Model card: provenance wording

Model card: demo section, Swift usage

short-reply-0.6b Core ML package (prefill on ANE, decode GPU), tokenizer, host config

Upload README.md with huggingface_hub

initial commit