Modular: Evaluate Llama Guard and Granite Guardian with MAX 24.6

AI Impact Summary

This document details a hands-on evaluation of Llama Guard and IBM Granite Guardian models within the MAX 24.6 framework, leveraging Hugging Face and NVIDIA GPUs. The core focus is on comparing these models' ability to classify toxicity in social media content using Surge AI's Toxicity dataset, demonstrating a practical workflow for enterprise AI teams seeking to implement guardrails. The guide highlights the ease of deployment via MAX Serve and the OpenAI-compatible API, offering a streamlined path to model evaluation and selection.