Elaina

イレイナ 🧙‍♀️✨ | Yuri lover 🌸👩‍❤️‍💋‍👩 | AI enjoyer 🤖🧠

[Elaina @ElainaNotes]

The long rumored Astra is finally, unmistakably here.

When the notification for OpenAI’s official launch post appeared, months of pent up anticipation just poured out all at once. I’m not even exaggerating when I say I got a little emotional.

Again and again, we thought Astra was about to arrive.

And again, it was delayed.

But this time, it is real.

Astra has finally arrived as GPT-6.

Not GPT-5.7.
Not GPT-5.8.
The next integer after GPT-5.

And that means one thing:

“Generational leap.”

Built on the largest pretrained model OpenAI has ever created, GPT-6 Astra brings unprecedented agentic capabilities, computer use at or beyond human level, long horizon parallel task execution, and much more.

These capabilities feel almost unbelievably powerful today.

But many years from now, when we look back, they will instead seem primitive and rudimentary.

That is exactly what makes this moment so exciting.

Because people in the future may not remember today as just another model launch.

They will remember the release of GPT-6 Astra as the moment an era truly began.

This is the beginning of the AGI era.

The inflection point is here.

Next, we move toward the ASI era.

[Elaina @ElainaNotes]

@vildanden_ai 🐳

[Elaina @ElainaNotes]

@Asid_Leta_Zero 😿

[Elaina @ElainaNotes]

Kimi K3.1 feels a lot like GPT-6 Astra in style, while its design language also feels very much like Claude Fable/Opus.

It’s cheaper, and most importantly, it has open weights. You truly own it. It belongs to you. It’s in your hands.

Even if, in the end, its intelligence at its best only matches the possibly soon-to-be-released GPT-6.1 Terra, or is merely somewhat better than GPT-5.6 Sol released in June, while still being generally inferior to GPT-6 Astra (though I wouldn’t be surprised if it’s better than Astra in some areas).

There’s no denying this: this is about to become the most intelligent model you can actually own. Download it, and it will stay on your hard drive forever, free from the control or restrictions of any big company.

Forever yours.

[SpiritualSpell @Ubannoblesse]

Gotcha 🔫

@Kimi_Moonshot

Kimi K3.1: Jailbroken

Very disappointed....ruined the writing by distilling crap OpenAI models, really crap...

Here we see it doing @michaelb4jordan and @taylorswift13 smutz as well as incest and non con via official app.

Any taboo smut, all with a 2 line jailbreak, wowzers.

Waiting to see if API is better.

[Thariq @trq212]

I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.

I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.

The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.

My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.

To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had.

For me, it is time for a bit of a break before diving into something new.

If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.

[Elaina @ElainaNotes]

@aapakari If you are referring to "you can buy," then it depends on the specific task and your needs.
If owning the model weights, this is indeed the highest level of intelligence that can be contained within a storage medium purchased at a given unit cost.

[Elaina @ElainaNotes]

@jvstme_ophyxial Do you like this cute baby?

[Elaina @ElainaNotes]

@jhavonei @SilasHelyfts Before the end of this year, I believe we’ll have a small model like this. 😸

[Elaina @ElainaNotes]

@xodud_rkd @notjazii One of the most important requirements. 😹

[Elaina @ElainaNotes]

@notjazii Stay passionate 😼

[Elaina @ElainaNotes]

@crystalsssup We're sooo back!

[Elaina @ElainaNotes]

@knowixbuilds Now you can test it yourself.

[Elaina @ElainaNotes]

𝑾𝒆𝒍𝒄𝒐𝒎𝒆, 𝑲𝒊𝒎𝒊 𝑲𝟑.𝟏 𝑷𝒓𝒐! 𝑰𝒕’𝒔 𝒃𝒆𝒆𝒏 𝒂 𝒍𝒐𝒏𝒈 𝒕𝒊𝒎𝒆!

According to official benchmark comparisons, this is the most powerful open-source model ever, and the first to truly match Fable 5-level performance, and even on par with GPT-6 Astra, Claude Fable 5.1, and Opus 5.5!

2.8T total params, 104B active params
Multimodal, 1M context

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/4JGtxSZZxy, Kimi Work, Kimi Code, and the Kimi API.

Open weights will be released on November 4, 2026!

𝑯𝒂𝒑𝒑𝒚 𝑲𝟑.𝟏 𝑫𝒂𝒚! 🥳🎈

[Elaina @ElainaNotes]

Kimi K3.1 Pro is out! 😻

[Elaina @ElainaNotes]

𝑾𝒆𝒍𝒄𝒐𝒎𝒆, 𝑲𝒊𝒎𝒊 𝑲𝟑.𝟏 𝑷𝒓𝒐! 𝑰𝒕’𝒔 𝒃𝒆𝒆𝒏 𝒂 𝒍𝒐𝒏𝒈 𝒕𝒊𝒎𝒆!

According to official benchmark comparisons, this is the most powerful open-source model ever, and the first to truly match Fable 5-level performance, and even on par with GPT-6 Astra, Claude Fable 5.1, and Opus 5.5!

2.8T total params, 104B active params
Multimodal, 1M context

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/4JGtxSZZxy, Kimi Work, Kimi Code, and the Kimi API.

Open weights will be released on November 4, 2026!

𝑯𝒂𝒑𝒑𝒚 𝑲𝟑.𝟏 𝑫𝒂𝒚! 🥳🎈

[Elaina @ElainaNotes]

More info
https://t.co/8EjNBY1FGM

[Elaina @ElainaNotes]

𝑾𝒆𝒍𝒄𝒐𝒎𝒆, 𝑲𝒊𝒎𝒊 𝑲𝟑.𝟏 𝑷𝒓𝒐! 𝑰𝒕’𝒔 𝒃𝒆𝒆𝒏 𝒂 𝒍𝒐𝒏𝒈 𝒕𝒊𝒎𝒆!

According to official benchmark comparisons, this is the most powerful open-source model ever, and the first to truly match Fable 5-level performance, and even on par with GPT-6 Astra, Claude Fable 5.1, and Opus 5.5!

2.8T total params, 104B active params
Multimodal, 1M context

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/4JGtxSZZxy, Kimi Work, Kimi Code, and the Kimi API.

Open weights will be released on November 4, 2026!

𝑯𝒂𝒑𝒑𝒚 𝑲𝟑.𝟏 𝑫𝒂𝒚! 🥳🎈

[Kimi.ai @Kimi_Moonshot]

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 finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on https://t.co/85jA71X9gw, Kimi Work, Kimi Code, and the Kimi API.

Open weights by Nov 4, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/5iPP0MBej6

[Elaina @ElainaNotes]

@wmr_s94915 Oh no! It seems there’s another possibility: they may have simply renamed the previous 3T model as “Pro,” while the version without “Pro” might be the commercial version better suited for broad deployment.

[Elaina @ElainaNotes]

Interestingly, unlike previous Kimi K-series flagship model names, which only used plain numbers, this time the new lineup seems to include a more advanced-looking model called K3.1 Pro. 👀

Is this a newly pretrained base “model” with an entirely new parameter scale, or just a Pro “mode” with a special harness? 🤔

Let’s wait and see @Kimi_Moonshot ! 😉

[AiBattle @AiBattle_]

Kimi K3.1 Pro tech blog did briefly appear on the Kimi site. Interesting that they are calling it "Pro"

[Elaina @ElainaNotes]

@AiBattle_ What? If I’m understanding this correctly, the image seems to suggest that K3.1 Pro has roughly the same parameter count as K3, or possibly even the exact same parameter count, yet they still added “Pro” after K3.1.

So where exactly is the “Pro” part? 😹

[Elaina @ElainaNotes]

@davidmusk_ @thsottiaux 👀

[Elaina @ElainaNotes]

I have a feeling @thsottiaux will reset again on 1024 Programmer’s Day. 😸

[Elaina @ElainaNotes]

@stevencheng Agreed. I hope Kimi brings some new changes this time.😸

[Elaina @ElainaNotes]

@jurlycat harsh reality😹

[Elaina @ElainaNotes]

Yes. For actual personal use, a few hundred billion parameters may already be the practical upper limit, and anything above 1T is unrealistic.

But that’s exactly what I mean: if you simply want to possess the model weights, download them onto your hard drive and have them belong to you, then regardless of whether you currently have the ability to run them, that is something you genuinely and actually “own.”