DeepSeek-R1 is a Mixture-of-Experts (MoE) model with 671 billion parameters, utilizing 37 billion active parameters per token. It has been trained through large-scale reinforcement learning, emphasizing enhanced reasoning abilities. The training process includes two reinforcement learning phases to refine reasoning patterns and align with human preferences, as well as two supervised fine-tuning (SFT) stages to establish both reasoning and non-reasoning skills. The model demonstrates performance on par with OpenAI-o1 in mathematics, coding, and reasoning tasks.
DeepSeek developed DeepSeek-R1 and released it on January 20, 2025. It supports a context window of 128K tokens. DeepSeek-R1 is an open-source model, with its code and weights freely accessible under a permissive license. It is capable of handling and producing text, making it ideal for tasks such as writing, summarizing, analyzing, and engaging in chat interactions.
In this guide, I’ll walk you through the process of installing and running Deepseek R1 on Ubuntu 24.04 using Ollama. To start, follow the first three steps for installing Ollama from this link.
Download Deepseek-R1
We can now download the DeepSeek model. It includes several distilled models based on Qwen and Llama architectures, each designed to address different performance and resource requirements.
These models come in various sizes. For instance, the 1.5b model is approximately 2.3 GB, the 7b model is about 4.7 GB, and the 70b model is over 40 GB. I opted for the default deepseek-r1 model.
The 1.5b model demands fewer resources, whereas models like the 14b and 32b are intended for more intensive performance. You can explore all the available models here.
ollama run deepseek-r1 input "Hello, how can I use this model for NLP tasks?"For a more user-friendly interface to interact with Deepseek R1, we’ll install OpenWebUI. You can follow the complete installation guide here: How to Install OpenWebUI on Ubuntu 24.04
Conclusion
You now have Deepseek R1 running on your Ubuntu 24.04 system with a user-friendly web interface! You can start exploring its capabilities for various natural language processing tasks using either the command line or OpenWebUI. Remember to monitor system resources while using the model and adjust settings as needed for optimal performance.