What it is
Ollama is a small program that downloads open-weight models and runs them on your computer. It gives you a chat in the terminal and an HTTP API on localhost, so your own apps can use a local model instead of a paid cloud API.
Why people use it
- Zero API bill. Once a model is downloaded, every prompt is free.
- Privacy. Prompts stay on your machine, which helps with client code or personal notes.
- Works offline, for example on a train or with bad Wi-Fi.
Quick start
bash
# macOS / Linux / Windows installers are on ollama.com
ollama run gemma3Call it from code through the local API:
bash
curl http://localhost:11434/api/generate -d '{
"model": "gemma3",
"prompt": "Explain recursion in one paragraph."
}'When not to use it
- Low RAM. On an 8 GB laptop, stick to small models (roughly 1–4B parameters), or it will be slow.
- Team or production serving. Ollama is built for one machine. For many users, look at dedicated inference servers.
FAQ
Is Ollama free?
Yes. The app is open source under the MIT license. You pay only for your own hardware and electricity.
Can I use it from Python or JavaScript?
Yes. It exposes a local HTTP API, and there are official client libraries.