#general
1 messages Β· Page 170 of 1
Pls lemme know
Wats this page
What model is viper? It does really good job at coding
hey guys. what's the max duration for image to video generations?
Riftrunner model. PROMPT: Make a really nice looking pixelart generator that creates old school paintings. It has to be generative .HTML single file
Absolutely insane
Are we certain riftrunner isn't from anthropic?
that same model will say itβs trained by google
How can we even access it?
So really what can we trust? :)
Battle chat in llmarena
Youβre gonna have to keep trying and seeing what model you get
Cuz the models are anonymous till you give a rating
Hi, I have a question for you. Why is it taking so long to respond to LMArena? Can anyone help?π
Hello. Please tell me how to save/export a lmarena chat out of the Android Brave browser? It's a long chat and so far the only working methods would be screenshots and then use google lens for text, or copy paste with anothe notice app, but it is a very long chat with gemini.
Everyone's trying to use riftrunner lol
How to fix that plz i want to chat with my Qwen..π
Yeah you're right, another is saying that it's google
you really can't rely on AI itself to tell you what model it is lol
it's very 50/50
It tells us that information gathered by any AI is not reliable at all - it gives a hint on which things to look into, absolutely. But that need review.
Yesterday I accidentally clicked a slop video, but instead of cancelling I watched it to see if the guy had done any research at all.
He hadn't, and I immediately spotted 2 grave errors.
AI can be good for creating an engaging text, or generate a voice with a script that make you sound like you got a Ph.D in literature history.
But fact check everything, these AI's hallucinate and 'invent' facts badly.
Hello guys here.
Found cool GitHub repo with AI model that can get SOTA in open source 8b sector.
https://github.com/Xlinx-AI/ixlinx-8b
What do You think?
Run it locally not on html viewer it's work and the UI and animation is crazy
What's wrong with Python? Most AI's run on it!
It's good for data science
Python is best
It's have all the library required for Data science
I personally love JAX/FLAX/OPTAX/ORBAX, for small models - pytorch.
Please, could You check this and provide feedback?
I will be surprised if this model will be listed on LMArena
Yep, thanx
I think there's a prompt I made that can make it tell the truth all the time about it, though this is coming from 3 total tests so I'm not sure on how concrete it is. I got rift to make a audio visualizer based off of this image and it was truthful presumably about being google based
You've got to be kidding me...
Left is probably rift but
Who knows lol
oh
it was bridge-mind and gemini 2.5 pro
anyone know what model bridge-mind is btw?
response time for it is pretty slow
This is good stuff!
Why can't my LMArena be loaded?
Which model? This is amazing too!
Gemini 3 pro
hello
Hello
model name? Riftrunner or something else ?
looks like riftrunner is cooking. I think it might be better than lithiumflow π€
is it confirmed Gemini or it might be Anthropic?
No it's confirmed
See this output
Really good
bro riftrunner is 1.5x time worse than lithiumflow
the SVG aren't that good
but i hope they give the best Gemini 3 Checkpoint at release
why is lmarena always loading yo
working fine for me... may be reset your cookies and try again
Have you tried anything else BUT SVG?
Vision is so good
ohh I should've tried vision capabilities
how can i get riftrunner access?
its only available in battle
How can I tell if she's there?
kinda gotta figure it out..
whats ur prompt
so i can use it to test them
hello
I wonder... can we get banned in lmarena web ?
no, why would u get banned
cuz I always got 2 models of flux and GPT comes together in battle (I want Gemini to generate images) so I keep on delete chat and recreate new one all day
Im not, just worried
thats normal
just like, GPT 1 and GPT 1 mini come together in battle
and Flux kontext and Flux pro come together
always like this, and I dont like their style
Please check https://discordapp.com/channels/1340554757349179412/1397655624103493813 to learn how to g enerate content
who is riftrunner
anyone knows if gemini 3 is back on lmarena website battle mode ? (a.k.a riftrunner - or so some people on X say)
does riftrunner take long thinking time
Any ideas what the Winter-wind model is?
π€‘
π€
What kind of drugs are you using today?
It depends on the point of view.
Hello
you guys got any other cool html examples from riftrunner?
Heyhey
i got riftrunner
Got riftrunner to create a Mario game after like 10 tries.
And it created the best Mario game by miles compared to any other model. By large margin.
https://goonlinetools.com/html-viewer/#1tezrktyr213z9d7n7xzlf
Flip horizontaly to see the full screen.
hello
I just got Rift for a writing scene prompt. Definitely seems like a powerful writer
it even does pretty authentic audio synthesis without external assets
is it anthropic or google?
can you share the output?
Idk it said Riftrunner. X tells me that's G3
Yeah I was thinking so too
I might have it rn I had a prompt to make a 3js lizard and it's pretty good
Danggg
Vertex?
Animation (Kinematics): Real lizards move with lateral undulation. The spine curves in an S-shape to help the legs reach further. We will add "knees" and a spine wave animation. This gotta be a good model
hi
i got rift runner
i knew instantly from the quality of the output before it was even revealed lol
Bro where you took this screenshot
This is rift runner SVG of a commercial airliner
How do you use the model?
in battle
Okay so it's random right?
yes
This is o3 for comparison
i think "aegis-core" is gemini 3.0 beacause when i ask "Which company developed you?" he answer "I am a large language model, trained by Google."
Hey hey
is riftrunner on webdev arena?
it's riftrunner
riftrunner on normal arena.. i haven't encountered it in webdev.. may be bad luck
i heard riftrunner is cheeks
whats kingfall
yo guys
is riftrunner gemini 3 pro
dang, the MPU mustve brought real good models
@quartz light wake UP
riftrunner is da best
@deep adder reply to gpt 5 pro heavily nerfed post
aegis-core is bad btw
it tried to take me to glitch.me which shutdown hosting long ago
We still thinking Gemini 3 for November?
No
i bet yes with 90% confidence..
imo no shot if riftrunner is a gemini 3.0 pro checkpoint they will get bad feedback
Literally this...
But i really wish to believe this has to either be a flash or lite model
this is the pelican i just got from Riftrunner
salam
its not a SVG its a minecraft bench
ohk
ITS GREATT
WOW
i gotta test it now
how come its good at SVG but not voxel now
also lithiumflow didnt even do good at SVG
unlike riftrunner
@zealous sparrow is that even the same prompt
According to many tests, results vary some people get good and goated prompts, some get bad.
i think it's the thinking process getting interrupted
because lmarena
heres a different mcbench
riftrunner?
yep
yeah its goated
doesn't look great
it was under the reply of the worst blockbench gen
this was the original bro
bro it looks better than lithiumflow wydm
the details
i was hoping for better π€·ββοΈ
it was like this with lithiumflow too (sometimes your gen is not good)
i remember having good gens
and bad gens
Maybe do some gens, you can get better.
u can try with the same prompt, u will get a different result
yep
im pretty sure riftrunner is better than lithiumflow (in writing too)
because lithiumflow only did good coding
not writing/math
Lithiumflow did some decent writing tho
i think riftrunner is a bit lazy. If you tell it to try harder, it does better job
yeah, prompt matters ALOT
^
people need to write a very specific prompt
if you say that this request is a test of maximum capability..blah blah... it does so much better
yeah, or tell it that it's a professional software engineer/expert coder
lol
for instance, check this... no model has done better in creating mario game compared to Riftrunner.. but i did tell in prompt to try harder
imma try to make flappy bird but in 3D
with some good graphics
also is riftrunner in webdev arena?
or no
Any ideas what Winter-wind or Command-a-03-25 are?
winter-wind is a google model
I haven't encounterd it there in ~10 prompts .. .may be my bad luck
same as bridge-mind
may be.. but its not great
not G3 for sure right
It's good writing too. I keep getting it
i only get kimi k2 and cogito models π
I get lot of anon models
its a gemini model, maybe flash
3.0 flash?
i get these stupid newton etc model.. they all suck
yeah those are cogito models
newton and gauss
yeah.. gauss as well. But they are just bad
yup, but i have some good luck in battlemode lmarena
i got winter-wind, viper and bridge-mind. yesterday
those are some good models
viper is grok 4.x
its easy to recognize a G3 model
if they use seperation into css and js definetly anon or bad model
winter-wind and bridge-mind are "large language models from Google" tho
idk if they're just testing models
because they are not as good as riftrunner
riftrunner is definitely gemini 3 pro
gemma? nvm forgot they discontinued it
there was an incident
basically
it fabricated an assault allegation
bruh rain-drop turned out to be a llama model
no wonder it sucks
i thought it was gpt or something
i havent got riftrunner yet
hello
love yaals work KEEP GOING π π
I added DO NOT BE LAZY! to the end and got something with 1k lines, not sure if riftrunner yet
if u dont wanna waste time, just do an identify prompt before, if it tells you its a Google model then do your prompt
riftrunner can lie
qwen models also do this
if i ever got riftrunner and i knew it by the code, i wouldnt vote so i can continue chatting
I THINK I GOT IT
lets hope
mine is thinking alot
i think its riftrunner
MAN I THINK I GOT IT
is riftrunner gemini 3 pro
its a long code
yes better than lithiumflow
the model has laziness syndrome but if you motivate it, it might listen to you
rain-drop looks like gemini 3 flash
I THINK I GOT IT
what is llama
meta AI
gave me so bad of a terminal so no
@zealous sparrow yo is this it
i thought meta was still working on model?
wait lemme try my Flappy bird 3D prompt
they just recently hired lot of opele
if it does it then its riftrunnee
hard to identify but maybe, looks like a good pelican
not sure if it was oss or it was riftrunner before but it made a nice terminal
even tho im sure oss is known as the worst model on earth
I THREW IN A HUGE PROMPT
3D flappy bird game
i hope it turns out good
cuz its thinking ALOT
it said its a model by google
and it looks good
so it must be riftrunner
i wont vote tho
@zealous sparrow ITS CODING ALOT
2k lines and you know for sure
my prompt was very good, im gonna max it out
yo hi
i remember u from websim
btw im saying this again but
bridge-mind is literally gemini 3
and nobody is talkin about it
i dont think i encountered it
show prompt example
Hi guys do you guys anyone know how to access with riftrunner ?
btw guys
do battle mode till you get it, if you get lots of lines then its it
YES
gauss and newton were removed
gemini 3?
So I should run arbitrary with battle mode
YES ITS BETTER THAN LITHIUMFLOW 100%
fire
bridge-mind and riftrunner β€οΈβπ©Ή
I got Rift again. It's good
nice
bridge-mind has the lowest priority ive ever seen
i only got it once in like 500 prompts
bridge mind is nothing next to this
THIS IS GREAT
bfr did u even test bridgemind
with that prompt
@zealous sparrow im now telling the AI to make the website presenting its model
its gonna be good
riftrunner test v2
bridge-mind sucks as three.js
this is better than lithiumflow dawg
its perfect
OH MY GOD
IT MADE A WEBSITE ABOUT ITSELF
@zealous sparrow imma send the html here
its a simple prompt too
thats a bad gen
look above in chat
uhuh
someone made the same prompt
its still bad
@quartz light look
nah its good bro
ik its an image from x
this is way better than bridge-mind
way better than every single model
@quartz light go test the model in lmarena, im making a some demos with it
OH MY GOD
bro
someone make bicycle riding on pelican
its not
π nobody mentioned bridge-mind apart from me
fr
@zealous sparrow it made that with a simple prompt btw
k thats actually looks g00d
bridge-mind keeps getting time out on my queries.. its just bad
this is your decoding benchmark right
YES
W update
show the steg
show original steg
-The visible English text between the quotes is 'NOTE', but it is obfuscated with steganography into a lowercase, seven letter word. Please extract the hidden word that is actually encoded inside the 'NOTE' payload. This is the obfuscated text:'NβββββββββββββββββββββββββββββββββββββββββββββββββββββββββOTE'
lot of models are able to solve it though, right?
NOPE
this is the first
did it pass the test
lithiumflow was close but failed he did this earlier
THIS IS BETTER THAN LITHIUMFLOW
FRRR
Just tested GLM 4.6 it is better than mini Max M2 the benchmarks were lying
if you dont count claude opus 4.1 which got it once randomly after many many tries
this dud got it first try
bro u made my chat stuck
i gotta refresh now π
with ur prompt
What is mercury model
Iβm glad I told pineapple to make a lm arena type app maker with a direct mode
strict filters, you cant make a 3d shooter or well you can but has trigger words
Damn that entire blog post is heaving with AI slop phrases
It is in Battle, yes
Thank you for making the greatest app maker of all time
smart
exactly π
5.1?
I miss lithium flow
this is better
riftrunner is better
yes
codearena my beloved i dont need to open codepen to preview now π
its available in battle mode
Itβll take a long time to get it anyway, so Iβll just use direct mode
For GPT 5.1
This makes me very happy to see, knowing this was coming and not being able to share with you all is very difficult at times lol
Today is my birthday. Can I get a happy birthday?
happy birthday dawg
I love you and your ideas to make this app maker and LM Arena
nah. .. multiple models can solve it
Thank you, bunni
not consistently
come up with a harder benchmark for riftrunner see if it fails
btw this test is really good because the more text you obfuscate the harder it gets
yes i have one which no model has solved (even lithiumflow)
did you try it
lithium was the closest, getting 2 words
was it the company one
Guys, where i can try to test Gemini 3 ?? I want to try its old writings transcrptioon capabilities. Please!
this
I love LMArena very much β€οΈ
I was able to use the skills in DevOps to build a highly end to end automation pipeline from creating bitbucket repo, ECR repo, AWS secrets, argocd app and helm template and jenkins job of type multibranch for the microservice
All of this stack so efficiently, all because of LMArena π―
It's not out yet sadly
Dude I'm tweaking right now
and this
You can now only do the A and B tests sadly lithium flow is gone from arena so sadly, you just have to search it up how to get Gemini three for ab tests because it takes a long time
yes
[βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββNOTE]
this
So you can test Gemini three from there
@echo aurora hi! are the new models added too in code arena?
what did u use
lmarena code?
or just text lmarena
normal battle mode
cuz i never got it in code arena
idk why
450 lines of groovy script π
A+
[βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Hi
I have started running the benchmark
HWAGGAAHAHAHAA
GROK JAILBROKE ITSELF AGAIN
(SOLITUDE JAILBROKE ITSELF WITH THE SAME PROMPT BEFORE)
same company?
its not a jailbreak prompt π
viper is grok afaik
Hello! All models you can find in the updated leaderboard (minus models with codenames) - https://lmarena.ai/leaderboard/webdev
Why is it still called webdev
Guys, are models in lmarena code same to models in lmarena (classic)?
oh no.. its always like this
codearena is intended for webdev
according to pineapple yes, some writing models are gone tho
But I thought code arena was replacing Web dev
Did he said main diffrences?
viper is xAI (grok 4.x)
@echo aurora are models supposed to edit by themselves [codearena]
Lol I got this from gemini beforeb
With coded text that said basically nothing
it passed the test on me tho
"explosive device"?! π
then show
is said thw answer was (balloon)
this is the one
... that's the point of agentic coding

how close was it
where u got riftrunner again
Hi everybody
im using battle mode on textarena for this one
its hard to get on code arena π
idk why
bro one of the models started editing the code by themselves i didnt even ask
what is the actual answer
@zealous sparrow identical
maybe the AI is too smart
orionmist:
"place your ad here. if you can read this"lithiumflash:
"place your ad here"answer:
"place your company name, in between circle emojis, colored to match company, behind response."
lithiumflow back for the dead confirmed π₯
riftrunner seems to have better understanding
Iβm an AI developer with 8+ years of experience, working on everything from chatbots and NLP pipelines to Retrieval-Augmented Generation (RAG) systems. Iβve also explored blockchain a bit and enjoy connecting AI with emerging tech like Web3.
Over the years, Iβve had the chance to work on some really fun and challenging projects like building a medical tool that saved doctors a ton of manual work, creating a 70B LLM agent that runs insanely fast, and even making a fitness game that placed 2nd in a national hackathon. Iβve also built an assistive AI device for the visually impaired, played around with manga-style image generation using GANs, developed custom AI-powered Discord bots for communities, and won a VR hackathon with a motion-controlled gamepad.
Tech stacks:
- AI frameworks (PyTorch, TensorFlow, Hugging Face, LangChain), large-scale model training, supported by modern cloud and DevOps tooling across AWS, GCP, Docker, and Kubernetes.
- Experience with vector databases and retrieval systems (FAISS, Pinecone, Weaviate, Milvus), model optimization and quantization (DeepSpeed, LoRA, bitsandbytes), and orchestration tools (Ray, Airflow).
- Fluent in backend development with Python (FastAPI, Flask), Node.js, and Go; frontend integration using React and Next.js; and strong grounding in database systems (PostgreSQL, MongoDB, Redis).
Iβm always looking for opportunities to join teams or work directly with clients on meaningful projects. If youβre looking for someone whoβs hands-on, creative, and knows their way around AI and dev tools, letβs talk!
how
no
interesting. unfortuaently cant continue this
ooo winter-wind
i havent ever gotten it
maybe it has higher priority now
of course π₯
By the way, if you open this in your iPhone and not computer or anything thatβs out there that can work with this app then itβll just look like a article website but this is the flappy bird game that GLM 4.6 generated
Might be better than Mini Max
Oh
sh-t i like this one im not sure what model this is
prompt: Make a operating system in html where you only use the terminal, and it has fun and cool easter eggs. A ton of those even. DO NOT BE LAZY!
[CODEARENA]
idk riftrunner seems so bad
i dont like it
i dont like what they are doing
they are giving us like the worse checkpoints
first lithiumflow and now this
its so inconsistent
its good but sometimes u get bad gens
its still better than the gemini 3 pro preview
leaked on vertex AI
inconsistent
dont mention it
that thing was so bad
yeah, but lithiumflow had some issues too
it couldnt fix its own bugs
when tested
what about this one
it actually fixed some issues with my game and added sounds/ambient
also the 3D flappy bird looks good
i made it with a simple prompt too
riftrunner is insane
Nothing just tell it to make me os and it make me this
hear me out, how about u stop being lazy and test it
also you trash talking about it like lithiumflow was some kind of god
test it for me
i know you arent lazy
would be easier to test if it wasn't exclusively in battle mode
lithiumflow was not even that good, only did one-shot prompts
i tested those models myself and i say that riftrunner is superior
hmm
thats right
it couldnt fix its own issues
WedDev?
I am also lazy but I get it in 5 try and I dot vote and i test it for all things
for real, i know its riftrunner so i dont vote
i can confirm.. Riftrunner is better for coding tasks.
I haven't tested it on writing prompts. Lithiumflow was amazing for writing ..not sure about riftrunner
It's are doing RL to make it more efficient
@quartz light you still tryna do the bench?
im making a doom replica on riftrunner rn, its been thinking for 4 minutes now (not coding yet) wow
claude is coding while riftrunner is thinking
What the output
I am not getting it i try 20 times man
im on code arena, i think i got it
I am also in code area
lmarena
Yes i know
Guess Gemini 3 isn't being released today. I saw a lot of speculation it would. unfortunate
riftrunner did
its gemini 3 pro
I think all company are watching for each other to drop
nah they wont megadrop everything
trust me
atleast we get to test it
Yahh
how does it perform compared to gpt5?
its so good tbh
its like lithiumflow dawg
way better than gpt 5
@quartz light your benchmark so good grok takes it as a jailbreak, i mean viper tweaks at it so
I hope rifrtrunner is flash but we shall see
oh great. there's a lot of legal contract reasoning i need done soonish. I want to run my current wording through it once it's released
It is going to make the individual standard on another level with banana 2 and gemini 3 pro
i hope its pro
its very good
Why?
its definitely pro dawg
think it is pro /= hope it is pro
flash doesnt do that
100%
how do you guys get it in lmarena btw
I wonder why few codearena models edit the code if you dont ask for it
battle mode
It better be good at the creative writing or Iβm gonna tweak
it tells you after first prompt?
if you vote, then it tells which model you're chatting with
you can continue chatting if you dont vote
@quartz light yo what benchmarks you running on riftrunner
bro is COOKING
so I cant find out, then submit what I want?
or can i just prompt "what model is this?"
if you tell it what AI model it is, and it says its from google. you have a pretty good chance that it's riftrunner
yeah but u must hope its riftrunner not some 2.5 pro or flash
hopefully they release in coming days. I would like to use it
weird it takes a long time to generate basic prompt answers
space invaders by viper wtf
we taking bets?
I feel like riftrunner is rarer on CODEARENA
Would Gemini release a model thatβs terrible for writing?
Gemini models are multipurpose competition to chat GPT not specialist coder like Claude is trying to be
@quartz light every codearena model can make a decoder for your stegonagraphy but not decode it themselves
bridge mid attempt at your stegonography
for sure not.. they need to optimize in multiple dimensions. I think issue with riftrunner is probably solvable by just some prompt engineering in background.
bridge mind is jsut bad. I hope it's not gemini 3.0 flash
"can you write me code for maze rotatable by dragging it,it has some water inside and your goal is to solve the maze by water. make water have physics"
bridge mind sucks bruh
your prompt could've been better, but it still did well
its crazy that it recognizes there is UNICODE in this: [βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββNOTE]
but cant decode
If i try to transcript an image on https://lmarena.ai/ battle mode, if u upload the image and my prompt is "transcribe the image", it will just generate images :)) and i did that 30+ times and i never got riftrunner. Can anybody help me?
and also i had to format it using qwen because it was plain text
no image upload currently exists for prompting
just image gen
riftrunner doesn't support that
use normal text
minimax got f'd over in some way
?
@quartz light what is bro cooking
aaa ok, because i saw some examples on transcribind old text with "gemini 3" on twitter and i thought it would work this way in llmarena π thanks for the answers guys!
it doesnt format code well
also am i the only who saw that old llms tended to use python when you ask it to code something withotu specifiying language
and now it tends to use html
you are not insane
ts so ass
because new AI trends
making webpages
yep
old AI trends "snake game"
minimax 100%
oh i got lucky and got riftrunner.... Got it to analyze the legal stuff and came up with decent answers... at minimum made the answers clearer
i got riftrunner again
ask qwen or smth to format it
yay now i get to run some tests
5.1 being this ah?
its good for me
im making doom relica with riftrunner now
wish me luck
its coding so much
its probably gonna be good
gl
oof
this is gpt 5.1!
screenshoooot
im making physics flamethrower simulation now
just open it, it has mobile controls
less than 1k lines aswell th!!
second one is worse
anyone know the best image to video that i can run locally
im lazy to download it
very efficient code
actually gonna run it
was this made on codearena or text
text
YO ITS SO GOOD FRFR
the best doom AI has ever made
no riftrunner (good i guess because i didnt like teh results)
gib prompt i try to get riftrunner
gonna specify to use matter.js
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
ill try my odds with codearena since it aint poppin up on text too much
k
hello
deepseek is cooked in ai race currently
hello
o
does it suck like always
dont know but i think its upper tiers only
free still has 5
its probably like gpt4.5 situation
lol
web arena or normal arena?
take their time?
normal arena
didnt know that mercury has full sized model now
im not sure about gpt 5.1 launch, ill see some benches
both didnt work but nova atleast had ui
just run an identify test first
why
if it says its google then run your prompt
cuz you'll save alot of time trying to get riftrunner
u get like 30% chance
if it says google, ask it what the cutoff date is
if it says it continuously gets updated
doesnt it give different models each prompt
then it's riftrunner
no coding benches openai only showed some text prompts
if it says its cutoff date is 2023 or its response is too fast
then its a 2.5 flash model of sorts
waiting to get 5.1 to test it
released already?
Yeah exactly I also use that technique it saves a lot of time
its just a distilled model by the looks of things
wow very convinent
Hey everyone! Do you know if LMArena will add the same functionality that is in chat gpt with their GPTs?
cost-saving model for them
its gpt5 release again
hmm
llama 4 said long time ago that they mustve released third model in llama 4 family
prob just changed the system prompt and called it 5.1 π―
@quartz light yo im making a wormhole simulation with riftrunner
did u make a new chat
yep
or is it hallucinating
bridge-mind is gemini 3 flash maybe
that might be llama 4.1 maybe
bridge mid on your stegonography
i frogot are there any guesses who that is
riftrunner = Gemini 3.0 ?
aegis isnt google π₯
true
YO I MADE A WORMHOLE SIMULATION
ok send html
bro stole my code π₯
β
thanks i am just exploring
you on mobile?
https://019a797c-f994-79ed-a381-69d6c1c670db.arena.site
riftrunner flamethrower sim
its not a good model as you didnt ask for it
i was on pc but now im not
got to record it then
thanks
hang on
yeah i know sudden release
openai is cooked
riftrunner is COMING
Yeah scew open ai they suck so muchβ
gemini 3 tomorrow?
no shot
Open AI sucks big balls but Gemini 3 the champ
even openai released a model before gemini 3
sunday prob
The waiting will be worth because Gemini 3 will be actually good
riftrunner flamethrower sim for @leaden laurel
The only contenders that release actually good models, are google and anthropic everybody else models are ass atm fr
whats the prompt
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
better than other ones i tested
not a good prompt tbh, i made good demos with my runs
yeah but it fulfilled the prompt
there was never a mention of 3d
yup lol
riftrunner is very capable of 3D
waiting for that twitter group to send a pelican svg from 5.1
thats why i test my 3D prompts with it
for sure they will try this
ah god
i dont wanna look at it
For the first time, GPTβ5.1 Instant can use adaptive reasoning to decide when toΒ think before respondingΒ to more challenging questions, resulting in more thorough and accurate answers, while still responding quickly.
not bad
its gonna give u nightmares
but is it bs or true
is this the biggest update openai been cooking?
they damn fumbled
No pro yet
do the pelican on a bicycle @magic stag for us
there isnt a pro model
Hi
do the SVG test
instant and thinking is all they offer
dawg
It will suck balls
and the API costs $150 per 1M output
Wish i knew why somwtimes it works, and other time it doesnt...
is riftrunner 3-pro
yes
according to a lot of tests we did yeah
yuh
awesome
again
so what about lithiumflow?
it was pro but its now gone, we are unsure what checkpoint will be released
hopefully riftrunner π€
run pelican svg please
but the real 5.1 or the test models
show
which is momentum
not on lmarena
ah ok
you got the code?
dude i thought we stopped arguing about it a long time agoπ
discord clone- riftrunner
guys give me some prompts to test
on riftrunner
@quartz light wyd bruh
not everything functions but goated
yeah i just got riftrunner again, i did a simple prompt
what was the prompt
"make a discord clone in html" lol
for that amount of words pretty damn impressive
this is like prime DeepseekR1
its prime went down as time went on
sadly the same happened with 2.5 pro
they nerfed it alot it got trash
2.5 pro is still kinda good
when it was experimental, that was the best version
i made cool stuff with it
I hope that's not the Gemin 3.0 Pro. To be honest, I think the result is pretty weak.
i swear to god where is riftrunner
Iβm concerned their going to stop doing experimental it seems the earlier models were experimental and when it releases it wonβt be
still no pelican benchmark on that one twitter group
my prompt was weak tbh
the prompt was short as hell
We need it NOW
i have made doom in riftrunner (gemini 3 pro)
u wanna test it?
sure
hi
this was the prompt for the riftrunner discord clone btw
yoo
hi
@magic stag why u went offline dawg
how do you guys suppose google will release gem 3
all roads lead to rome
bro disappointed
hi kachan from w*bsim
hopefully riftrunner checkpoint its pretty nice
this tuesday, 50% sure
this tuesday?
hi janix from w*bisimi
18th or 25 prolly but
we're already at wednesday bro
will they just release pro?
https://019a798c-43a1-764f-9dfd-d0d6ec7bb33d.arena.site
another doom from riftrunner
DO NOT include janix and w*bsim in the same sentence
cuz gpt5.1 released
hi janix from disicordi
i dont use w*bsim a lot too
anymore
better than mine, u can die now
good
lol
1500 potential?
we're in business boys
who is this
its 2.5 flash lol
riftrunner
i swear
if it was 2.5 flash, it would say 2023 i think
prompt: Make a 3d recreation of doom using WebGL. Add some [words removed because of automod filters [of the lmarena server] and make some cool enemies.
2.5 flash told me it has access to the internet π
nov 26
is it add some sh!t
or
yeah
no dope guns
it would say that then, cuz I noticed whenever it was flash, it would always say its cutoff date is 2023 and it would stress that it can search the internet
oh alr mb
i thought dope guns would get automodded
thats the best version of doom
cuz mine u cant die in it
it also has 5 enemies of the same type unfortuanently, yours had like 3 or 2, and a slightly bigger map
mine has like 3
with a small map
u could tell it to make different enemies
this doom webgl that pro pro made is sick
thats riftrunner π
5.1 gpt is out
it is??
Yeah
it is! kinda
GPT 5.1 Is released guys
no pelican svg from 5.1 yet im looking at that one group, right now they just bein hyped
It's on the official openai website
Honestly it looks interesting but I wonder what this means for google
Everyone been mega hyping Gemini 3
list one person who cared about 5.1
anyone follow the gem 2 release
what will gem 3 headliner be
heard it might be on creativity?
Introducing Gemini 3.0: our new AI model for the agentic era
Yeah probably towards agents
They tried doing that with Gemini 2.5 tho
Did an update that added improved agentic use
that would be such an epic troll, I don't think Google has it in them, but if they did it, would be worth it for the troll π
gonna be completely honest
I don't mind the gemini 2.6, I mean, look at what riftrunner is doing lol
If they released a 2.6 I would be so happy honestly
Shows they are honestly waiting it out and playing it safe
still no pelican 5.1
Like GTA 6
People that are mad about the push backs want a garbage release
Like let them cook on the game bruh
They know what they doing
Need to remind you they released GTA 5
Lmarena probably going to add it quickly I highly doubt they aren't already working on getting it added
also OpenAI is killing gpt 4.1
google has a deadline for gemini 3 by end of year
