How To Install Ollama And Gemma 4

In the video below, we show how to install Ollama and Gemma 4

---

AI has been with us for some time now but as useful as it is, a big problem is privacy or rather the lack of it

You’ve really no idea what happens to any information you provide and so there is the risk of any sensitive data being leaked

As a result, Local AI is becoming more and more popular

Now fairly recently, Google released their Gemma 4 model and it sounds useful when you’re considering local AI without necessarily having high end hardware

But how do you install Ollama for instance and use Gemma 4 with it?

And what’s Gemma 4 like to use?

Install Ollama:
Now in order to setup a local AI, you first need a platform like Ollama

So the first thing we need to do is to download Ollama

To do that we’ll point our web browser at their website
https://ollama.com/download

There are options here for macOS, Linux and Windows

So make sure to select the correct OS and follow the download instructions

Now Windows is always easy, you just have to click on the Download button

So sticking with Windows, once the file is downloaded, double click on this to start the installation process

Now click Install and that’s basically it. It will install Ollama and it will then launch it ready to use

Choose AI Model:
Ollama is not an AI solution in itself, think of it as more of a frontend if you will

In which case you have to pick an AI model for it to use

To test this out, click the New Chat option in the left hand pane

Click on the drop down menu in the bottom right corner and you’ll see in this video example it offers gemma4:26b to download locally

NOTE: To use a cloud model you’ll need to setup an Ollama account

But your choices will vary as I’ve installed this once before and back then it offered me gemma4:12b and qwen3.6

That doesn’t seem to be much choice but if in the text field you type gemma for instance you will be able to select a default model for gemma

Now at this stage things looked a little confusing to me as it seemed like this model needed to be downloaded but clicking on it didn’t accomplish anything

And while you could open a terminal or command session and instruct Ollama to download a model, you can just type something in the chat window like

“hello who are you?”

Ollama will then download and install the selected model, load it and it should then respond

So you don’t necessarily have to open a terminal and download models that way

Prompt and Image Testing:
Now like most people I could do with some more money so maybe I could ask this AI to write a small story for me so I can sell it and get rich

How about this for a prompt?

“tell me a kid’s story about a puppy who goes into the garden, meets a sparrow and they have fun. Keep it to under 150 words”

Now how long it takes to get a reply very much depends on the capabilities of your computer but you should expect some response

OK there’s clearly more to me becoming a best selling author than I thought but hopefully as you can see you ask this AI for assistance just as you can with the typical cloud based ones

Now, it probably won’t be as responsive as anything you’ve used in the cloud, it likely won’t have access to as much information and the quality may not be as good

But the key point here is that because the AI is running locally, the information exchanged is exchanged to stay local

Now the next test is can it identify things in images?

So I’ll drag an image into the chat and ask it

“where was this photo taken and what is it”

Granted, this is an iconic place but all the same it was able to work with a photo

Code Testing:
Now why write code these days if you can get an AI to do the work for you?

No matter how good you are, chances are an AI can do this much quicker

Granted you still should check the code afterwards, as even AI isn’t that good, even if does somehow manage to understand what you ask

So I’ll give it a basic prompt

“write a simple html page that counts down from 10 to 0 then changes the background to a random colour. it then repeats this countdown. include everything in a single file”

Once it’s finished we’ll click the copy button to put this code into the clipboard

Then open Notepad, paste in the code and save this as an html file

The OS typically associates html files with a default web browser, so if we double click on the file we should see the results

Now granted this isn’t doing much, but when you think about it that would normally take quite a while to put together yourself even if it sounds like quite a simple task

Maths Testing and Reasoning:
Sometimes you’ll want an AI to solve a maths problem so let’s try this one

“a train leaves station 1 at 08:10 at a speed of 60mph. at 09:25 a train leaves station 2 at 75mph and arrives at station 1 at 11:35, at what time will the first train reach station 2”

To be fair we aren’t taking into account complexities like acceleration or deceleration but the goal here is to see how it handles maths

What’s really useful is you don’t just get an answer, you also get to see the reasoning and this is extremely important

At this stage you can’t be 100% certain you’re going to get an accurate response and you do have to proof check the work

But again, being able to do these sort of things can save you time

Summary:
Hopefully as you can see it’s fairly easy to setup your own local AI but we’ve only scratched the surface

You don’t need expensive hardware to get started, although the more hardware you can throw at this the better

But what’s particularly interesting, is once you’ve got your own AI setup, you can start thinking about using agents to take advantage of it