The Model block offers three ways: the built-in list of international models (0.5B to 7B params), Hub search by keyword, or a custom org/name.
- Hub search: lists results sorted by downloads; one click selects one.
- HF token: required for gated models; it stays in memory for that run only and is never written to the run's on-disk config.
- Mirror: enable hf-mirror.com if huggingface.co is unreachable from your network.
- The first model download takes a while and lands in the HF cache; later runs load instantly.
Tip: the smallest preset is a light, stable starter, and it is the one this tutorial uses.
