[MarcScott]

I've run it on a Pi 5 with 8Gb, and get about a token a second

[theshrike79]

M-series are a LOT faster :)

Even my M1/16GB gets decent speeds. 7+ tokens/second with llama3