DEEP SEEK R1: Future of Artificial Intelligence

Ever since Open AI’s own ChatGPT was released in 2022, the development of AI models has been on the rise. Companies like Google released Bard; later Google Gemini, Anthropic packaged Claude AI and many more. 

In 2024, a Chinese research lab built a competitive model in just 2 months at a cost of approximately 6 million dollars; the biggest story that will change the future of tech around the world.

Google and Open AI spent years and a lot of resources to build their AI language models. What makes Deep seek impressive is China used half of the tech used by their counterparts and only a fraction of the cost. China is making it open source meaning anyone can view, modify and use the software. 

On January 21st, 2025. Deep seek released its latest model the R1 following the V3.

Features of Deep Seek R1

Open Source

All deep seek models are Open Source meaning the users explore, modify, and deploy with in certain technical limits. This fosters innovation in the AI community.

Reasoning

Some of the exceptional features of this model is the ability to logically assess complex reasoning and comprehension. It scores more than all the other models; Claude 3.5 and ChatGPT 4.0. This deep seek model excels in reasoning with a score of 91.6 percent.

Encoding

This AI model has advanced encoding capabilities overtaking the products in the market already like Claude and ChatGPT. It swiftly supports text and image encoding.

Maths

Also, this AI model scores accuracy of 90.2 percent in Maths 500; meaning it can easily reason out complex Maths Problems.

Interface

Its interface is quite similar to that of other AI models like Claude and Gemini.

It possesses a deep-think button below the chat. This helps activate the model’s thinking capabilities and helps solve complex problems similar to what we get with chatGPT premium version; difference is this is totally free. It also shows you the thinking process; how it got to the solution.

Web search

It also has a web feature to search the internet for updated information on your question.

Uploading and Summarizing

Deep seek has the capability to upload documents for analysis; a maximum of 50 documents 100 mbs each. 

Prompts to try out Deep Seek

Thinking Ability

Question is “I have three brothers. Each of my brother has two brothers. My sister also has three brothers. How many brothers and sisters are there?”

Correct answer is 5; but some AI models failed to reason it out but Deep Seek got it right. With a thorough show of the process on how it got the answer.

Creative Writing

Question is write a story about “A boy who got lost in space”

Review: Claude 3.5 emerges as the best creative writing model. Deep seek has a long way to go in creative writing.

Web Abilities

In order to assess its ability to search the internet, I sued the prompt “What is the price of the apple stock”. It delivered more than expected.

 Coding

This model has superb coding abilities.

Deep Seek R1 vs V3 relationship

• Both of these API models are available through Deep seek and Hugging FAce.
• Deep seek-R1 has shown superiority in reasoning tasks and coding compared to the later.
• Both models are open-source and their code is available for public use.
• The deep think free feature allows up to 50 messages a day. The cost for Deep-seek R1 is 2.19 dollars per million tokens while V3 is 0.28 per million tokens for output.

What does the future hold?

AS Open AI prepares to release Open AI o3, I am looking forward to see how this growing competition shapes the future of reasoning models. Share you thoughts about the future of AI below and check out our Gaming and Technology category for more AI related articles.