Over the last one week, the entire AI industry has been buzzing over the release of Deep Seek R1. I initially thought, what’s all the fuss about? However, I spent some time yesterday playing with Deep Seek and doing some research about it, and yes, I also had my Eureka moment—I finally get the hype. Here are some of my initial thoughts.
Ridiculously Good LLM for FREE!!!!!
The first thing that got my attention was that Deep Seek is open source!!! Nahhhh, this is game-changing! If you don’t understand what this means, let me break it down. Prior to now, the best AI LLM models were created by OpenAI (in my opinion) , and while OpenAI initially started off by making some of their initial projects open source, they eventually switched to a for-profit model. For example, ChatGPT today is currently not open source, and because of this, they offer different price tiers for accessing different capabilities. I, for example, am a Plus user ($20 per month), but from my initial interaction with Deep Seek, I am not sure how much longer I will be. :). Could this potentially disrupt Open AI’s monetization plans ?
Brilliant with Reasoning and Coding
Another thing I tried yesterday was testing out its reasoning and coding capabilities, and in my opinion, so far, Deep Seek is the clear winner. I am currently learning how to code with Python, and ChatGPT has been a wonderful tutor. However, Deep Seek equally performed well with breaking down Python syntax for me in a language I can understand.
I also asked them to write HTML code for a Tetris game that I could play offline, and while they both did a decent job with this, I liked the results from Deep Seek much better. The Tetris game it produced had a better overall feel.
I then uploaded the generated HTML lines of code to both Chat GPT and Deep Seek to ask which code was better. While Deep Seek immediately told me its code was better at the first prompt, Open AI’s Chat GPT first beat around the bush comparing different aspects but when forced to choose one decided that overall Deep Seek was the better code. And Deep Seek produced this in few lines of code. However I do not think this is an apple for apple comparison as I will repeat this test when Open AI’s O3 model is made public.
Other Thoughts
From my first experimentation, overall I still prefer ChatGPT’s user interface in general. It feels a little more premium, and I love the fact that they have projects where I can organize different prompts. However, is this something I am willing to pay $20 a month for?
Also, I enjoy generating AI images for presentations and other things. Currently, Deep Seek is unable to do that, so ChatGPT is still a clear winner here. However, I can equally generate free AI images using Meta AI on my phone, so it might not justify the $20 a month fee for me.
Over the next week, I intend to test out the RAG (Retrieval-Augmented Generation) functionalities of both models, I am curious to see how they both compare in this area, and maybe I will write a blog about that too.
Leave a Reply