typeof

Open source AI model on par with o1 👀

https://github.com/deepseek-ai/DeepSeek-R1

Staff Eng at Iron Fish 🐟 • Building BlockSocks 🧦 • Denver, CO 📍

Are these the chinese models?

Man, local first, open source, competitive models, (happily) landed sooner that i was expecting.

🐵 ← 🙈🙉🙊 → /phil →  phil.cockfield.net, db.team →  
ƒ(🌳) →  🧫 /sys  ←  @cell <T>

Yup, it’s a Chinese model. Interesting how close behind the US they are, and how they’re choosing open source. I’m def here for it

Super interesting - the "thinking" (reasoning steps) are coughed up before the response.  Damn this is cool.

(this is the 14B model running on my MBA, 24GB memory hardward)

I haven’t read the paper, but it’s linked in the repo if you wanna learn about how it works