You can use Javascript or Wasm to interface with the AI binding, think of it of as the SDK, but the inference task itself runs natively on top of a ML runtime a

You can use Javascript or Wasm to interface with the AI binding, think of it of as the SDK, but the inference task itself runs natively on top of a ML runtime and the models are loaded into GPUs.