Llama 3.1 Nemotron Safety Guard 8B NIM
About the Microservice
The microservice serves a GPU-accelerated LLM model for multilingual content moderation rather than general knowledge. It detects harmful content in user messages and bot responses in Arabic, Chinese, English, French, German, Hindi, Japanese, Spanish, and Thai.
The model was trained using the Nemotron Safety Guard Dataset V3 and cultural adaptation techniques. The dataset has a commercially friendly license and is publicly available for community analysis and development.
Audience and Best Use
Developers can integrate multilingual content safety into their applications using the standard OpenAI API interface and libraries such as LangChain. The Getting Started page shows how to classify safe and unsafe content.
You can also integrate the container with NVIDIA NeMo Guardrails and customize the prompt template. See Integrating with Guardrails for an introduction and links to detailed instructions.
License
Related Information
- CultureGuard: Towards Culturally-Aware Dataset and Guard Model for Multilingual Safety Applications, arXiv:2508.01710.
- Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails, arXiv:2501.09004.
- Nemotron Safety Guard Dataset V3 on Hugging Face Hub.
- The microservice is based on NVIDIA NIM for LLMs.