Your First Model
Ollama is the simplest way to run models locally: one program that downloads models, runs them, and serves them to you and your programs. This chapter gets you chatting in minutes.
Install it
Linux — one line in a terminal:
curl -fsSL https://ollama.com/install.sh | sh
Mac or Windows — download the installer from ollama.com and run it. (No command line needed for the install itself.)
Check: confirm it's there:
ollama --version
You should see a version number printed back. If the command isn't found, the install didn't finish — re-run the step above.
Download a small model
Start small. tinyllama is a 1-billion-parameter model — about 0.8 GB by the size formula — so it fits almost anywhere and downloads quickly:
ollama pull tinyllama
Check: the download completes, and then:
ollama list
shows tinyllama in your local model list.
Chat
ollama run tinyllama
If the model isn't downloaded yet, run fetches it first — pull
and run in one step. You'll get a >>> prompt. Type something:
>>> Write a haiku about a rainy Tuesday.
The model answers, word by word, right in your terminal. Ask a few things. Notice it's small — answers are simple, sometimes clumsy. That's a 1B model being honest about its size.
Two commands to know inside the chat: /help lists the chat
commands, and /bye exits back to your terminal.
The other commands you'll actually use
| Command | What it does |
|---|---|
ollama pull <model> |
Download a model |
ollama run <model> |
Chat with a model (downloads it first if needed) |
ollama list |
Show your downloaded models |
ollama ps |
Show models currently loaded and running |
ollama rm <model> |
Delete a model to free disk space |
ollama show <model> |
Show details about a model |
ollama --version |
Show the Ollama version |
Check: run ollama ps while a chat is open in another
terminal — you'll see your model listed as loaded. Close the chat
with /bye and check again later: Ollama unloads idle models on
its own to free memory.
That's the whole loop: pull, run, chat. The next chapter is about getting better answers out of the conversation.
Next: Talking to Models