Meta: Llama Guard 4 12B

Llama Guard 4 12B is a Llama 4 Scout-derived multimodal pretrained model, specifically fine-tuned by Meta for robust content safety classification. It excels at moderating both LLM inputs (prompt classification) and LLM responses (response classification). Similar to previous versions, it generates text output indicating whether content is safe or unsafe, and if unsafe, lists the violated content categories. This model is aligned with the standardized MLCommons hazards taxonomy and designed to support multimodal Llama 4 capabilities. This advanced model combines features from prior Llama Guard models, offering content moderation for English and multiple supported languages. It boasts enhanced capabilities for handling mixed text-and-image prompts, including multiple images. Llama Guard 4 is integrated into the Llama Moderations API, extending robust safety classification to both text and images. With a context window of 163K tokens and a max output of 4K tokens, it's a powerful tool for content safety. Access it for FREE on Multi AI, with pricing at $0.18 per 1M input/output tokens.