This model differs from the OpenLLaMA models on the Hugging Face Hub, which primarily use the LLaMA architecture.