meta-llama/Llama-Guard-4-12B
Preview
Llama-Guard-4-12B is Meta's specialized natively multimodal content moderation model designed to identify and classify potentially harmful content. Fine-tuned specifically for content safety, this model analyzes both user inputs and AI-generated outputs using categories based on the MLCommons Taxonomy framework. The model delivers efficient, consistent content screening while maintaining transparency in its classification decisions.
Usage note: With respect to any multimodal models included in Llama 4, the rights granted under Section 1(a) of the Llama 4 Community License Agreement are not being granted to you by Meta if you are an individual domiciled in, or a company with a principal place of business in, the European Union.
Key Technical Specifications
Model Architecture
Built upon Meta's Llama 4 Scout architecture, the model is comprised of 12 billion parameters and is specifically fine-tuned for content moderation and safety classification tasks.
Performance Metrics
The model demonstrates strong performance in content moderation tasks:
- High accuracy in identifying harmful content
- Low false positive rate for safe content
- Efficient processing of large-scale content
Technical Details
| FEATURE | VALUE |
|---|---|
| Context Window (Tokens) | 131,072 |
| Max Output Tokens | 128 |
| Max File Size | 25MB |
| Token Generation Speed | - |
| Input Token Price | $0.20 per million tokens |
| Output Token Price | $0.20 per million tokens |
| Tool Use |  |
| JSON Mode | |
| Image Support |  |
Use Cases
Content Moderation
Ensures that online interactions remain safe by filtering harmful content in chatbots, forums, and AI-powered systems.
- Content filtering for online platforms and communities
- Automated screening of user-generated content in corporate channels, forums, social media, and messaging applications
- Proactive detection of harmful content before it reaches users
AI Safety
Helps LLM applications adhere to content safety policies by identifying and flagging inappropriate prompts and responses.
- Pre-deployment screening of AI model outputs to ensure policy compliance
- Real-time analysis of user prompts to prevent harmful interactions
- Safety guardrails for chatbots and generative AI applications
Best Practices
- Safety Thresholds: Configure appropriate safety thresholds based on your application's requirements
- Context Length: Provide sufficient context for accurate content evaluation
- Image inputs: The model has been tested for up to 5 input images - perform additional testing if exceeding this limit.
Get Started with Llama-Guard-4-12B
Unlock the full potential of content moderation with Llama-Guard-4-12B - optimized for exceptional performance on Groq hardware now:
pip install groq1from groq import Groq
2client = Groq()
3completion = client.chat.completions.create(
4 model="meta-llama/Llama-Guard-4-12B",
5 messages=[
6 {
7 "role": "user",
8 "content": "Explain why fast inference is critical for reasoning models"
9 }
10 ]
11)
12print(completion.choices[0].message.content)