#codename-discussion
1 messages Β· Page 6 of 1
Sure man
Hello
<@&1349916362595635286>
GWEEN BEANπππππππππππππ
Hey guys where i can create ai video here?
Only on website
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.
god I'm just getting spammed with tetra models right now and they all... just suck really badly
it's killing basically every matchup by default
we all cooked
Heey everyone
Is this model still available in this arena? Like: gpt5.5 high and 5.4 high.
yes, but not direct chat. If u get luck u can only get this in battle mode. Reason is cost
i tried it (gpt-5.5-[x]high) out in battle, it sucks, when trying to create an engine for jungle chess in Rust
opus is likely better
(the Rust engine of gpt was⦠blargh)
but the GUI was great (it one-shotted a beautiful and functional interface, in Rust)
ok, the Tetra thing is getting genuinely ridiculous. I've had Tetra 5/5 times across two different sessions now
Is there a way to blocklist models by a regex pattern or something? lol
are you getting tetra for the same prompt, or similar prompts? or entirely different from each other?
Completely different prompts across a wide range of themes/tasks
Personality kind of feels like GPT or Gemini, guessing Gemini as itβs coming out next week
recraft-v4.1-utility
"Special Ed classroom yearbook group picture"
And second is from Recraft V4
Notice difference
Another from recraft-v4.1-utility
Hello eveybody, I have a question... I have always limitation now. I only can do 2 or 3 battles per day. Is it normal ?
Are you referring to Video Arena?
whatever openhard-1.0-search-nocot-0506 is supposed to be
name sounds like macrohard which elon musk announced earlier this year
also openhard's whole answer is hallucinated the sources do not match its response at all
God Elon is such a manchild. Macrohard isn't even original
Has anyone ran into this Gemini 3.0 flash model that people are talking about I guess thatβs different than the Gemini 3 flash model thatβs on the leaderboard? I saw people saying itβs actually 3.5 lol I canβt keep up with these rumors
I got luxor-alt
the gemini 3 flash i got in battle mode actually beat opus 4.7 thinking on a task lol. it's really powerful and quick.
yeah, but can it beat Opus-4.6-thinking ?
(which still is better than 4.7)
and can it beat recent GPT?
3 in a row wow
Kiravel is solid
I got Luxor and Luxor alt
are they good
probably a google model
O
You should do more test
does response sound like gemini
or claude
Hmmm!
ask it about tawian
taiwan
I want to know if that's an chinese model
Ok
Alr bro arena just teasing me
damn you got gpt 5.5 xhigh
really rare
Maybot 26
is it better for vibecoding than Opus-4.6-thinking?
(when writing engines for niche boardgames)
luxor has been extremely spammy with emoji in my experience so far
I don't usually do more than 1-2 prompts per battle so I don't have a ton of responses from it
I got it when ironically chatting about something relating to ancient Egypt
Nice
nano banana 3 is better than 2.
I've just witnessed a battle between korin and gpt 5.5 xhigh. Haven't found anything here when searching for korin, does anyone know something?
Excuse me Because I'm still learning how to use it The required number of tokens has been used. Please start a new chat to continue What should I do next? I just need to start a new chat, right? Can't I restore the chat history? Thank you.
Yeah you'll want to start a new chat session. These chat sessions are going to have context limits which you can learn about here: https://help.arena.ai/articles/3975292349-arena-troubleshooting-session-token-limits.
Can't I restore the chat history?
Sorry to say we do not have this feature.
Thank you very much.
when pasting my script in ai chat it gives me this error Trace ID: d1487078-769f
Would you mind sharing this in #1417174113092374689 ?
is an unlogoed muse-spark model secretly gemini 3.2/3.5
new model "kyrin" in code arena
and "kymir" too
Kymir flip flops a lot for me in code suggestions. While the presentation is convincing, I do not trust its output very much.
What's this new model? It's pretty good for images
Got 'flashfennel' for a single image edit test, not bad
Maylynx's visual understanding is solid.
Was hit with Kymir twice in coding, did well the first time but was a flop the second.
maylynx-alpha
new model
Just got this one too in image edit
blue-crab
Nice picture
ellsworth
"u2-preview" in text arena
what do u think about ts model?
and "mizar-beta"
hi
New Claude Model!!
It made an apple clone but with a phone named nova https://019e5086-8c92-7766-8864-e19c1052cf39.arena.site/
cuz it says
it doesnt want to copy
probably a chinese distill
the phone and other models are good
I think it is a claude model
look at the result
another here and it gives heavy claude vibes https://019e507b-ebdd-788f-9b19-831faf618d28.arena.site/
All I typed was "luxury"
This could be a new sonnet model
or opus
anthropic has never done pre-release testing on arena before, it would be weird for them to do it now
oh
but the model keeps saying its claude π
also
QWEN 3.7 MAX IS SO GOATED
some Chinese models can do that
tested that yesterday and i got a loopy mess, at least it got itself out of its loop (qwen 3.5 is infamous for entering infinite loops)
I honestly dont
well that's because you're coding, qwen and hard world knowledge questions don't mix well
u haven't seen mimo v2.5 pro
this server seriously needs a honeypot against this bot plague..
(a digital venus fly trap)
is Mimo better in coding than all other AI models?
frontend maybe
What is king-crab
Prompt" Lost Media video about The Lost Game of Talking Pineapple
not entirely; sometimes it goes into a CoT that last 10 minutes
ok, ty for info
btw, i just encountered dark-matter
seems to be a decent model
unfortunately, it doesn't tell, what it is or who build it
hmmm, talk to it about taiwan and tianan square
maybe its gemini 3.5 pro
unfortunately, the showstopping-bug resurfaced for me (in battle mode)
but adding web-search re-enables access, for some reason
that would be the first time, a google model hid its identity
would it? well, nano banana was a codenamed modelβ¦
i mean, for text models
if you asked any gemini model, they always admit having been created by google
(even for experimental versions, which only were in arena-battle, iirc/afaik)
king-crab (Grok)
Just got king-crab too in t2i, REALLY good for my test
snow-crab
Now thats insane
@woeful junco
Must be another Nano Banana model
@keen bridge well gemini said it doesn't have SynthID
$now?
Snow crab is Reve 2.0?
Kelly
-> Image Model
-> Editor: OpenAI (almost certain, it gave exactly the same result as gpt-image-2 but had some mistakes on text and image)
-> Gpt-image-2-fast-alpha (whatever it's called, it's a cheap version of gpt-image-2)
"caudipteryx" appeared for me on t2i.
kelly
It was mentioned a while back but I got rising-sun on Search battle and yeah woof is it bad
@river kettle what was the prompt, so I could try it if I get a codename model image
Btw, have you heard of mai image 2.5
anyone know about phantom_brush ? getting it a lot in the battle ...
from my first tries its really good !
hello Kjei i am getting good result with that prompt : create infographic stating your model name and creator and model stat overview
I got phantom_brush once and now I'm only getting Flux and wan π€’
runway_gen sucks
"kelly" isn't good at all either
Just ask for a "detailed infographic" and list everything you want to know. mai 2.5 is nothing special.
Sooo, it turns out that this was a new iteration of Qwen 3.7 Max..
Really? Where's the link for the post
Wow it was REALLY good