Member-only story
Introduction to NousResearch’s Hermes 3
A New Standard in AI Models
Author
- Tohfa Siddika Barbhuiya (ORCID: 0009–0007–2976–4601)
Introduction
Instruct (or “chat”) tuned models have become the primary way in which most people interact with large language models. As opposed to “base” or “foundation” models, instruct-tuned models are optimised to respond to imperative statements.
NousResearch’s Hermes 3 is a breakthrough in Generalist Instruction Models, offering advancements in Language Models and versatile AI Capabilities. Hermes 3 contains advanced long-term context retention and multi-turn conversation capability, complex roleplaying and internal monologue abilities as claimed by the developers. Hermes 3 training data aggressively encourages the model to follow the system and instruction prompts exactly and in an adaptive manner. Hermes 3 was created by fine-tuning Llama 3.1 8B, 70B and 405B, and training on a dataset of primarily synthetically generated responses. These models are not only powerful but also uniquely aligned to follow system and instruction prompts with unparalleled precision and neutrality. The model boasts comparable and superior performance to Llama 3.1 while unlocking deeper capabilities in reasoning and creativity. This article deep dives into the innovative…