# Nous Research Nous Research is an applied research group and open-source AI collective. The team builds and releases open-weight [[Large Language Models (LLMs)]], fine-tuning datasets, and tooling, with a stated emphasis on decentralization, transparency, and preserving human agency in the AI ecosystem. They are best known for the [[Hermes]] series of fine-tunes (e.g. Hermes 2, Hermes 3) built on top of popular open-weight bases (Llama, Mistral, Qwen), as well as for pushing the boundaries of distributed and decentralized model training (e.g. DisTrO, DeMo optimizers, Psyche). Main features and contributions: - Hermes series — general-purpose instruction/assistant fine-tunes known for steerability and neutral system-prompt behavior - Open-weight releases published on Hugging Face with permissive licensing when possible - Research into distributed/decentralized training across heterogeneous hardware and unreliable networks - Forge / Nous Chat — inference products that expose their models and others - Active community around fine-tuning, dataset curation, and agentic workflows Common uses: - Running capable open-weight assistants locally or self-hosted - Starting points for further [[AI Fine-Tuning]] - Research on decentralized training and alignment alternatives to centralized labs ## References - Official Website: https://nousresearch.com/ - Hugging Face: https://huggingface.co/NousResearch - X / Twitter: https://x.com/NousResearch - Discord: https://discord.gg/NousResearch - GitHub: https://github.com/NousResearch ## Related - [[Hermes]] - [[AI Open Weight Models]] - [[Large Language Models (LLMs)]] - [[AI Fine-Tuning]] - [[Open Source]]