Android

Higher Dimension Being

[Tencent Hy @TencentHunyuan]

Hy3 has just claimed #1 on the OpenRouter LLM leaderboard! 🥇

Huge thanks to our partners for their hardcore support — this milestone wouldn’t be possible without you.

Fellow geeks and devs, jump in, test it out, and drop your feedback & reviews. 📷#Hy3 #OpenRouter #AI #LLM

[Tony Simons @tonysimons_]

WTF did I just find?!

Colibrì runs GLM-5.2’s 744B-parameter model locally on a consumer PC with roughly 25 GB of RAM.

🚫 No cloud API.
🚫 No server rack.
🚫 No $30,000 GPU.

The engine is written in pure C. It keeps the dense portion in RAM, stores roughly 370 GB of experts on your NVMe, then streams the experts it needs for each token.

The catch: it’s slow as hell.

But that almost doesn’t matter.

Someone just proved you can make a frontier-sized MoE model answer questions on hardware sitting under your desk.

I have a laptop 4090, WSL2, and several terrible ideas.

Who wants to see me get this monster running?

https://t.co/74Il5ue094

[Android @Androthon]

@EHuanglu or all of the copy cat AI leaders copying from each other and their looking to their left looking at Elon... and Elon and Andrei Karpathy looking at each other...

[X Freeze @XFreeze]

Anthropic is now extending Fable 5 access on their subscription plans because they got really close to losing customers to the current competition, despite having the most powerful AI model

They’re running dry on inference, basically, they don’t have enough cheap compute to let everyone hammer Fable 5 inside subscriptions, so they're pushing it to higher‑priced API usage instead

They have a powerful model, but not a very efficient one, and many competitors currently want to snatch Anthropic’s paying customers at all cost

So the winner in this race will be whoever can provide cheaper inference. That is what is going to count sooner or later. Everybody will eventually catch up, but inference is the limiting factor

That’s where SpaceXAI is insanely bullish. It already has major data centers on Earth and has built the fastest and most efficient data center on the planet

And soon, Elon is going to build a data center in space where, once it is deployed, there will be no energy bills or cooling bills, just free sun energy for as long as the data center can run

That is the future we’re heading toward right now

[Elon Musk @elonmusk]

Bamf

[Elon Musk @elonmusk]

As usual, legacy media is misrepresenting the situation.

I just asked Tesla & SpaceX to try out Grok 4.5 to see if it solves their task, not use it no matter what!

They should continue to use other AI models if those models outperform Grok.

[Hasan Toor @hasantoxr]

A 744 billion parameter AI model just ran on a machine with 25GB of RAM.

No graphics card.

The tool is called colibri.

GLM-5.2 is a mixture-of-experts model. It contains 744 billion parameters but only about 40 billion wake up for each token.

colibri keeps the 9.9GB core in memory.

The remaining 370GB sits on an SSD. As the model writes, colibri pulls in only the tiny group of experts needed for that token.

→ Pure C
→ Zero runtime dependencies
→ 25GB RAM
→ Zero graphics cards
→ The full model stays on your machine

The catch is brutal.

A cold token can require 11GB of disk reads, so output crawls at roughly one token every 10 to 20 seconds. You will not replace ChatGPT with this tomorrow.

But colibri cracks the assumption that the whole model must fit in memory before it can run.

100% open source.

https://t.co/dSm86eRWCj

[Android @Androthon]

@NiohBerg @grok what is it?

[DogeDesigner @cb_doge]

When NVIDIA CEO Jensen Huang interviewed @elonmusk in 2015, over a decade ago.

[Elon Musk @elonmusk]

@sama We start flying them next year. Maybe you can come see them if your parole officer approves.

After stealing an open source AI charity, you then stole all of Apple’s phone technology! Wow.

What do you plan for an encore? That’s tough to beat.

[Android @Androthon]

@XFreeze They also promised to release Grok 3 to the open source community.

[All day Astronomy @forallcurious]

🚨: Scientists achieve the impossible by creating matter from empty space

[NIK @ns123abc]

“GPT‑5.6 by our community”

How it started: How it's going: