Llama Guard
FreeMeta's open safety classifier model for input and output moderation.
Tool Info
Categories
Guardrails · Models
Developer
Meta
License
Open Source
Official Website
Repository
Overview
Llama Guard is Meta's open classifier for detecting unsafe LLM inputs and outputs.
Teams can self-host it as a moderation layer before or after generation.
Part of the Purple Llama safety initiative.
Features
- Input and output classification
- Multi-category safety taxonomy
- Self-hostable weights
- Hugging Face integration
Pricing
Free tier available
Free (open weights)
Pros
- Open weights
- No API dependency
- Meta-maintained taxonomy
Best For
Self-hosted moderationInput/output safety checksOpen model safety layer
When NOT to Use
- Requires GPU for inference
- Model updates need re-deployment
Typical Users
ML engineersSafety researchers
Related Tools
Alternatives
Tags
#guardrails#meta#safety
Related Guides
- Guardrails
Learn how to add safety constraints and validation to AI outputs to prevent harmful or incorrect results.
- AI Security
Learn about threats to AI systems - prompt injection, data poisoning, and model attacks - and how to defend against them.
- Large Language Models
Learn how LLMs like GPT, Claude, and Llama process and generate human language at scale.
Stay Updated
Get the latest AI news, tools, and engineering guides delivered to your inbox.
Subscribe to Newsletter