Hermes (NousResearch)
Fine-tuned open models built for tool use, function calling, and agentic reasoning.
NousResearch's Hermes series are fine-tuned open-source models optimized for agentic tasks — tool calling, structured JSON output, multi-step reasoning, and system prompt adherence. Built on top of Llama and Mistral base models, Hermes models consistently outperform their base counterparts on function calling benchmarks and are a popular choice for self-hosted agentic pipelines.
Open SourceModelsTool UseAgentic
Published Mar 10, 2026