Model Tag:

Nous Hermes: Overview of the Innovative AI Language Model

Nous Hermes, developed by Nous Research, boasts two primary variants based on different architectures: a 13 billion parameter model built upon the original Llama framework and both 7 billion and 13 billion parameter models based on Llama 2. All these models are general-use and have been trained using identical datasets, ensuring versatility and reliability in applications.

System Requirements

Before running Nous Hermes, it is important to be aware of the memory requirements:

• 7 billion parameter models generally require a minimum of 8GB of RAM.
• 13 billion parameter models typically need at least 16GB of RAM.

If you encounter issues when using higher quantization levels, consider switching to the q4 model or closing other memory-intensive applications.

Exploring Model Variants

Ollama provides various quantized variants of the Nous Hermes models, optimized for local execution:

• Nous Hermes Llama 2: This variant is the original Nous Hermes model based on Llama 2.
• Nous Hermes Llama 1: This variant is based on the original Llama model.

Model Aliases

For your convenience, Nous Hermes models come with various aliases:

Aliases
latest, 7b, 7b-llama2, 7b-llama2-q4_0
13b, 13b-llama2, 13b-llama2-q4_0