Nous Research

Updated 2026-04-06

AI research lab focused on open-source models and agent frameworks. Best known for the Hermes model series, fine-tunes for tool use and agent tasks, and for the Hermes Agent framework.

Hermes Models

Nous Research regularly publishes fine-tunes of base models such as LLaMA or Mistral, optimized for structured output, tool calling, and reasoning. In the open-source community, Hermes is often treated as a reliable benchmark for agent capability, better tuned for “do something” than for “explain something.”

Hermes Agent Framework

An open-source agent that can run on your own server, persist context, and teach itself new skills. It is conceptually similar to OpenClaw, but more research-oriented.

Key traits:

  • installable with a lightweight script
  • model access via Nous Portal, OpenRouter, or your own endpoint
  • multi-platform support across chat apps, email, and CLI
  • multiple sandboxing options
  • dozens of built-in skills plus community skills
  • research features such as batch trajectories and RL-training integrations

One important distinction: this is not the same thing as the local Hermes Agent service in this vault.

Connections