Local AI runs on modest PCs - no RTX needed; efficient small models work on CPU and iGPU. Sub-1B models feel instant for ...