gpt-oss-20b on a Mac mini (16GB): yes, but get 24GB
llama.cpp's own guide runs it on a 16GB Mac with one flag. But on a Mac the CPU and GPU share one pool of memory, so the whole model still has to…
Aliteq
Lena Fischer · AI & Local Compute Editor
The short answer
Yes, but 16GB is tight and 24GB is comfortable. llama.cpp's official gpt-oss guide says gpt-oss-20b "can run efficiently on Macs with at least 16GB RAM", but notes that "Macs don't allow to utilize…
16GB Mac: works with --n-cpu-moe 12 -c 32768 (llama.cpp's own Mac command). Close other apps.
24GB Mac: fits fully on the GPU at 32K, with ~8 GiB left for macOS
Memory bandwidth: M6 170 GB/s vs M4 120 GB/s (Apple). Generation speed follows it.
Buying for this? Get the 24GB M6. Memory can't be upgraded later.
Aliteq
Read the full story
gpt-oss-20b on a Mac mini (16GB): yes, but get 24GB