How to Run DeepSeek R1 Locally with OpenClaw

DeepSeek R1 has taken the AI world by storm. Its performance rivals top-tier proprietary models like GPT-4o and Claude 3.5 Sonnet, yet it’s open weights and can run on your own hardware.

When you combine DeepSeek R1 with OpenClaw, you get the holy grail of personal AI: a highly intelligent, fully autonomous agent that runs entirely on your machine, with zero API costs and total privacy.

Here’s how to set it up.

Prerequisites

• Hardware: A Mac with Apple Silicon (M1/M2/M3) or a PC with an NVIDIA GPU (8GB+ VRAM recommended for 7B/8B models).
• Software: OpenClaw installed (guide).

Step 1: Install Ollama

Ollama is the easiest way to run local LLMs. If you haven’t already:

1. Download Ollama from ollama.com.
2. Install and run it.

Step 2: Pull DeepSeek R1

Open your terminal and pull the DeepSeek model. The “distill” versions are great for most consumer hardware:

# For 8GB RAM/VRAM (Fast, good reasoning)
ollama pull deepseek-r1:7b

# For 16GB+ RAM/VRAM (Better reasoning)
ollama pull deepseek-r1:14b

# For 32GB+ RAM/VRAM (Excellent reasoning)
ollama pull deepseek-r1:32b

Step 3: Configure OpenClaw

Run the OpenClaw configuration wizard:

openclaw config

Select Ollama as your provider, and enter the model name you just pulled (e.g., deepseek-r1:7b).

Or, edit your config file directly (~/.openclaw/config.json):

{
  "llm": {
    "provider": "ollama",
    "model": "deepseek-r1:7b",
    "baseUrl": "http://localhost:11434"
  }
}

Why DeepSeek + OpenClaw?

1. Cost

Running OpenClaw with DeepSeek is free. You don’t pay per token. You can leave your agent running 24/7 to monitor your email or Discord without worrying about a surprise bill.

2. Privacy

DeepSeek R1 running via Ollama never sends your data to the cloud. OpenClaw processes everything locally. Your calendar, emails, and files stay on your disk.

3. “Reasoning” Capabilities

DeepSeek R1 is a “reasoning” model (like OpenAI o1). It “thinks” before it answers, making it exceptionally good at complex tasks like:

• Code Audit: Ask OpenClaw to review a local git repo.
• Data Extraction: Have it parse messy PDFs or websites.
• Planning: Ask it to plan a travel itinerary based on your calendar constraints.

Performance Tips

• Context Window: DeepSeek supports large contexts. OpenClaw automatically manages context, but larger windows require more RAM.
• System Prompt: OpenClaw’s default system prompt works well, but for DeepSeek, you might want to encourage “chain of thought” by adding “Let’s think step by step” to your custom instructions.

Conclusion

The combination of OpenClaw’s tools (browser, file system, apps) and DeepSeek’s intelligence is powerful. You have a free, private employee that works tirelessly on your machine.

Ready to try?

npm i -g openclaw
openclaw onboard