PICK YOUR COMPUTER
Each page: the top-ranked free model that fits, for each job, one command, and one fix.
MAC
16 GB16 GB MAC 10.7 GB for models (estimate)
24 GB24 GB MAC 16 GB for models (estimate)
32 GB32 GB MAC 21.3 GB for models
36 GB36 GB MAC 24 GB for models (estimate)
48 GB48 GB MAC 32 GB for models (estimate)
64 GB64 GB MAC 48 GB for models
96 GB96 GB MAC 72 GB for models (estimate)
128 GB128 GB MAC 96 GB for models (estimate)
192 GB192 GB MAC 144 GB for models (estimate)
256 GB256 GB MAC 192 GB for models (estimate)
512 GB512 GB MAC 384 GB for models (estimate)
NVIDIA GRAPHICS CARD
8 GB8 GB NVIDIA CARD 8 GB for models (the card's memory)
12 GB12 GB NVIDIA CARD 12 GB for models (the card's memory)
16 GB16 GB NVIDIA CARD 16 GB for models (the card's memory)
24 GB24 GB NVIDIA CARD 24 GB for models (the card's memory)
32 GB32 GB NVIDIA CARD 32 GB for models (the card's memory)
TWO 24 GBTWO 24 GB NVIDIA CARDS 48 GB for models (estimate: both cards added up)
AMD RYZEN AI MAX PC
64 GBAMD RYZEN AI MAX 64 GB 48 GB for models, after you set it
128 GBAMD RYZEN AI MAX 128 GB 96 GB for models, after you set it
NVIDIA DGX SPARK
ONEONE DGX SPARK Picks from recipes tested on one Spark
2XTWO DGX SPARKS Picks from recipes tested on 2 Sparks
3XTHREE DGX SPARKS Picks from recipes tested on 3 Sparks
4XFOUR DGX SPARKS Picks from recipes tested on 4 Sparks
NOT LISTEDAMD Ryzen AI Max PRO 400 (Gorgon Halo). Not officially supported by Ollama or LM Studio yet. The Ollama list names the older Max+ 395, 390 and 385 only. source
Not sure? Mac: Apple menu > About This Mac. Windows: Task Manager > Performance > GPU, "Dedicated GPU memory".