The AI Stack
Sign in
L4 · Models & Training

Nous Research

Trains world-class open-source language models like Hermes using distributed infrastructure.

Updated April 2026

Overview

Segment
Frontier Foundation Model Labs
Posture
Open-Weight Frontier

Product overview

Nous Research develops the Hermes series of open-source LLMs optimized for reasoning, long-context handling, and user alignment, hosted on Hugging Face with millions of downloads. Researchers, developers, and enterprises use them for local inference, chat interfaces like Nous Chat, and advanced applications via Forge Reasoning API. They stand out from closed frontier labs like OpenAI by fully open-sourcing models, data, and methods, plus pioneering decentralized training via Psyche network on Solana.

Revenue model

API usage-based pricing per token via Forge API (e.g. Hermes 4 70B: $0.13/$0.40 input/output per 1M tokens; Hermes 4 405B: $1.00/$3.00); open-source models free; funded by $85M+ VC (Paradigm-led Series A).

Moat

  • Proprietary Technology
  • Platform Effects
  • First Mover

Nous Research's key competitive moat lies in its pioneering open-source AI models like Hermes, emphasizing user-aligned, neutrally aligned systems with full transparency in data curation, training methodologies, and distributed GPU infrastructure via Psyche, enabling broad collaboration and customization that differentiates it from closed proprietary AI giants.