Hermes 4: step-by-step

Step by step

Follow along

1. 1

   Choose how you will access Hermes

   For most beginners the simplest paths are: (a) via API through OpenRouter — no local hardware needed, or (b) running locally with Ollama — free but requires a capable machine. The steps below use Ollama as the beginner-friendly local path.

2. 2

   Install Ollama

   Download and install Ollama from ollama.com for macOS, Linux, or Windows. Ollama runs as a local server that manages model downloads and inference, so you do not need to configure Python environments or GPU drivers manually.

3. 3

   Pull the Hermes 3 model

   Open a terminal and run 'ollama run hermes3' (defaults to the 8B model, ~4.7 GB). Ollama downloads the model weights automatically on first run and drops you into an interactive chat prompt immediately after.

4. 4

   Chat with Hermes locally

   Type your message at the prompt and press Enter. Hermes 3 is trained for strong instruction-following and multi-turn conversation; it also supports function-calling and JSON-mode outputs for structured tasks.

5. 5

   Access Hermes via API on OpenRouter (alternative path)

   Create a free account at openrouter.ai, add credit, then use the model ID 'nousresearch/hermes-4-70b' in any OpenAI-compatible API call. This lets you run the larger 70B variant without local hardware.

6. 6

   Call the Ollama API from code

   With Ollama running, send a POST to http://localhost:11434/api/chat with a JSON body containing model 'hermes3' and a messages array. The response is OpenAI-chat-compatible, so existing client libraries work without changes.