Which AI models can my PC run?
The answer comes down mostly to your graphics card's VRAM. Pick your card below to see which local AI models fit, or scan your PC for an estimate based on measured hardware that also accounts for your RAM, disk space and installed acceleration backends.
NVIDIA graphics cards
AMD graphics cards
Browse by VRAM instead
Know how much memory your card has but not its exact model? Start here.
Using a Mac?
Apple Silicon shares one pool of memory between the CPU and GPU, so it has its own pages by memory size.
How much VRAM do you need to run AI models locally?
Compare a specific build and context rather than parameter count alone. Weights, KV cache, runtime overhead and buffers all use memory. The 8/12/16 GB comparison holds those settings constant, while the calculator lets you change them.
VRAM is not the only constraint. A model that does not fit can still run by offloading part of itself to system RAM, which can change response speed. The PC scan measures all of it — VRAM, RAM, disk and whether CUDA, ROCm or Vulkan is actually available on your machine.