Which models fit on my Mac?
Every popular open model on one chart: pick your Mac memory, quantization and context, and see which ones fit and which is the largest you can run.
How to read it
Each bar is the memory a model needs at the chosen quantization and context: weights, KV cache and a flat 1 GB of buffers. The dashed line is what the GPU can use on a Mac with that much unified memory, about three quarters by default. Dark bars fit under it.
The axis is logarithmic because the models span three orders of magnitude, from about 7 GB to well over a terabyte.
Fitting is not the same as being fast. Generation speed depends on memory bandwidth and on how many parameters are active per token, which is why mixture-of-experts models like Qwen3.6 35B-A3B run much faster than their size suggests.