We're excited to announce the addition of a new model to Private LLM, the Hermes 2 Pro Llama-3 8B. This model, based on Meta's Llama 3 8B, is now available for both iOS and macOS versions of our app.

For iOS devices with 8GB or more RAM, such as the iPhone 15 Pro or iPad Pro and newer, we offer a 3-bit quantized version of the Hermes 2 Pro Llama-3 8B model. This ensures optimal performance on mobile devices while maintaining the model's capabilities. On the macOS side, we provide a 4-bit quantized version of the model, which is compatible with all Intel and Apple Silicon Macs.

Screenshot of Hermes 2 Pro Llama 3 8B running on iPhone — Hermes 2 Pro Llama 3 8B running on iPhone

What is Hermes 2 Pro Llama 3 8B?

The Hermes 2 Pro Llama 3 8B model is an upgraded and retrained version of Nous Hermes 2, built upon an updated and cleaned version of the OpenHermes 2.5 Dataset and the JSON Mode dataset developed by Nous Research. This model maintains the excellent general task and conversation capabilities of its predecessor while excelling at JSON Structured Outputs and improving on several other metrics.

According to benchmarks, the Hermes 2 Pro Llama-3 8B model surpasses Llama-3 8B Instruct on AGIEval, GPT4All Suite, TruthfulQA, and BigBench. This demonstrates the model's strong performance across a variety of tasks and evaluation criteria.

Screenshot of Hermes 2 Pro Llama 3 8B running on Mac — Hermes 2 Pro Llama 3 8B running on Mac

We believe the addition of the Hermes 2 Pro Llama-3 8B model to Private LLM will provide users with an even more capable and versatile AI chatbot experience while maintaining the privacy and security they expect from us. We hope you enjoy this new addition to Private LLM and look forward to hearing your feedback.