# Hermes
> Disambiguation; this note is the **Hermes LLM family**. The agent harness that shares the name is [[Hermes Agent]]. Both ship from [[Nous Research]] and are commonly referenced together; they are different artifacts.
Hermes is the flagship open-weight [[Large Language Models (LLMs)|LLM]] series from [[Nous Research]]. The Hermes models are instruction-tuned fine-tunes built on top of popular open-weight bases (Llama, Mistral, Qwen), known for strong steerability, neutral system-prompt behavior, and first-class support for tool use and structured output — which has made them a common backbone for agentic workflows.
## Key characteristics
- **Open-weight** — distributed via Hugging Face under the Nous Research org
- **Steerability** — designed to follow user/system prompts without an opinionated built-in persona
- **Function calling and structured output** — first-class, aimed at agent integration
- **Base-model diversity** — Hermes variants exist on top of Llama, Mistral, and Qwen bases (e.g., Hermes 2, Hermes 3, Hermes 4)
- **Agent-friendly** — commonly used as the model backbone for autonomous agent stacks alongside closed-weight options
## In the agentic coding ecosystem
Hermes is cited alongside [[OpenClaw]] as one of the autonomous agent tools whose workloads [[Kimi K2.6]] targets (Moonshot's K2.6 launch singled out Hermes as a representative agentic deployment). In practice, Hermes-based agents run on stacks like [[Ollama]] or vLLM and plug into agent harnesses the same way closed-weight models do.
## Why it matters
- Demonstrates that open-weight fine-tunes (not just base models) can be competitive substrates for serious agent work
- Pairs naturally with the [[AI Open Weight Models|open-weight]] frontier ([[Kimi]], [[Qwen]], [[Deepseek]]) to form a fully-open agent stack
- Backed by [[Nous Research]]'s broader bet on decentralized training and user-sovereign AI
## References
- Nous Research Hugging Face: https://huggingface.co/NousResearch
- Nous Research website: https://nousresearch.com/
- Kimi K2.6 launch (agentic tool context): https://www.kimi.com/blog/kimi-k2-6
## Related
- [[Hermes Agent]]
- [[Atropos]]
- [[Nous Research]]
- [[Large Language Models (LLMs)]]
- [[AI Open Weight Models]]
- [[AI Agents]]
- [[AI Tool Use]]
- [[AI Fine-Tuning]]
- [[Kimi K2.6]]
- [[OpenClaw]]
- [[Ollama]]