How we estimate speed

To generate each token, the machine reads the model's weights from memory once. So the speed ceiling is memory bandwidth divided by the bytes read per token. Real machines reach a fraction of that ceiling.

The numbers we use

Machine typeShare of ceilingSource
mac80%Calibrated on a measured run
x86-igpu60%Assumption, pending measured runs
unified-x8670%Assumption, pending measured runs
unified-arm70%Assumption, pending measured runs

Memory: a model needs its file size plus about 1.5 GB for context. Macs with 36 GB or less give the GPU about two-thirds of RAM by default; bigger Macs about three-quarters. Mixture-of-experts models read only their active experts per token, so they run faster than their size suggests. Model sizes come from ollama.com (4-bit Q4_K_M). Prices are US list prices where we have a source; otherwise we leave the price blank.

Help make it measured

Every submitted run replaces an estimate with a real number. Submit yours.