Alright, you've learned about the tools available, like Ollama and LM Studio, which act as helpful assistants for running Large Language Models on your computer. Now it's time for the most rewarding part: actually running a model and seeing it generate text based on your input.

This section provides practical steps to get you started. We'll use a relatively small model for this first exercise to ensure it runs smoothly on a wider range of hardware. Remember from the previous sections, you should have already installed either Ollama or LM Studio. Choose the tool you installed and follow the corresponding steps below.

Running a Model with Ollama (Command Line)

If you opted for Ollama, you'll interact with it using your terminal or command prompt.

Download a Model (If You Haven't Already): Let's download a small, capable model. We'll use phi3:mini, a model developed by Microsoft known for its good performance relative to its size. Open your terminal and run:
```
ollama pull phi3:mini
```
You'll see download progress indicators. This might take a few minutes depending on your internet speed. Once downloaded, the model is stored locally.
Run the Model Interactively: Now, start an interactive session with the model using this command:
```
ollama run phi3:mini
```
Ollama will load the model (this might take a moment, especially the first time) and then present you with a prompt, often looking like >>> Send a message (/? for help).
Interact with the Model: Type your prompt directly after the >>> and press Enter. Let's try asking it to do something creative:
```
>>> Write a short story about a robot who discovers gardening.
```
The model will process your request and generate a response, streaming the text output directly to your terminal.
Continue the Conversation: You can continue interacting. The model remembers the previous parts of the conversation within its context window (as discussed in Chapter 5). Try asking a follow-up question.
Exit the Session: When you're finished, you can exit the Ollama interactive session. Type /bye and press Enter, or on most systems, you can press Ctrl+D.

Basic workflow for running a model interactively using Ollama via the command line.

Running a Model with LM Studio (Graphical Interface)

If you prefer a graphical interface, LM Studio makes running models straightforward.

Launch LM Studio: Open the LM Studio application you installed previously.
Download a Model (If Needed):
- Go to the "Search" tab (often represented by a magnifying glass icon 🔍).
- In the search bar, type phi3 mini instruct. You'll see various versions. Look for a GGUF format model, preferably one with Q4_K_M or Q4_0 in the name, as these offer a good balance of size and quality. For example, you might find Phi-3-mini-4k-instruct-q4_0.gguf.
- Click the "Download" button next to the model you choose. You can monitor the download progress in the bottom section of the application.
Load the Model for Chat:
- Click the "Chat" tab (often represented by speech bubbles 💬).
- At the top, you'll likely see a dropdown menu labeled "Select a model to load". Click it and choose the Phi-3 model you just downloaded.
- LM Studio will load the model into memory. This can take some time, and you can usually see the progress or status messages on the screen (often on the right-hand side). Wait for it to indicate the model is ready.
Interact with the Model:
- Once the model is loaded, you'll see a chat interface similar to online chatbots.
- Find the text input box at the bottom. Type your prompt there. Let's try a simple instruction:
```
Explain what a Large Language Model is in one sentence.
```
- Press Enter or click the "Send" button.
- The model will process your request, and its response will appear in the chat window above.
Continue Chatting: You can type more prompts and carry on a conversation just like you would with the command-line version. LM Studio manages the interaction history for you.

Basic workflow for running a model interactively using the LM Studio graphical interface.

A Quick Note on Performance: Loading the model (transferring it from your storage drive into your computer's RAM or VRAM) is often the slowest part, especially the first time you run a specific model after starting the application. Once loaded, generating text should be relatively faster, though the speed still depends heavily on your hardware (CPU, GPU, RAM) and the size of the model.

Congratulations! You have successfully downloaded and run your first Large Language Model locally. You prompted it, and it generated text based on your input, all running entirely on your own machine. In the next chapter, we will look more closely at how to effectively communicate with these models through prompting.