Update README.md

README.md

@@ -9,14 +9,21 @@ pinned: true
 # The Llama Family
 *From Meta*
 
-Welcome to the official Hugging Face organization for Llama, Llama Guard, and Code Llama models from Meta! In order to access models here, please visit a repo of one of the three families and accept the license terms and acceptable use policy. Requests are processed hourly.
+Welcome to the official Hugging Face organization for Llama, Llama Guard, and Prompt Guard models from Meta! 
+
+In order to access models here, please visit a repo of one of the three families and accept the license terms and acceptable use policy. Requests are processed hourly.
 
 In this organization, you can find models in both the original Meta format as well as the Hugging Face transformers format. You can find:
 
+Current:
 * **Llama 3.1:** a collection of pretrained and fine-tuned text models with sizes ranging from 8 billion to 405 billion parameters pre-trained on ~15 trillion tokens.
+* **Llama 3.1 Evals:** a collection that provides detailed information on how we derived the reported benchmark metrics for the Llama 3.1 models, including the configurations, prompts and model responses used to generate evaluation results.
+* **Llama Guard 3:** a Llama-3.1-8B pretrained model, aligned to safeguard against the MLCommons standardized hazards taxonomy and designed to support Llama 3.1 capabilities.
+* **Prompt Guard:** a mDeBERTa-v3-base (86M backbone parameters and 192M word embedding parameters) fine-tuned multi-label model that categorizes input strings into 3 categories - benign, injection, and jailbreak. It is suitable to run as a filter prior to each call to an LLM in an application.
+
+History:
 * **Llama 2:** a collection of pretrained and fine-tuned text models ranging in scale from 7 billion to 70 billion parameters.
 * **Code Llama:** a collection of code-specialized versions of Llama 2 in three flavors (base model, Python specialist, and instruct tuned).
 * **Llama Guard:** a 8B Llama 3 safeguard model for classifying LLM inputs and responses.
 
-
 Learn more about the models at https://ai.meta.com/llama/