Model · Judgment model

Llama-Guard-4-12B

A gated downloadable 12B natively multimodal safety classifier pruned from Llama 4 Scout that emits safe/unsafe verdicts as generated text, governed by the Llama 4 Community License.

Developer

Meta

Distribution

Downloadable weights

Artifact licensing

Restricted artifact terms

License scope

Reviewed release artifacts and mandatory terms; not an assessment of training code or training data openness.

Licenses

Llama 4 Community License · Llama 4 Acceptable Use Policy

Boundary

This record represents the meta-llama/Llama-Guard-4-12B release at Hugging Face commit 87acb4b94e930c3d679e6e7ee9d57e2feab9ea71. The earlier Llama Guard 3 releases, the Llama 4 Scout base it was pruned from, quantizations, fine-tunes, the Llama Moderations API, and independent serving providers are separate model or service boundaries. The model acts as a classifier but generates text verdicts, so it is recorded as a multimodal language model with the safety-classifier role carried in prose.

Open in the directory → Open official model page ↗

Editorial ratings appear in the directory beside the profile they belong to and are never compared across collections.