Nous Research is a decentralized AI research group focused on developing open-source, human-centric language models and AI tools. Founded in 2022 or 2023 (sources vary), it emphasizes advancements in large language model (LLM) architecture, data synthesis, fine-tuning, and decentralized training. Key projects include:
Hermes 3: A fine-tuned LLM based on Llama 3.1 (8B, 70B, 405B), designed for enhanced reasoning, creativity, and multi-turn conversations, rivaling top models while remaining open-source.

DisTrO: A distributed training framework that reduces inter-GPU communication, enabling AI model training across global internet-connected GPUs, demonstrated by pre-training a 15B parameter LLM.

Psyche: A decentralized platform on the Solana blockchain for collaborative AI training, aiming for accessibility and fault-tolerant model development.

Nous-Forge: An upcoming AI orchestration composer for integrating multiple models.