# Your first local chat with Ollama

After installing Ollama, run ollama run qwen2.5:1.5b to download a small model and start a local chat.

Sources checked: 2026-10-17

## 1. Install and start Ollama

Follow the installation steps for [Mac](/learn/mac), [Windows](/learn/windows) or [Linux](/learn/linux). Open Terminal or PowerShell and run `ollama --version`. Keep the Ollama app or service running.

## 2. Download a small local model

Run:

```sh
ollama run qwen2.5:1.5b
```

The official model page checked on 17 October 2026 lists a 986 MB Q4_K_M download. That is the download size, not total RAM use. The first run downloads it; later runs can reuse it. This model name selects a local model, not a cloud model.

## 3. Ask a question

At the chat prompt, type: “Explain a rainbow in three short sentences.” Wait for the reply. Type `/bye` to leave the chat.

In the terminal, `ollama list` shows downloads, `ollama ps` shows loaded models, and `ollama stop qwen2.5:1.5b` unloads this model.

## 4. Optional: use a Hugging Bay GGUF

Download a complete GGUF supported by Ollama from [Hugging Bay](/downloads) and check its SHA-256. In a plain-text file named `Modelfile`, put the following line, replacing the quoted path with your actual file path:

```text
FROM "/absolute/path/to/model.gguf"
```

Open a terminal in the folder containing Modelfile and run:

```sh
ollama create my-local-model -f Modelfile
ollama run my-local-model
```

On Windows, use the real path with forward slashes, such as `C:/Users/you/Downloads/model.gguf`. The model architecture and chat template must be supported; a GGUF extension alone is not enough. See [Ollama's import guide](https://docs.ollama.com/import).

## Sources

- [Ollama quickstart](https://docs.ollama.com/quickstart)
- [Ollama Qwen2.5 1.5B model](https://ollama.com/library/qwen2.5:1.5b)
- [Ollama settings and privacy](https://docs.ollama.com/faq)
- [Import a model into Ollama](https://docs.ollama.com/import)
- [Ollama command reference](https://docs.ollama.com/cli)

[Next: check what stays on your device](/learn/privacy)