My $600 Mac Mini Runs a 35B AI Model

thoughts.jock.pl

4 pointsdanebalia3 comments

> 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.