Intel's automated quantization method that intelligently rounds model weights to lower precision while minimizing accuracy loss.