Artificial Intelligence

What is DeepSeek R1? Here’s How To Use It!

A groundbreaking AI has entered the scene, capable of achieving everything that OpenAI’s most advanced systems can achieve. The best part? It’s not only significantly cheaper than OpenAI’s top-tier o1 model but also open-source, making it accessible to all.

DeepSeek R1 is a “model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step.” This approach is explained in the research paper published by the Chinese AI lab, DeepSeek.

This means that there is an AI training method where humans initially provide the model with the desired outputs in the SFT, helping it comprehend what is good and wrong. This paves the way for RL, the next phase in which the model produces a number of outputs, and their quality is evaluated by humans. This procedure is iterated over and over again until the model masters the art of reliably producing good results.

So what’s the deal with installing DeepSeek on a PC? You just need good tools and some technical skills for now.

DeepSeek R1 - how to use it on PC and mobile, plus how to install it locallyDeepSeek R1 - how to use it on PC and mobile, plus how to install it locally

For the time being, DeepSeek R1 is reserved for academic use. Anyone can use it for personal or professional purposes because it is open-source and licensed by the Massachusetts Institute of Technology (MIT).

How to Use DeepSeek AI?

A few pieces of technology can work together to make artificial intelligence (AI) work its magic. To start, the open-source platform Ollama creates a dedicated area on your machine for AI models to execute, eliminating the requirement for cloud computing. Rest assured, your data remains securely in your possession.

From complicated mathematics to real-time decision-making, DeepSeek R1 can handle it all with its 671 billion parameters.

A variety of R1 models are available to you, with parameter counts ranging from 1.5 billion to 70 billion. Everything you need to run them is this:

  • 7B Models: Can run smoothly on a standard 8GB RAM system.
  • 13B Models: Requires at least 16GB RAM for optimal performance.
  • 33B Models: Demands a hefty 32GB RAM or more to function efficiently.

The research paper from DeepSeek states that the cost to run DeepSeek R1 is approximately ₹684 for every million tokens. When comparing OpenAI’s o1 model to other models, the input cost is approximately ₹1,281 and the output cost is an astounding ₹5,130 per million tokens.

Ollama puts the power of the cloud at your fingertips, eliminating the need for subscriptions and ongoing fees.

Step-by-Step Setup for Ollama

The setup is easy.

For Windows Users

  1. Visit the official Ollama website and download the installer for Windows.
  2. Run the installer by double-clicking it and follow the on-screen instructions. Ensure you have at least 4GB of free storage space.
  3. Open Command Prompt and enter the following command:
    $env:OLLAMA_DEBUG=”1″ & “ollama app.exe”  
  4. File locations:
    • Logs and updates: %LOCALAPPDATA%\Ollama
    • Binary files: %LOCALAPPDATA%\Programmes\Ollama
    • Models and settings: %HOMEPATH%.ollama

For Mac Users

  1. Download the installer directly from Ollama’s website, or use Homebrew, a package manager, by typing:
    brew install ollama
  2. Ensure your system is running macOS 11 Big Sur or later.

Exploring DeepSeek

  • To start using DeepSeek, open your terminal and type: ollama run deepseek-r1:8b  
  • This will automatically download and launch the model.
  • Keep in mind that larger models demand more computing power. It’s a good idea to start with smaller models and gradually scale up.

Customizing the Experience

For a tailored interface, you can use Chatbox to tweak settings easily.

  1. Select Ollama in the Chatbox.
  2. Enter http://127.0.0.1:11434 as your API host.
  3. That’s it—no licensing fees, subscriptions, or payments required.

DeepSeek R1 is designed to handle complex tasks like data analysis, coding challenges, and more. With its open MIT license, you have the freedom to customize, build, and use it for personal or professional projects.

Ollama makes AI accessible without the need for advanced technical expertise or expensive infrastructure. Just download, install, and start exploring its potential. The flexibility of the MIT license lets you adapt the tool to your unique needs, whether it’s a personal hobby or your company’s next innovation.

Sponsored
Huma Ishfaq

Share
Published by
Huma Ishfaq
    Sponsored

Recent Posts

WhatsApp’s ‘View Once’ Fails at Privacy—Here’s Why

NEW DELHI: A critical security vulnerability in WhatsApp View Once has been discovered, allowing iPhone…

16 mins ago

WhatsApp Beta Enhances User Experience with Chat Bar Fix and AI Chatbot Tab

After installing the 2.25.3.2 beta version of WhatsApp for Android, users encountered an issue with…

2 hours ago

Etsy Opens Doors to Pakistani Sellers with Payoneer Payments

Etsy is a world-renowned platform that offers a wide variety of exclusive, vintage, and handcrafted…

3 hours ago

Ducky Bhai Loses 290,000 Subscribers Amid Creator Backlash Following Sham Idrees’ Allegations

The long-standing rivalry between popular Pakistani YouTubers Sham Idrees and Ducky Bhai(Saad Ur Rehman) has…

4 hours ago

Marvel’s Spider-Man 2 Now Available on PC: Here Are the System Requirements You Must Know

Marvel’s Spider-Man 2, released on Steam today, January 30th, has set a new benchmark for…

4 hours ago

Punjab Govt Cracks Down on Speeding with New 60km/h Motorcycle Limit

Punjab: The Punjab government has implemented a 60km/h speed limit for motorcycles across the province,…

4 hours ago