Deepseek is a powerful open-source language model that is suitable for research, code generation and many other AI applications. With Ollama, you can easily run Deepseek locally on your computer – without having to rely on a cloud connection. In this article, you will learn how to install, configure and securely use Deepseek with Ollama.
Deepseek is a modern Large Language Model (LLM) that has been specially optimized for high performance and efficiency. The DeepSeek R1 model was released in January 2025 and attracted international attention because it is comparable to the currently strongest model OpenAI GPT-o1 and is also available for free as open source and was created at a fraction of the cost.
Deepseek R1 in Ollama hosted locally on your own computer
Deepseek R1 is suitable for many typical use cases such as:
Automated text creation and text analysis
Creation of code and technical documentation
Researching and summarizing large amounts of data
Chatbot development
What advantages does Deepseek offer over other AI models and tools?
Deepseek R1 offers the opportunity to use a powerful AI model similar to GPT-o1 without passing on data to US companies or other organizations and also to have full control over data and the AI system thanks to its local usability on your own computer or server. This is particularly interesting for companies that need to ensure data protection and governance as well as high system availability and independence from external IT services.
How can you use Deepseek R1 with Ollama on your own server?
Ollama is a tool for Windows, MacOS and Linux. It allows you to run modern open source AI models directly on your own computer. The advantages of Ollama in combination with Deepseek:
Local execution: no data leaves your own computer.
Easy installation: Ollama makes the setup of Deepseek uncomplicated.
Better control: Adjustments and optimizations are possible directly on your own system.
This makes it easy to test and use the basic Deepseek model.
What requirements does Deepseek R1 need on my own computer?
When using Deepseek on your own server, there are limits set by the hardware of the computer. The Deepseek R1 model is therefore available in several versions, from small to large. Depending on how powerful the computer is, you can choose a larger model:
Parameters of the Deepseek model:
Deepseek R1 is currently available in the parameter versions 1.5b, 7b, 8b, 14b , 32b, 70b and 671b. Reading example: The “Deepeek R1 8b” model is based on 8 billion parameters. The number of parameters can be roughly imagined as the number of neurons in the “brain of the AI”, the neural network (note: however, parameters work differently to real neurons)
Computing power required to use Deepseek:
Deepseek can be run on a modern Windows computer or MacBook M1-M3 with approx. up to 8b
Larger models of up to 70b can be operated on a Hetzner server
On a cloud platform such as Azure, AWS or GCP, the 671b model can be used with scalable cloud hardware
Installation: Set up Deepseek R1 with Ollama on your own computer
For the installation you need a Win/Mac/Linux computer that is as up-to-date as possible, approx. 5 GB of free hard disk space, the free Ollama tool and a suitable Deepseek model. This can be done in approx. 5 minutes even by non-experts with slightly advanced computer skills.
Here’s how to do it step by step:
1. Install Ollama
Ollama makes it easy to install and use various AI models such as Deepseek, Mistral and others. However, it does not have a graphical user interface, so the AI model is started by text command. Ollama is available for macOS, Linux and Windows.
For Windows and MacOS:
Here is a simple installer with which you can install Ollama with one click: https://ollama.com
For Linux and macOS:
Professional users can install Ollama with the following command:
curl -fsSL https://ollama.ai/install.sh | sh
2. Download Deepseek model in Ollama
Once Ollama is installed, Deepseek can be downloaded and started with a single command. To do this, start a terminal and enter the following command. The model is then automatically downloaded from the Ollama repository and saved on the local computer. This takes approx. 10 minutes for a 7b model, as approx. 4-5 GB are downloaded.
After the successful download, the model can be started. Deepseek is now ready and can be used for queries via the terminal.
ollama run deepseek-r1:7b
Example: Creating a computer science ebook for 7th graders with Deepseek
As an example, we give Deepseek R1 (7B) the task of helping us to create an ebook. It should teach Python basics for students in grade 7.
Help me create an ebook for 7th graders on the topic 'learning to program with Python'
Result:
Deepseek R1 is a reasoning model, just like OpenAI GPT-o1. This means that it first thinks explicitly and develops a plan on how to reach its goal. This greatly improves the quality of the results compared to other models, but takes a little longer.
Deepseek R1 in Ollama hosted locally on your own computer
Result: Table of contents and text for the ebook
As a result of the reasoning phase, Deepseek has created a sensible structure adapted to the “Class 7” target group. The result is already good even for the small 7B model and can of course be further refined through better prompting.
Deepseek R1 in Ollama: Result of the ebook creation prompt
The next steps can be to translate the ebook from English into the desired language. Deepseek supports numerous languages in high text quality, so this step is very easy. For further fine-tuning, copy the text into Word or a similar tool, add more paragraphs, images and visualizations and then create a PDF from it.
Use cases: How can Deepseek be used with Ollama?
1. Research and data analysis
Deepseek is ideal for the automated processing of large volumes of text. One example:
"Summarize the latest research results on AI-based diagnostic procedures."
The model provides a concise summary and thus helps to process information efficiently.
2. Code generation and debugging
Deepseek can help developers to write code and analyze errors. Example:
"Write a Python function that sorts a list of numbers."
The model provides a complete implementation of the desired function.
3. Chatbots & automations
Deepseek can be used to create interactive chatbots that run completely offline. This is ideal for privacy-sensitive applications or internal business solutions.
Conclusion & outlook
Deepseek with Ollama offers a powerful way to use the popular AI model securely locally without exposing your data across national borders. Even with the small 7B model in our example task for ebook creation, the quality of the results is already very usable. It is particularly suitable for developers, companies and researchers who value data protection and performance.
Test Deepseek for yourself: Install Ollama and start your first AI model directly on your computer!
