Why does RAM matter more than GPU memory?

Local runtimes fall back to CPU and system RAM when GPU memory runs out, so system RAM is the binding requirement. A GPU with enough memory makes generation faster; it does not decide whether a model fits.

What is a quantization build?

A compressed version of the model weights. Smaller quantizations shrink the download and RAM needs at some cost in output quality; each model page lists every build with its numbers.

What happens with models that do not fit?

They still run in idapt through cloud routing, with no hardware limits. Local runs are for the models your machine can hold; they are not billed.

All tools

Free tool

Can your machine run it?

Enter RAM and GPU, get a verdict for every self-hostable model and build: fits with GPU, fits on CPU, or too big.

Last reviewed July 17, 2026

How this is computed

Verdicts use the same compatibility kernel as idapt's paired-computer fit filter: RAM decides whether a build fits (the runtime spills to CPU and system memory when GPU memory is short); a GPU only changes speed. Requirements come from each model's published quantization builds. Your inputs stay in the browser and are remembered locally for next time.