someone tested the unreleased gemini 4 carbon checkpoint in arena
> asked for a random 30 sec horror scene in a jungle
> turn your sound on, the audio is the scary part
argon was already good at audio, this one looks wild. who's testing it next?
kimi k3.1 pro is now #4 on artificial analysis
> above most of the closed frontier
> weights reportedly drop nov 4
who's actually still paying opus prices?
three stealth models walked into arena to fight opus 5.5
> carbon (rumoured gemini 4)
> kibitz (rumoured kimi k3.1)
> kalmia (rumoured qwen 4)
which one came closest?
kimi k3.1 pro is officially out
> 2.8t params, 104b active
> 1m context
> builds full games in unreal, unity and godot
is this the best open model yet?
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at
4 models, one prompt: "build a cozy 3d bedroom"
> opus 5.5: ~$9.00
> sonnet 5.5: $3.01
> sol 6.1: $1.23
> luna 6.1: $0.05
180x price gap
which room would you sleep in?
someone made google antigravity text them first
> hooked it to telegram
> built a proactive skill on top
> now it pings them about daily tasks
would you let an agent message you first?
someone built a virtual macbook with claude
> open the case
> pull the battery
> remove the fans
> take out the logic board
every part comes out in order
how far can this go?
kalmia (rumoured qwen 4) vs gpt 6.1 sol ultra
same benchmark
> early testers say kalmia is already on sol ultra's level
> still not quite opus 5.5 though
and it hasn't even launched yet
who did it better?
🔴 Qwen 4 Kalmia is looking kinda scary 👀
I tested it on my Magnetar benchmark, and ngl, the results caught me off guard.
It feels pretty close to GPT 6.1 Sol Ultra, though I’d still put Opus 5.5 a step ahead for now.
The wild part? It could be priced around Haiku 5.5.
A little slower than some models I’ve tested, but if the quality holds up at that price, we might have a serious new contender on our hands.
No official benchmarks yet, so I’m keeping an open mind.
But now I’m curious 👀
**If Qwen 4 Kalmia really delivers GPT-level quality at Haiku-level pricing, would you switch to it, or are you still sticking with Claude and GPT?**
I wanna hear your takes. 🔥
kalmia (rumoured qwen 4) just built a drivable 3d car
> took ~20 mins
still a bit buggy but this is a stealth model on arena
which one should it build next?
one thing about email marketing is that creating the actual email is only one part of the job.
there's the copy to write, the design to get right, the mobile version to check, and the delivery to worry about.
think about doing all of that across multiple campaigns.
you can easily spend more time managing the process than working on the ideas themselves.
this is more of a reason you need to find @Migma_AI worthy of its functions.
I mean:
⤪being able to describe what you want ⤪generate an on-brand email draft
⤪make edits in plain language
⤪prepare everything for delivery without constantly switching between tools makes the process a lot more manageable.
btw what do y'all think of the video 😉?
2.7M Followers 47 FollowingThe official handle for NVIDIA. Blog: https://t.co/JAn5eKOTBT Support: https://t.co/6ln5FVnA2o All our social media: https://t.co/Uc56dL57Dh
193K Followers 8 FollowingMakers of Devin, the first AI software engineer. We are an applied AI lab building end-to-end software agents. Join us: https://t.co/4Ss9hvpjRG
5.5M Followers 4 FollowingOpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. We’re hiring: https://t.co/dJGr6LgzPA
191K Followers 3K Following👩🏻💻 Everything in Chief, The New Things
📚 "I AM NOT A ROBOT" (@harpercollins)
📺 Chief tech analyst @nbcnews
🏆 Emmy winner + mom
1K Followers 542 Followingthis dude writes and designs too || my content secrets live in the library || cw tutor @thePW3acad || my highlights says it all