Comparisons inside Meta's Purple Llama safeguards: Llama Guard, Prompt Guard and LlamaFirewall, and how their versions and licenses differ.
Content safety against prompt attack detection: Llama Guard 4's 12B multimodal classifier for hazard categories in prompts and responses against Prompt Guard 2's 22M and 86M classifiers for injections and jailbreaks, with sizes, context limits and where each sits in a pipeline.
Moving from Llama Guard 3-8B or 3-11B-vision to the single 12B multimodal Llama Guard 4: multi-image support, hazard taxonomy, hardware needs and the change from the Llama 3.2 to the Llama 4 Community License.
Tell us what you're working on, whether it's a deployment, an audit, a security test or a cyber range. You'll speak with an engineer who can help you scope it.