How to Install and Run DeepSeek AI on Ubuntu/Debian (No GPU)

DeepSeek is a cutting-edge AI model designed for natural language processing, offering powerful capabilities such as text generation, summarization, and reasoning. It can run locally on Linux, making it an excellent choice for users who want privacy, control, and offline access to AI.

One of DeepSeek’s strengths is its flexibility while it can run on CPU-only systems, performance is significantly improved when using a dedicated GPU. On a CPU, response times may be slower, and larger models may require substantial RAM. With a GPU, DeepSeek can generate responses much faster by leveraging parallel processing, making real-time interactions more seamless.

This guide will walk you through the installation and setup of DeepSeek on Ubuntu or Debian-based Linux distributions, ensuring you can get started with AI on your own machine, whether you have a high-end GPU or not.

In this tutorial you will learn:

• How to install and configure Ollama for running DeepSeek
• How to optimize system resources for best performance

Prerequisites

Before we begin, ensure your system meets the minimum requirements. While DeepSeek can run on a CPU-only machine, having a high-performance processor and sufficient RAM will improve execution speed.
If a compatible GPU is installed, Ollama will automatically detect and utilize it for accelerated processing. If no GPU is found, a message will be displayed indicating that the model is running on the CPU.
No manual configuration is required.

Installation Steps

Video

1. Install Ollama: Ollama is required to run DeepSeek models. It provides an optimized local runtime for running machine learning models efficiently.First, ensure that curl is installed on your system. If it’s not installed, you can use apt install curl with:

   $ sudo apt install curl

   Once curl is available, download and run the official Ollama installation script:

   $ curl -fsSL https://ollama.com/install.sh | sh

   After installation, verify that Ollama is installed correctly by checking its version:

   $ ollama --version

   Additionally, ensure that the Ollama service is running with:

   $ systemctl is-active ollama.service

   

2. Download DeepSeek-R1: Now, fetch the model you want to run. DeepSeek-R1 models vary in size, balancing speed and accuracy based on your hardware capabilities. Larger models provide better reasoning and accuracy but require more RAM, VRAM, and disk space.To install the 7B model as an example, run:

   $ ollama pull deepseek-r1:7b

   Available DeepSeek-R1 Models, Hardware Requirements and Recommendations

   Model	Parameters	Disk Space Required	Minimum RAM	Recommended GPU VRAM	Performance
   deepseek-r1:1.5b	1.5 Billion	3 GB	8 GB	4 GB	Fastest, low memory usage, less accuracy
   deepseek-r1:7b	7 Billion	15 GB	16 GB	8 GB	Good balance of speed and accuracy
   deepseek-r1:8b	8 Billion	17 GB	24 GB	10 GB	Similar to 7B, slightly improved reasoning
   deepseek-r1:14b	14 Billion	30 GB	32 GB	16 GB	Better understanding, needs more RAM
   deepseek-r1:32b	32 Billion	70 GB	64 GB	24 GB	High accuracy, slower response times
   deepseek-r1:70b	70 Billion	160 GB	128 GB	48 GB	Very accurate, slow inference speed
   deepseek-r1:671b	671 Billion	1.5 TB	512 GB+	Multiple GPUs, 100 GB+ VRAM	Cutting-edge accuracy, extremely slow

    

   Choosing the Right Model:

   • 1.5B – 7B models: Best for everyday tasks, chat applications, and lightweight inference.
   • 8B – 14B models: Balanced models offering improved reasoning while staying relatively efficient.
   • 32B – 70B models: Highly advanced, suitable for research and deep analysis, but require substantial resources.
   • 671B model: Requires data-center-level hardware. Used for cutting-edge AI research.

   ---

   ---

   NOTE
   Even with 512+ GB RAM and multiple GPUs with 100+ GB VRAM, the DeepSeek-R1:671B model remains slow due to its massive 671 billion parameters, requiring an immense number of calculations per response. While multiple GPUs improve overall throughput, they don’t significantly reduce latency for a single request, as data movement, memory bandwidth, and computational limits create bottlenecks. Even high-end AI infrastructure struggles with this scale, making smaller models (7B–14B) far more practical for real-time applications. The 671B model is best suited for research and large-scale AI experiments, where precision outweighs speed.

   If you are unsure, start with deepseek-r1:7b as a general-purpose model.

3. Begin Using DeepSeek: Once the model is downloaded, you can start interacting with it directly.To run the DeepSeek-R1 model, use:

   $ ollama run deepseek-r1:7b

   You can explore more advanced usage and configurations in the Ollama Documentation.

   

4. Use a Local API for Integration: If you need to interact with DeepSeek programmatically, enable the API.

   $ ollama serve & curl http://localhost:11434/api/generate -d '{"model": "deepseek-r1:7b", "prompt": "Hello, how are you?"}'

   

Conclusion

DeepSeek offers various model sizes, each with different hardware requirements. If your system struggles with larger models, consider using a smaller variant like `1.5b`. Running DeepSeek without a GPU is possible, but optimizations will improve efficiency.