At DigitalOcean, we’ve been closely observing the rapid advancements in open-source Large Language Models (LLMs) as they catch up to the capabilities of commercial models. A significant feature of these models is reasoning—the ability to think logically and sensibly.

Traditionally, LLMs operated in a linear fashion; they would respond to prompts with straightforward answers, lacking any self-correction mechanisms. This limitation affected their ability to process complex problems, especially in language-based mathematics. However, the emergence of reasoning models, like OpenAI’s latest O1 series, has transformed the landscape. These new models effectively bridge the gap between human and machine reasoning across various tasks, including coding, math, and scientific inquiries.

In response, the open-source community has been equally industrious. The release of the DeepSeek R1 series of LLMs marks a significant milestone, showcasing public access to models capable of advanced reasoning.

In this tutorial, we’ll learn how to run DeepSeek R1 on DigitalOcean’s GPU Droplets using Ollama. Readers will be guided through the process of setting up a GPU Droplet, installing Ollama, and utilizing DeepSeek R1 for reasoning tasks.

Prerequisites

• A DigitalOcean account for accessing GPU droplets.
• Basic experience with a Bash shell, as terminal commands will be necessary.

Setting Up the GPU Droplet

To begin, create a new GPU Droplet via the DigitalOcean interface. It’s recommended to choose the “AI/ML Ready” OS and select a single NVIDIA H100 GPU, unless you’re planning to leverage the largest model, which has 671 billion parameters.

Once your machine is operational, proceed to the next step.

Installing Ollama & DeepSeek R1

Utilizing the Ollama developers’ work will allow for efficient model deployment. Open a web console window via your GPU Droplet management page and navigate to your desired working directory.

Run the following command in the terminal to install Ollama:

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

After the installation completes, you can execute DeepSeek R1. Ollama provides various model sizes, and for optimal performance using a single GPU, the 70 billion parameter model is recommended:

ollama run deepseek-r1:70b

This initial run will take some time to download approximately 40GB of data, but you will then be able to interact with the reasoning model.

Reasoning with DeepSeek R1

DeepSeek R1 is adept at handling intricate tasks in math, coding, and scientific reasoning. The model operates on a framework of principles that guide its decision-making and problem-solving capabilities, effectively utilizing metalogic reasoning to derive solutions.

For a practical demonstration, let’s consider a math problem:

A package heavier than 11 lbs and 8 oz is labeled "heavy." Gloria has packed 6 flowerpots, each weighing 1 lb and 12 oz, along with 5 oz of packing material. Will her package be labeled as "heavy"?

The model will systematically compute the package’s total weight, weighing the flowerpots and the packing material, then comparing it to the threshold to conclude whether it meets the criteria for being labeled as "heavy."

Closing Thoughts

Through this tutorial, we’ve explored how to execute DeepSeek R1 on DigitalOcean’s GPU Droplets using Ollama. The model’s ability to facilitate reasoning tasks, such as math and coding, is impressive. Additionally, we’re eager to delve deeper into R1’s architecture, model training methodologies, and what empowers its robust reasoning capabilities in future installments of this series.

Thank you for engaging with the DigitalOcean Community—explore our comprehensive offerings in compute, storage, networking, and managed databases.

Welcome to DediRock, your trusted partner in high-performance hosting solutions. At DediRock, we specialize in providing dedicated servers, VPS hosting, and cloud services tailored to meet the unique needs of businesses and individuals alike. Our mission is to deliver reliable, scalable, and secure hosting solutions that empower our clients to achieve their digital goals. With a commitment to exceptional customer support, cutting-edge technology, and robust infrastructure, DediRock stands out as a leader in the hosting industry. Join us and experience the difference that dedicated service and unwavering reliability can make for your online presence. Launch our website.