Install Deepseek in 60 Seconds

Deepseek is the hot new model that everyone is talking about. OpenAI is so worried about this model taking over the news cycle that they dropped their new o3 model in response.

There is a lot of brilliance to how this model is built and their marketing strategy is to let everyone use the tool for free.

But that's not a good idea...especially if you're a business user.

You can run the model on your local machine and get all the benefits of using this powerful new AI model, without transmitting your data around the world.

I'm going to show you exactly how to install Deepseek in less than one minute.

Step One - Install LM Studio

Head over to https://lmstudio.ai/ and install the version that matches your operating system.

The installation process will be slightly different for each operating system, but double clicking the file should start the process.

Step Two - Open LM Studio

The first time you open LMStudio, this is the welcome screen.

Click on "Get Your First LLM"

If you're already past that screen, you can click the magnifying lens on the left sidebar.

This will open the search dashboard where you can see many amazing open source AI models to choose from.

Step Three - Choose Your Model

By default LM Studio will show the most popular models that can run on your machine on this page.

There are already many versions of Deepseek, so we want to enter that into the search bar to choose the right one for your machine.

There are multiple versions of each model to choose from.

Click this small icon and LM Studio will show you a list of all the available versions of the model you can choose from.

There is a lot of information about each model on this page.

LM Studio will show you the size of each AI as well as if it will run on your machine.

Green means it will run great.

Red means your machine can't handle it.

In between are the models that will run slowly on your machine.

Choose the DeepSeek model from unsloth and choose the model that will run on your machine.

Step Four - Download Your Model

Find a version of DeepSeek that will run on your machine.

My new M4 Mac Mini can handle all the options, so I'll download the 16.07GB model.

When you find the model you want, click the DOWNLOAD button in the bottom right corner.

A download window will pop up with the progress of your download.

Wait for this to complete.

Step Five - Activate DeepSeek

When your download completes, you will see the option to load the model.

Click this button to load the model.

If you already closed this window there are several other was to find this button.

The download button will change color and now say "Use in New Chat."

This will also load the model.

You can also leave the search section and click on the little chat icon.

This will take you to the main chat window.

Click the bar at the top of the screen to choose from any models you have installed.

Customize Your Model

LM Studio will give you a bunch of options to modify your version of DeepSeek.

If you really like to get under the hood and tinker, you have that power.

There is no need to change any of these settings if you chose a model that LM Studio marked in green.

Click "Load Model"

Using DeepSeek on Your Local Machine

The beauty of running a model locally is that none of your information is transmitted through the internet.

You can run this AI without an Internet connection now that you've downloaded it.

Keep in mind that there is a tradeoff in speed.

The larger the model you download, the slower it will run on your machine.

The accuracy leaves something to be desired.

When asking about myself, DeepSeek got the answer completely wrong.

Neither of these books were written by ANY Jonathan Green.

They weren't even written by the same person.

This may be a function of DeepSeek being a reasoning model.

Rather than lie, ChatGPT's new o3 model claims that there is no universally accepted winner for this category.

It took DeepSeek 33 seconds to get the answer wrong and ChatGPT-o3 24 seconds to claim the answer is unknowable.

Fortunately, ChatGPT-4o remains my faithful companion and got the answer right.

This answer has changed significantly since the last time I asked six months ago, which shows that ChatGPT-4o is not static.

It's still improving...even if they don't update the name.

Use Different AI Models

Not every model is right for every situation.

Reasoning models are better for solving complex problems and coding.

The beauty of LM Studio is that you can load multiple models in your machine and switch between them for different tasks.

You can quickly test to see which models are the most useful for your personal use cases.

Those bar graphs they all use each time a new model is released aren't nearly as useful as testing a model to see if it answer the question correctly.

Ready to Take Advantage of AI’s Real Potential?

As the influence of AI grows, knowing how to strategically implement it becomes a vital competitive skill. Ready to see how AI can directly impact your business?

Take our quiz to discover how AI can boost your revenue and streamline operations: Take the Quiz