meta-llama/Meta-Llama-Guard-2-8B

Meta Llama Guard 2 is an 8 billion parameter LLM safeguard model developed by Meta, based on the Llama 3 architecture. It is specifically designed for classifying content in both LLM inputs and responses, identifying and categorizing unsafe content across 11 harm categories aligned with the MLCommons taxonomy. This model acts as a content moderation tool, generating text outputs that indicate safety status and violated content categories, and significantly improves over its predecessor, Llama Guard, in F1 score and AUPRC.