Llama Guard 4 12B

Llama Guard 4 12B is Meta's content safety classifier, built to detect and evaluate harmful content in both user prompts and model outputs.

Strengths

Operates at scale with a 163k-token context window and improved safety detection over its predecessor Llama Guard 3 8B.

Best for

Filtering harmful content in moderation pipelines and safety-critical applications where prompt and output evaluation is required.

Limitations

Designed as a safety classifier rather than a general-purpose language model; not intended for generation or reasoning tasks.