Llama Guard 3 8B is a specialist, not a generalist — it exists solely to classify whether conversational content is safe or unsafe, not to generate responses. It operates like a focused auditor, scanning inputs and outputs against a defined taxonomy of harm categories and returning structured verdicts. Its narrow focus means it excels at its specific task but has no conversational ability outside of safety classification.