Model purpose
llama-guard-4-12b is described as a multi-modal model that classifies safety for input prompts as well as output responses. It sits in the content safety category rather than being a general-purpose chat or reasoning model. Readers searching for a llama guard content safety model will find it listed under Meta in the NVIDIA model catalogue.
- Classifies safety of input prompts
- Classifies safety of output responses
- Multi-modal input handling
Classification tags
The catalogue entry applies a set of tags that indicate how the model is intended to be used. These are content safety, guardrail and content moderator. The tags distinguish it from the other Meta models listed alongside it, which carry tags such as image-to-text, reasoning, chat, text-to-text and large language models.
- Content Safety
- Guardrail
- Content Moderator
Availability and access
The model is published on build.nvidia.com within the Meta section of the NIM API catalogue. It is offered with a free endpoint. Unlike several neighbouring Meta entries, the listing marks it with the free endpoint label only.
- Listed under Meta on build.nvidia.com
- Free Endpoint availability
- One of four Meta models shown on the page
Sources
Last verified 02 Oct 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.