# PolicyLM-1.7B

- [Download PolicyLM-1.7B on HuggingFace](https://huggingface.co/musubilabs/policylm-1.7b)
- [Musubi Announcement](https://www.musubilabs.ai/blog/introducing-policylm-1-7b)
- [Model Card](https://huggingface.co/musubilabs/policylm-1.7b)
- [Hosted version on Baseten](https://www.baseten.co/library/policylm-1b/)

PolicyLM-1.7B is an open-weights (Apache-2.0) moderation classifier from Musubi Labs. Give it a message and a policy written as short plain-language categories, and it returns a 0–1 score for every category in one pass, with no text generated. It is built for places that need a decision right away, such as live chat, DMs, usernames and comments, and runs on a single 24 GB GPU, a CPU or Apple silicon.

### Helpful Resources
- [Quickstart and inference helper](https://huggingface.co/musubilabs/policylm-1.7b) (the `inference/` folder in the model repo)
- [Questions, bug reports and results on your own data](https://huggingface.co/musubilabs/policylm-1.7b/discussions)