#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/openclaw/setting-up-your-ai-assistant/choosing-an-ai-model-local-with-ollama

░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████████████████████████████████████████████████████████████████████████████████████████████████████████

#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/openclaw/setting-up-your-ai-assistant/choosing-an-ai-model-local-with-ollama

Repovive

Choosing an AI Model: Local with Ollama - Setting Up Your AI Assistant | OpenClaw | Repovive

Repovive.

Choosing an AI Model: Local with Ollama

Local LLM with Ollama

4 min

What if you don't want your conversations leaving your machine? Run a model locally with Ollama.

Install Ollama from ollama.com, then pull a model:

ollama pull llama3.3

OpenClaw auto-discovers a running Ollama instance on localhost:11434. No API key needed.

Local models are great for privacy but come with trade-offs. You need a GPU with at least $8$ 8 GB of VRAM for usable response speeds. Smaller models produce shorter, less accurate answers compared to cloud options. If your assistant handles sensitive data, the privacy benefit outweighs the quality gap. Can you live with that trade-off? Only you can decide.