My $600 Mac Mini Runs a 35B AI Model

Re:My $600 Mac Mini Runs a 35B AI Model by bigyabai on Monday April 13, 2026 @03:56PM (#47757052)

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

Wave "bye bye" to your write cycles.

[ Parent | on HN ]

Re:My $600 Mac Mini Runs a 35B AI Model by RobMurray on Monday April 13, 2026 @05:54PM (#47758349)

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

[ Parent | on HN ]

Re:My $600 Mac Mini Runs a 35B AI Model by bigyabai on Monday April 13, 2026 @06:15PM (#47758631)

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.

[ Parent | on HN ]