Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

DeepSeek-R1 on Mac Studio 192GB 🪄

ggerganov's profile picture

Georgi Gerganov

59,308 subscribers

259,545 Aufrufe • vor 1 Jahr •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

12 Kommentare

Profilbild von Georgi Gerganov
Georgi Gerganovvor 1 Jahr

The model is quantized by @UnslothAI

Profilbild von Solar Heavy
Solar Heavyvor 1 Jahr

- Moonlight

Profilbild von Unsloth AI
Unsloth AIvor 1 Jahr

Looks soo good! 🎇🐋

Profilbild von Daniel Nakov
Daniel Nakovvor 1 Jahr

come on apple, new studio

Profilbild von Georgi Gerganov
Georgi Gerganovvor 1 Jahr

M4 studio with 256GB would be the final boss

Profilbild von unfazed.
unfazed.vor 1 Jahr

@JudeSkrr il a encore drop

Profilbild von Vasilii Kharitonov 🌍
Vasilii Kharitonov 🌍vor 1 Jahr

Tried R1 32B 8bit on M4 Pro with 48 GB memory (ggml-org/DeepSeek-R1-Distill-Qwen-32B-Q8_0-GGUF), seems to run great 👍 Memory usage is 43.3/48.0GB, eval time is about 7 tokens per second.

Profilbild von GP
GPvor 1 Jahr

I need your cli setup 😏

Profilbild von Woody Lu
Woody Luvor 1 Jahr

Could you share the noise level (in DB) of the mac when running it locally?

Profilbild von Georgi Gerganov
Georgi Gerganovvor 1 Jahr

Not sure how to measure it. It only makes noise for very long contexts, otherwise it's completely silent. And when it makes noise, it's not really loud IMO.

Profilbild von Alexander Kvamme
Alexander Kvammevor 1 Jahr

What profiling tool are you using? Much cleaner than htop

Profilbild von Georgi Gerganov
Georgi Gerganovvor 1 Jahr

asitop

Ähnliche Videos

we just added an early preview of DeepSeek-R1 to the fullmoon TestFlight you can use it on iPhone, iPad, Mac, and Vision
0:20

we just added an early preview of DeepSeek-R1 to the fullmoon TestFlight you can use it on iPhone, iPad, Mac, and Vision

Jordan Singer

52,763 Aufrufe • vor 1 Jahr

DeepSeek R1 is now active on FluxAI.
0:49

DeepSeek R1 is now active on FluxAI.

Daniel Keller

23,528 Aufrufe • vor 1 Jahr

deepseek-r1-distill-llama-70b on groq
0:50

deepseek-r1-distill-llama-70b on groq

Developers Digest

143,277 Aufrufe • vor 1 Jahr

DeepSeek R1 + FactSet = Financial Research on Steroids
0:34

DeepSeek R1 + FactSet = Financial Research on Steroids

Aravind Srinivas

582,192 Aufrufe • vor 1 Jahr

OpenAI's nightmare: DeepSeek R1 on a Raspberry Pi
4:17

OpenAI's nightmare: DeepSeek R1 on a Raspberry Pi

Jeff Geerling

23,725 Aufrufe • vor 1 Jahr

o3-mini-high (left) vs. deepseek-r1 (right) results from the first try deepseek-r1 is cracked... wtf!
0:30

o3-mini-high (left) vs. deepseek-r1 (right) results from the first try deepseek-r1 is cracked... wtf!

elvis

719,763 Aufrufe • vor 1 Jahr

Watch DeepSeek-R1 (8B) "think" in realtime on a regular computer.
1:43

Watch DeepSeek-R1 (8B) "think" in realtime on a regular computer.

Vincent Kennedy

43,351 Aufrufe • vor 1 Jahr

🚨 DeepSeek just dropped DeepSeek R1-0528 and it's INSANE for coding Built an entire game with Nebius Studio in three turn. The improvement from previous versions is wild. - API access available - Safe to use If this is R1... DeepSeek R2 is going to be absolutely nuts. How to access 👇 1/3
0:44

🚨 DeepSeek just dropped DeepSeek R1-0528 and it's INSANE for coding Built an entire game with Nebius Studio in three turn. The improvement from previous versions is wild. - API access available - Safe to use If this is R1... DeepSeek R2 is going to be absolutely nuts. How to access 👇 1/3

AshutoshShrivastava

61,058 Aufrufe • vor 1 Jahr

Grok 4's coding skills are worse than DeepSeek-R1 (the old version, not even the newer DeepSeek-R1-0528)! 😅 #Grok4
0:54

Grok 4's coding skills are worse than DeepSeek-R1 (the old version, not even the newer DeepSeek-R1-0528)! 😅 #Grok4

ShanghaiPanda （ Account deactivation）

118,909 Aufrufe • vor 11 Monaten

DeepSeek R1 + Perplexity is INSANE 🤯
13:04

DeepSeek R1 + Perplexity is INSANE 🤯

Julian Goldie SEO

168,119 Aufrufe • vor 1 Jahr

DeepSeek-R1 + Windsurf is INSANE! 🤯
12:13

DeepSeek-R1 + Windsurf is INSANE! 🤯

Julian Goldie SEO

46,671 Aufrufe • vor 1 Jahr

Sarvam-M now has 8.6 lakh downloads on Hugging Face, beating DeepSeek-R1!
1:58

Sarvam-M now has 8.6 lakh downloads on Hugging Face, beating DeepSeek-R1!

Caleb

55,857 Aufrufe • vor 1 Jahr

Project Veritas: Make DeepSeek R1 truth seeking
0:42

Project Veritas: Make DeepSeek R1 truth seeking

Aravind Srinivas

96,555 Aufrufe • vor 1 Jahr

Build anything with DeepSeek-R1, here's how...
18:30

Build anything with DeepSeek-R1, here's how...

Julian Goldie SEO

3,469,539 Aufrufe • vor 1 Jahr

DeepSeek-R1 + RooCode is INSANE (FREE!) 🤯
13:19

DeepSeek-R1 + RooCode is INSANE (FREE!) 🤯

Julian Goldie SEO

48,802 Aufrufe • vor 1 Jahr

Automate Anything with DeepSeek-R1, here’s how
19:40

Automate Anything with DeepSeek-R1, here’s how

Julian Goldie SEO

880,544 Aufrufe • vor 1 Jahr

Latest Fireship video featured my run of DeepSeek R1 on M4 Mac Minis. Apple Silicon dominates in memory/bw unit economics, ideal for huge MoE models like R1 at batch_size=1 (the real-world use-case). SOTA AI may come from China, but it will run on American hardware.
0:21

Latest Fireship video featured my run of DeepSeek R1 on M4 Mac Minis. Apple Silicon dominates in memory/bw unit economics, ideal for huge MoE models like R1 at batch_size=1 (the real-world use-case). SOTA AI may come from China, but it will run on American hardware.

Alex Cheema - e/acc

79,775 Aufrufe • vor 1 Jahr

⚡️New "Thinking" UI with DeepSeek R1 Distilled models, running completely locally on iPhone
0:58

⚡️New "Thinking" UI with DeepSeek R1 Distilled models, running completely locally on iPhone

Private LLM

188,633 Aufrufe • vor 1 Jahr

DeepSeek-R1 is blowing up right now, but we're not surprised. And not just because we’ve been working closely with DeepSeek to bring these models to production. We've been betting on powerful, open-source models like DeepSeek-R1 from day one. 1/n 🧵
0:59

DeepSeek-R1 is blowing up right now, but we're not surprised. And not just because we’ve been working closely with DeepSeek to bring these models to production. We've been betting on powerful, open-source models like DeepSeek-R1 from day one. 1/n 🧵

Baseten

59,327 Aufrufe • vor 1 Jahr