Models / meta / Llama Guard 4 12B

Llama Guard 4 12B

JSON →
metallm

A 12-billion-parameter safety classifier from Meta designed to detect harmful content and enforce content policies in LLM inputs and outputs.

textjson-modestreaming
Specs
Context & limits
8K
context window
8K
max output
Lifecycle
Dates
released2025-04-01
knowledge cutoff2024-06-01
Resources