Meta releasing Llama Guard 4 on Hugging Face Hub

AI Impact Summary

Meta has released Llama Guard 4, a 12B dense multimodal safety model, alongside two new Llama Prompt Guard 2 models, designed to mitigate unsafe content generation in large language models. This release includes open model checkpoints available on the Hugging Face Hub, along with an interactive notebook for easy experimentation. The Guard models evaluate both text and image inputs, offering flexible moderation pipelines for filtering prompts and generated responses, and supports multilingual moderation.