Llama Guard 3 8B INT8 is a specialist, not a generalist — it focuses exclusively on content safety classification rather than open-ended conversation or generation. It evaluates inputs and outputs for harmful content categories, acting more like a gatekeeper than a collaborator. The INT8 quantization keeps memory footprint lean, making it practical to deploy alongside other models in a pipeline.