Live data from Hacker News

My $600 Mac Mini Runs a 35B AI Model

thoughts.jock.pl

Post reply on HN

Print view View original

1–4 of 4 posts

My $600 Mac Mini Runs a 35B AI Model

#1

Post by danebalia » Mon, Apr 13, 2026, 7:52 PM UTC

My $600 Mac Mini Runs a 35B AI Model
thoughts.jock.pl

Re: My $600 Mac Mini Runs a 35B AI Model

#2

Post by bigyabai » Mon, Apr 13, 2026, 7:56 PM UTC

> The 35B Trick (Your SSD Is the New GPU Memory)

Wave "bye bye" to your write cycles.

Re: My $600 Mac Mini Runs a 35B AI Model

#3

Post by RobMurray » Mon, Apr 13, 2026, 9:54 PM UTC

bigyabai wrote: post #2

> The 35B Trick (Your SSD Is the New GPU Memory) Wave "bye bye" to your write cycles.

why? it's mostly reads. the weights are static.

Re: My $600 Mac Mini Runs a 35B AI Model

#4

Post by bigyabai » Mon, Apr 13, 2026, 10:15 PM UTC

RobMurray wrote: post #3

bigyabai wrote: post #2

> The 35B Trick (Your SSD Is the New GPU Memory) Wave "bye bye" to your write cycles.

why? it's mostly reads. the weights are static.

llama-cpp's process is, but macOS itself will swap hard when 10-14gb of memory is paged for LLM inference. Dense models especially would thrash zram.

Post reply on HN