Downloading and compiling two models in your browser via WebGPU. This can take a while on first load — they are cached afterwards.