Download models

$ ollama pull llama3.1

Start interactive CLI

See Model library in the GitHub docs for models and CLI commands to download.

e.g.

$ ollama run llama3.1

Run a once-off request

Pass as an argument e.g.

$ ollama run llama3.1 'Tell me a joke'

Read a text file:

$ ollama run --nowordwrap llama3.1 "Summarize this file in one sentence: $(cat file.txt)"

Read an image file such as with bakllava or llava image models:

$ ollama run --nowordwrap bakllava "What's in this image? ./photo.jpg"

Access via HTTP

When a model is requested, it will be loaded and used.

Then you can find it under running models here:

$ ollama ps

Stop

See conversation on GitHub here. There is no builtin way to stop ollama. Use your system manager or use pkill.

On macOS, find the icon as running apps and click Quit.

If you started with ollam serve, then just stop that process to stop Ollama.