of easily made aistudio more expensive or heavier limits
this aint svg but voxels, yet its still better
But they didn't
because it's only a bit. and maybe the only reason they stayed free is because they wanted to create a killer gemini 3 π
we don't know that yet. they might give us flash for free, and charge for pro π
we will see but like
so far only API was really paid
i hope you're right, honestly
Me as well
Man I remember the feeling of using aistudio with 2.5 pro for the first time
i think a lot of epople will be disappoint and a bit mad if you're not, but it also wouldn't be that surprising of google if they did charge
Felt like a superpower
agreed
wen GPT 5.1?
It's already released
released and the svgs arent what we wanted
i remember cpl years ago when chatgpt first released lol now just feels like old nokia phone
I highly doubt that they will suddenly start charging money. Their models have always been accessible to everyone free of charge and there is a 99% chance that this will remain the case.
It's crazy that I will be known as someone that lived when the first ai released
this is gpt 5.1 (Polaris Alpha). sort of comparable:
not saying better, just comparable
how is the test better than release π
openai logic π
first llm, first flying car, humanoid robots
we were promised automated cars 5 years ago π
we will see all that stuff
heres some news
So its going to be Gemini 3 pro Image
Aka Nano Banana 2
someone get riftrunner to remake pokemon
remake pokemon?
the gameboy version?
give me a good prompt so i can do my best
Make a game where you fight little creatures to capture them and make them fight for you, something like the game pokemon.
try this one
wait why dont you just use the game ip
is it copyrighted
i didnt get riftrunner in like ten prompts
Of course iOS is preferred again... Oh come on Google.
might reject if you use a copyrighted ip, or just make you a clone
wow i got it in like every 3 prompts
ios paid money for google
i only got bridgemind
bridge-mind is like gpt5.1 level
riftrunner is way better than gpt5.1
yeah we wont be getting gpt 5.1 on lmarena too soon
google actually being more responsible than openai is crazy
Why should they?
it is kind of ironic ios is preferred, given android is their OS π
yeah sure
So gpt 5.1 has no API?
where is liftrunner on LMArena
Battle arena
where is it
where is the channel on discord
@deep adder iOS was the market leader in the United States for a very long time, but since Q2 2025, they have been on par with iOS. In 2026, Android will also overtake iOS in the United States. Android is already the market leader in every other country around the world.
It's not on Discord, you'll want to go to -> https://lmarena.ai/
where (on riftrunner)?
i dont see it
maybe a secret?
go to battle mode and spam prompts till you get a generation
battle mode
iOS been winning since June 2022 apparently:
For models that are using a codename, you'll have to use the Battle mode and there is a chance you'll be sampled that particular model
Did anyone get gpt 5.1 access on chatgpt yet
nope but we got some pelican svgs in chat from someone who did
Is Polaris alpha still there
probably s
Is Polaris alpha on lmarena battle arena
riftrunner is
and its 100x better
Yeah riftrunner just takes forever to get to
gpt 5.1 wont be long on lmaerna
Is riftrunner Gemini 3
there is no API on release
Y'all thinking
we did too many tests to not conclude it not being gemini 3
if it sonnet tho
nah
gpt5.1 will never make doom as riftrunner did lol
struggling with pelican svgs what makes you think this model can do it
havent gotten riftrunner for a while now
That's new..
Whats raptor
my chat has reached limit, since i did 7-10 prompts on it
Just got it in arena
trash
amazon AI
wdym limit
if u chat with them without voting, there's also a limit
What prompt should I try running with arena for riftrunner
but like for one convo or
this tweet is truth and only truth
map of the world by random ais
ah yes SQUARE
only gemini 2.5 pro could really make a proper map
perfect flat earth now has square earth
i wonder how riftrunner does on this now
good work gang
by willow
blob earth
thats prehistoric earth
by grok 4 fast
its a good close try
earth map made by 1600s travellers
would want to see someone try
well i gotta try to get riftrunner again π
Im using codearena and trying but my rng sucks
i aint using codearena
that doesnt give me riftrunner
gave me a few times and i made the doom in it
thats why ur doom was better probably
cuz mine was in the text battle
I feel like codearena has a different sys prompt
payload is obfuscated obv
if i dont get it rn ill try tommorow if its still on the arena
lot of thinking from this model i got rn
uuh
who dis
Is there a way to stop this? Its annoying
no it has to happen
which model made this
Damn
whats your prompt
make an svg map of the world
did it generate after you refreshed
ah yes Antarctica is a trapezoid
everything here is upside down lell
ok
it was winter-wind
thinking opus
Llama is so going to revolutionize AI
nvm raindrop is llama
what bout winter wind tho
apparently its a google model
no way in hell
willow has ratelimits btw
cuts off randomly when i prompt in battle
thats prob an error
like 4.5 sonnet
its the only model which is doing this
whats that
glm 4.6
A benchmark i like for ai is greentext writing
riftrunner makes some nice ones
even getting certain info right
How good is GPT 5.1 thinking?
Didnβt it just release?
1 hour ago
raindrop kinda bad in what i experienced
I bet riftrunner has like 1500 on the arena
It's sooooo good
i still dont get it for svg map
I love the new lmarena code thingy
but it lacks the cool models
Like codex
opus thinking
etc
I wish i could choose from the direct chat within code arena
its gpt 5 with a different sys prompt
i guess
opus thinking is in battle mode
I think it should be in the direct chat as well for code arena
also codex
given that gpt 5 high is also not there
Is this ai
I can't believe anything anymore
nano banana 2
made by gpt-oss
did u get riftrunner yet
no
Is this a good image for Google ai pro that I'm paying for @gaunt spade
yea ig
but nano banana 2 is gonna make it better
Make it better fr
u get 2k quality
no
Cool
not sure on gemini
So about Gemini 3 is it gonna be free or wat
same thing
Does google ai pro even unlock a new model
I feel like I'm scammed
I dunno if the image quality is the same
shitted in pants at first
Can Gemini make this fr
eh it was just image
I'm gonna try doing this
For me that just means infinite free gens
Probably in dataset...
Any way to figure out what video generation tool makes the clips?
Ask it to do something different maybe that it needs to be intelligent for
Oh I see
Like make each country's size depend on their population
It can do so in app tho..
finally
Gemini app and the 2.5 flash model
Did it cook or did it burn
uhh weell
Yeah the flash model sucks
ais still sucks in world maps
Gemini 2.5 pro did this so much better sadlu
The paid model is better tbh
Wait is Gemini pro paid?
I just realised
even minimax make graphics better
its not google lol
btw bridge mind is gemini 3
but with like
every limit imaginable
like it can only go up to around 3800 tokens
and it errors every time so if u load the chat after it cut off then u wont see response bcuz lmarena
also it doesnt have image input support
riftrunner has image input support
absolutely new benchmark bicycle riding on pelican svg
willow or gpt5.1 or smth failed
holy #### that is the worst result ive ever seen
@gaunt spade look at ts π
damn, i never thought i would see the most trash SVG of the pelican until now
fr
gpt 2 would do better π
sonnet 4.5 is pretty silly
ur prompt is a bicycle over the pelican?
ye
btw heres my tool for u: https://svg.on.websim.com
Step 1. Make agi
Step 2. Lobotomy
even the lobotmised version acts better than gpt
they look like ___
Who made this
QWEN3-MAX
how
||for anyone with OBS (oatmeal brain syndrome) this is fake||
wdym? isnt it on https://chat.qwen.ai already
Qwen Chat offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.
i mean that qwen3 made this svg kekishe
ohh
dalle 3 is still avaible eventhough its been 2 years
somehow, there are still people who consider dall-e-3 their favorite image gen model
I've heard people say dall-e 2 was their favorite as recent as 2-3 months ago, that model is still available on the api btw
Glm 4.6
Did a pretty good job
That's not bad
Good details on the bird
Glm doing a better than gpt 5.1 and sonnet 4.5?
Job*
There better cropped
yeah 5.1 is rolling out, I got access to it
it's better than I was expecting I think
WebSim!
if 5.1 kicks the arena in the nuts i have a feeling that it MIGHT prove it a fad
preference afv is about sucking the user not actual content
hello, how can i get riftrunner?
lmarena battle mode, vote to reveal model, chance to get riftrunner
Has it not been released in Direct Mode yet? And is it Gemini 3 as people expect?
- no
- nobody knows for sure
codenamed models are never available on direct mode
Okay Thanks β€οΈ
π
yall please try this
Gpt 5.1 before gemini 3.0 pro is wild
β
it just doesn't do it well
It feels way more like that specifically on the arena but wtv
check it out
got riftrunner to reason ontop of its existing reasoning!!!
56kb.. nice
I haven't tested the html file it gave yet
any free online generator with loras support? there's 2 for sdxl but output is very bad even with proper workflow and only one lora support (under 400mb)
guys
its good
maybe
like, it's not mind-blowing, and my opinion on models tends to go down after the new model smell wears off
but right now I'm finding it better than expected
probably
right now I feel like 5.1 is comparable to gemini 2.5 pro in many ways (and it's probably a lot better at coding), so gemini 3 should be better, but it's also possible that gemini 3 just feels like an iterative improvement
I mean pretending that 2.5 pro is even comparable in coding to gpt 5 or claude is ridiculous
I mean, on lmarena it's still the best at creative writing
5.1!
This came at a great time for me
nah
oh
idk, but you won't be 100% sure too
nobody really knows for sure
or thye don't want to release it for whatever reason
it not impossible tho
adn there's might be more reason to why they aren't realeasing it to the public
I don't know?
we don't need to know everything
isn't the checkpoints good tho?
it looks pretty good
yea but it good
also gpt 5.1 bechmark not out yet?
There's stuff under the codename polaris alpha already
tbf, there's no way google has nothing good in the ai space, they aren't meta
If I had to guess, they don't really know how to make a better all purpose model than 2.5 pro, all the new checkpoints they have are all coding-focused
And until 2.5 pro leaves first on lmarena there isn't a major reason to show off a better model.
meh, ok
yeah lol
Grok fooled me, I genuinely thought it was Claude telling me he couldn't do that π
2.5 pro is number 2 or 3 in creative writing sonnet 4.5 and Opus are both tied for 1
Anyone who claims fekin deep seek or kimi is the best justβ¦.
Has terrible taste
5.1 is, without any clear contest, the most sycophantic model I've ever seen
Wait, Brother Whale roleplays real good π³
oh boy!
Worse than 4o?
agree
raptor isn't that good
educated guess
I use this benchmark. Deepseek was the former champ. 4o was up there. 5.1 is actually crazy here
K do how do we go to the lm video arena arena
Please head to #1397655624103493813 for a detailed guide on how to use the bot
Why qwen3-max-preview out performs qwen3-max-2025-09-23 (stable version) in text arena?
idk
praying that riftrunner is gemini 3 flash, not pro
or some kind of "thinking-low" version of pro
it is the same/regressed on my comic transcription task π have been desperate for google's image understanding all year so this is a bitter disappointment
2.5 pro is actually pretty good at ocr, so arguablly the only direction for them is down.
grok isn't in contention for my personal task, it's too dumb, but i found that gpt5-thinking-high was a very strong image understanding update that almost equals gemini 2.5 pro. it's way better than any 4o or o1
i will also note that raptor-vision is quite decent at my task, a bit short of gemini 2.5 pro. if it's deepseek they have levelled up the same way openai have
my task is about annotating comic panel transcriptions with detailed description of body poses and expressions
π€ well then
Helloo ππ»
image understanding is mysterious. but my task also involves a bit of intelligence inferring what's important to note, so o4-mini is quite bad at it, anything mini is pretty bad at it
What Is Grok best at
6
6
3
Search
bro fr when is gemini 3.0?
what's the VIPER model?
wahoo they'll fix the image understanding
create a 58 seconds video with this text: clouds raining money
create a 58 seconds video with this text: clouds raining money
not working
How much video generate in a day
5
Because the preview is better
Is gpt 5.1 better that sonnet 4.5? π€
im finding 5.1 is really good. it's been improving my legal reasoning answers quite a lot over 5
Am i the only having this issue that the generated pic wont reload unless i reload the page?
sounds VERY promising
(though i hope it is not censored.)
nvm it is not open weight and behind login page that doesn't work
whats gpt 5.1s context window?
When will Gpt-5.1 be available on LmArena? I want to compare the battles
horrible at instruction following
me too, but the download button still works when this happen
Please fix this issues
ernie reminds me a lot of gemini for some reason, i like it so far, great to see those bench stats, thanks for sharing π
Hi. I'm new here. How is everyone doing?
@leaden sun
image gen
Anyone see this? GLM-4.6 is that powerful?
Sadly gpt 5.1 has arrived
and sadly its not that good
sadly there's no API
and not even 1 real benchmark from them
So they focused on personality, style and varying output lengths. But I'm gonna assume at best it is beyond gpt5 performance by a margin of error. Otherwise they would have shown it lol
Not impossible that it performs slightly worse than 5.0 on benchmarks either
It is able to write many lines of HTML/JS/CSS without any errors, so not surprised to see it so high.
I wonder how it compares to Kimi K2 Thinking.
Oh is it mainly for ChatGPT?
Any news about when gpt 5.1 is going to be available in API and therefore on lmarena?
yes
Their system card gonna satisfy all the safety enthusiasts but no one else lol
https://cdn.openai.com/pdf/4173ec8d-1229-47db-96de-06d87147e07e/5_1_system_card.pdf
A.I is crashing
why when i upload an image into direct chat it directs me to image gen models
Ayo???
Gemini 2.5:
https://gemini.google.com/share/b0152cb91a29
Although they say it's only for canvas idk
how do they know it's gemini 3?
eh just ab testing nothing new
Nope
We are getting it every single time
This currently with method
2.5 pro in ai studio 1 weeke ago
im getting this too. thought it was just me
It still says Gemini 2.5 how can I get Gemini 3.0?
Yes it's not "ab testing" it's ALWAYS gemini 3, probably tomorrow or in 12 hours we'll get it everywhere not just in canvas
cool seems real, cant be bothered going on mobile, but yeah should be in ai studio today prob
No models yet
hello guys, is it not possible anymore to use screenshots in text models ? i used to send screenshots to describe my problems, but now when i do it, the interface tells me to select an image model instead of text model
No you just press X on the image button above the textbox (not the image itself ) and it goes back to regular text model
when i do it i get this:
Try dragging it to prompt
yeah i get the same message
Or try pressing add a document instead of add an image
There's a way you just gotta find it
both dragging and dropping, and adding a file by the button do it
thanks bro, i just saw a thread about this, it appears that im not the only one facing this problem and the dev team are working on it. Thanks a lot guys
Idk i had that problem too but i know i did it some way
The fact that it made the size of each key exactly how it should be and spacing is correct ... This is insane
do anyone here knows about Google Gemini JailBreak?
Gemini 3 will be available in AiStudio this week or next. 100%
It's coming out so soon for a simple reason: if it's already this good, what's the point of Google keeping it hidden?
!!!!!!!!!!!!!!!!!!
hey, so.
which is better
https://openai.com/index/introducing-gpt-5/ or https://gemini.google.com/share/d6bc6dc5fbba
5.1 flopped
wdym whic is better
Imagine training gemini 4 on quantum computer
atp they should just drop it on ai studio
feedback is only positive from that one group
Was this 1 prompt or?
1
Damn
wish i could share the canvas but college account wont let me
on the topic of AI doom
damn i got so beaten https://x.com/ucfufd92897/status/1988908056395030667
thats nuts
so like
my AI doom was the best one
till this sh drops
minimap
sound effects
bruh how
The fact that ball rotates while moving and clouds are going slowly with each cloud moving at different pace and the sun and the game over screen...
https://gemini.google.com/share/29721ada09b3 BICYCLE RIDING ON PELICAN SUCCESFULL??
I have some high hopes Gemini 3 full release this november, because just softlaunching on app is bruh
does google have a conference today
not that i know of but i think they said something
about building with gemini on a discord event
wait
Ready to learn how to Prompt Like A Pro? Join us for our next Discord event tomorrow, 11/13, at 11:30am PT to watch Google Deepmindβs Director of Product Management @triswarkentin spotlight real-life use cases and prompts, then offer tips to take your prompting skills to the next
yeah
maybe they are routing us to gemini 3 just for that lol
That might mean we'll get nano banana 2??
yo thats cool
how much usage you got
i got like 100 prompts a day
or idk
its a college account
so we got a business plan
can you make one i request rq
flamethrower sim where you can spawn some objects to burn
also gpt 5.1 is also not immune to seahorse emoji
this is nuts
what is this exactly
mq 9 reaper simulator
needs some improvements
but for first try this is actually insane
what stays in education stays in education?
this is just blacked out for you?
Wait dod you have to sign for college somewhere or did you just enter with college account?
i have a college account
which has a business plan
Me too
just send the code?
ok
space invaders is the easiest shtuff you can make
AGI
huuuh
i doubt this is a svg
this is either iframed or image
check code
or give me it ill check
from my understanding
this links to a script
yeah just how i thought
https://unpkg.com/world-atlas@2.0.2/countries-110m.json - this is a file that contains country positions and names
polygons and stuff
half authentic since its not really drawn
not from memory
which model made that other one
same
that one was better tho
prob because of the json file reference
it knew coords and shapes
not bad
the first model that generates a terrain on roblocks that you can call decent
Where is gemini 3.0
Only on canvas?
space invaders by gpt 5.1
will you ask it to make a lava terrain
i'm finding the 5.1 model is much better. quite happy with it's output. could always have better reasoning but it's not bad at all
instant or thinking
thinking. but both are better
GUYS IS GEMINI 3 OUT
only on android canvas
and only on the pro model
wtf is android canvas
android app canvas
i would say its more of a rollout than region
since i got it
im in EU
free users only get a few gens from 2.5 pro daily
hold on im trying to do a fix
you get few prompts free for 2.5 pro everyday
paid users get more
but its canvas
we dont know the full rollout because ith asnt happened
@zealous sparrow can u recreate the Doom game, to see if its better than riftrunner
Was already done
Yes
wait can u forward it or reply with it
It's there
it was from first prompt??
https://x.com/ucfufd92897/status/1988908056395030667
bro i got cooked
so, gemini 3 coming next weeks by the looks of it.
im happy that its better than riftrunner
fr
@stray aspen can u do some three.js tests with it
thats the preview models tho, not the official release
do yall have the paid subscription for it
no i just have many accounts
but @zealous sparrow said free users dont get gemini 3
but ill guess buying advanced soon
no its avaible for free
u can select it?
free user get gemini 3?
how u know its gemini 3
its out?
yeah 2.5 pro. why u lying
nah
@stray aspen how do we know if its gemini 3
https://www.reddit.com/r/Bard/comments/1ovvmjo/not_hype_posting_gemini_3_is_rolling_out_to/ just check ts or smth
bro trusting a reddit post
on mobile?
im on canvas yeah
android
but it says "gemini 2.5 pro" under the model name
oh i forgot did u sit in app version
wow
i have the app downloaded
???
did u writed there?
yea
<@&1349916362595635286>
its only for gemini 2.5 pro that moves u to gemini 3
then ye it is
free user only get 5-6 usage per day
its the same as gemini 2.5
pro
so if i run out of gemini 2.5 pro, thats gemini 3 credits now?
its also thinking alot
more than 2.5 pro would
no i didnt really say that
try to make something else
i said
im making 3D flappy bird rn
that free users only get a few generations of 2.5 pro daily + canvas [gem3
And paid get more!
u need to buy gemini ultra to get more
yeah i finally got it
can it create image?
idk, canvas is like a coding section
can i ask some people to run it for me if i get ideas for prompts
i cant use it
now fr
btw i once put old minecraft code in gemini 2.5 pro and it tried to make creeper
give me ur prompt
ok i actually would like to see result with gem 3
u threw in the whole code?
damn
yes
it seems like Chatgpt has been neutered. They blocked legal advice.... mfers
why would u ask chatgpt for legal advice. contact a lawyer
"can you make single file html code for physics simulation of flamethrower, you drag it by left mouse button and if you hold space while dragging it, it launches fire particles with gas-like physics, they get turned into smoke and then disappear. use matter.js"
can you try this one
Maybe we don't have money for a lawyer lmao
Not all 'legal advices' are really that deep... It litteraly is written somewhere on the government website or something
gemini app is on android, so i can make it but i gotta give u the code cuz im on my phone rn
imma put the prompt in
what if the chatbot hallucinates?
k thanks
i know, i cnat sue it on android
oops
holy misspelling
Well we're aware of those risks... It should just give something he thinks and than note below not completely deny the legal advices
not everyone will go to site to verify.
i have some problem with it
ran out of my usage and im not bothered to go on my phone rn
lneduo got lots of prompt
bc college acc
wait ur using it on web?
make new acc for more usage
or app
its only on app
if only we could get a college acc
no i did one gen on mobile and that was all i used it for
which checkpoint do we think google took
definetly not riftrunner
gemini ai ultra gets 500 prompt a day to test
@leaden laurel
its on point lol
@leaden laurel also i made ur prompt
identical to riftrunner..
lemme check
nah its even worse
eeh i dont think its really gemioni 3
ehm
https://g.co/gemini/share/1116949c6da0
here's movile bersion
hmm, wha
but it looks better
people spit at riftrunner for being the worst gemini 3 checkpoint
but maybe it was the best?
ill ask it myself
not really bruh, it made good gens for me
legal strategy advice. It's better than many lawyers. I know because i've been dealing with lawyers, and chatgpt has to rewrite their arguments into something that makes sense.... or sometimes the lawyers arguments are literally incorrect
Gemini 3 come out so i can get better legal advice
which is better riftrunner or lithiumflow
People legit hate riftrunner..
they need to release x28
we dont even know what checkpoint we got
im going to so be laughing if its riftrunner
people will be so mad
And i will be happy that it got a shot
Whatβs the best AI to Edit pdf?
Although judging off the doom clone..
https://gemini.google.com/share/9fa043d54cd0 it looks a little better
It's near identical to riftrunner..
did u play with the prompt
or is it the same
same
thats really cool af
It made the fire, make the water disappear
And also the sand is sticking somewhere but the water won't stick the same way
Im like so sure the gemini 3 released is riftrunner
bro how are your examples so good
the doom clones being identical
doesnt help me think wrong about it
the doom clone on X looks alot better than mine
u need to write prompt better
for mobile
iOS too
on gemini app
Think about it.. They are identical..
smth smth
only on gemini app anywhere
no, they have working doors
on there
canvas mode
also the map is way bigger
u have to use gemini 2.5 pro + canva mode. it doesnt explicity call it gemini 3
yeah they wont release that for free
Desktop vs mobile Gemini, I think itβs legit
its deep think prob
bro why does it always give it a hat now
Gemini 3 is a freak
i think we either got riftrunner or lithiumflow [no actually what am i thinking of]
this is the gemini 3.0 canvas?
fr no other models comes close
riftrunner is still on lmarena right
yea
Honestly if google gives us riftrunner im not mad
The twitter group only wants the best checkpoints
which are most likely gemini ultra
riftrunner is what i imagine gemini 3 as not too good and not too bad
riftrunner is very good
riftrunner is goated
imo i think riftrunner is better than the canvas model because the flamethrower sim had better particles
the doom clone from canvas is better than riftrunner but prob because of some prompting
Have we ever wondered what kingfall is
Why buy Gemini, Chatgpt, and other subscriptions when you have LmArena?
We are waiting for gpt 5.1 on LmArena!
hope riftrunner will be flash 3.0 and lithiumflow is pro
lmarena is quite limited unless ur just doing 1-2 shot requests
still no API btw
riftrunner is too good to be flash
hopefully riftrunner is the general release checkpoint
Perhaps this is the only drawback. It's strange why Api is still not available
the model was in the confidential category so i doubt its any time soon
thats an internal model for testing only
yeah
hopefully the app version isnβt garbage like the app version of 2.5 pro
why did google even add a confidential model into aistudio
there isn't a big difference between riftrunner and lithiumflow
gpt 5.1 cant do svg right so i doubt
gpt 5.1 is literally a chat model for women who seek fantasy
support their delusional ideas
i benchmark models by seeing how good they can make a henry stickmin game
if you dont know what a henry stickmin game is
you pick a choice
see it play out
sometimes theres QTEs
give me da prompt
That are the testing models..
hmm, maybe because Gemini 3 didn't have a release yet
you will feed to gem 3?
ye
henry stickmin like game in html with animations, endings and episodes, and achievements.
gl
I think gemini 3 will top lmarena by a good margin but will be on par/slightly worse than claude/gpt for coding/stem in real use
I love codearena because the amount of time it takes to generate gives me time to study
5.1?
code arena isn't that good tbh, some models just slow it down
still use it, it might take long to gen but
gives me time to study
because i am a very distracted person
lolwut
if your asking why his message disappeared
@stray aspen what did u create so far
gemini 2.5 isn't even good at writing
it was good when it was experimental
bet no one here remembers gemini experimental 1206
it is, you just gotta constrain the hell out of it or you get purple prose soup
I have to say, I expected better...
Barely any animation. It's just the stickmen jumping up and down..
Im goin to try to run this with riftrunner
pretty sure its still on the arena
i love how the other choices immidiately kill you
there's only one ending w
in henry stickmin games, yeah some choices do kill you
also yeah, i said multiple endings..
Try asking him to write a Minecraft clone in Canvas
u need multiple ending
Hmm
mine has trees
how do i know if riftrunner is still on the arena
guys
I think ios gemini app canvas reoutes to gemini 3?
nice
oh yeah saw this too
google cooking
did i really just get two claude models on codearena
no i know one of them must be lying!
this is much cooler than gemini 3
this is gemini 3 pretty sure?
minimax m2 likes to like that its claude
why sonnet 4.5 soo low?
hi guys, I know gpt, sonnet, but what is glm 4.6?
general language model
yes l use 4.6,it is pretty good.
chinese Zhipu AI company
I think it's an cheaper alternative to Sonnet 4.5.
because people judge off of UI
This can tell glm 4.6's priceπ€£
bruh why is riftrunner 77%
x28 π
Gemini 3 Giga
there is no gemini 3 ultra coming
lithiumflow above ecpt π
what kinda benchmark is ts