Quick Facts

  • Level: L3 (Dev Setup)
  • Type: Local LLM runtime
  • Maker: Ollama (open source)
  • Runs on: Your machine (macOS, Linux, Windows)
  • Access: Fully local - nothing leaves your machine

What It Solves

Ollama lets you run open-source LLMs on your own hardware. No API keys, no cloud, no data leaving your machine. It's the privacy-first approach to AI coding assistance.

Where It Belongs

APL 3 with a unique advantage: AAL is inherently limited because nothing leaves your machine. This makes it ideal for sensitive codebases where data privacy is non-negotiable.

What It's Good At

  • Privacy-sensitive environments (no data leaves your machine)
  • Offline development (works without internet)
  • Cost-free inference after initial setup
  • Experimentation with different models (CodeLlama, Mistral, Phi, etc.)
  • Integration with other L3 tools as a backend
  • Learning how LLMs work by running them directly

What It's Not For

  • Users without decent hardware (8GB+ RAM recommended, 16GB+ ideal)
  • Tasks requiring the latest frontier model performance
  • Complex agentic workflows out of the box (it's a runtime, not an agent)
  • People who need managed infrastructure

Risk Profile

Ollama has a unique risk profile - it's both L3 in capability and inherently private. The main risks are:

  • Model quality: Local models are less capable than frontier cloud models
  • Hardware requirements: Running large models needs significant RAM and GPU
  • Over-trust: Just because it's local doesn't mean the output is correct

Getting Started

Install Ollama, pull a model, and start generating. The setup is straightforward:

  1. Install from ollama.com
  2. Pull a code model: ollama pull codellama
  3. Start using: ollama run codellama

Related Models