app·AI Tools

Ollama

CLI and API for running open models locally with simple pull-and-run commands.

FREEOPEN SOURCELOCAL-FIRSTSELF-HOSTED

Setup guide · 20–30 min

Visit Ollama →

Quick facts

Price model: Open source
Starting price: Free
Best for: Developers · Local API endpoints · Model experimentation
Replaces: OpenAI API for local tasks, Hosted LLM sandboxes
Platforms: MacWindowsLinux
Last verified: 2026-06-22

Why it's listed

The standard way to wire local models into your toolchain without cloud API invoices.

Ollama packages models into easy installs (`ollama run llama3`) and exposes a local API for apps like Continue, Open WebUI, and custom scripts. No token billing—just your electricity and hardware.

The catch

Jump to setup guide ↓

Command-line first; less hand-holding than LM Studio for non-technical users.

How to set up Ollama

Minimum path to a working local model you can chat with in the terminal and wire into other apps—no OpenAI invoice.

Time: 20–30 min
Difficulty: Moderate
Verified: 2026-06-22

Before you start

8GB RAM minimum; 16GB+ recommended for useful models
macOS, Windows, or Linux
~5–10 GB free disk per model you pull

Install Ollama
Download the installer from ollama.com for your OS and run it. On Mac/Linux you can also use: curl -fsSL https://ollama.com/install.sh | sh. Confirm it works: ollama --version
Pull your first model
Start small: ollama pull llama3.2:3b (fast on laptops) or ollama pull llama3.2 (better answers, needs more RAM). Downloads once—no per-message fee.
Chat in the terminal
Run ollama run llama3.2:3b and ask a real question. Type /bye to exit. If responses are gibberish or hang, your model is too large for your RAM—pull a smaller tag.
Confirm the local API
Ollama serves http://localhost:11434. Test: curl http://localhost:11434/api/tags — you should see your pulled models listed. This endpoint is what Continue, Open WebUI, and scripts connect to.
Optional — add a chat UI
If you want a browser interface, install Open WebUI (listed in our directory) and point it at localhost:11434. Keep Ollama running in the background.

Troubleshooting

Model download fails or is very slow: Check disk space and retry. On corporate networks, try off-hours or a home connection.
Runs but answers are empty or cut off: RAM is the bottleneck. Pull a smaller quant (e.g. :3b) or close other apps.
Port 11434 already in use: Another Ollama instance is running. Quit duplicate installs or change OLLAMA_HOST in env docs.