How to Use Local Models with OpenClaw and Ollama

Why Local Models?

• Zero API costs — run as many conversations as you want for free
• Complete privacy — no data leaves your machine
• Offline capable — works without internet after model download

Setup

1. Install Ollama from ollama.com
2. Pull a model: ollama pull llama3.3
3. Configure OpenClaw to use Ollama as a provider

Recommended Local Models

• Llama 3.3 70B: Best quality for local (needs 48GB+ RAM)
• Llama 3.1 8B: Good for 16GB RAM machines
• Qwen 2.5: Strong multilingual support
• Mistral: Fast and efficient

Hardware Requirements

Local models need significant RAM. A 7B model needs ~8GB RAM, 13B needs ~16GB, and 70B needs ~48GB. GPU acceleration (NVIDIA CUDA or Apple Silicon) dramatically improves speed.

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull llama3.1

# Configure OpenClaw
// In ~/.openclaw/openclaw.json
{
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollama/llama3.1"
      }
    }
  }
}