Local AI models by graphics card

On Windows, Task Manager shows it under Performance, GPU, as dedicated GPU memory. Two cards with the same memory fit the same models, so there is one page for each size.

On a Mac, memory is shared and the answer is different: models by Mac.