The file is read straight from your disk into this tab. Scanning runs on your GPU (WebGPU if available) and the models are cached after the first run.