I Tested ChatGPT and DeepSeek Side by Side - Here's the Clear Winner
DeepSeek is open-source, but the biases are glaring, and it's just not trained well enough to compete.
You may know, DeepSeek dropped, and within a week, the entire internet lost its mind. Everyone’s talking about how “insane” it is.
Some so-called experts are even saying it’s better than ChatGPT.
I don’t go with the hype. I do results.
So, I put both ChatGPT and DeepSeek to the test — head-to-head, across different levels — to see which one actually delivers.
That’s exactly what I’m breaking down today.
Since DeepSeek is still new, so the servers can get overloaded, making it very slow. So I won’t worry about speed and will just see how well it works.
Note: If you found this post helpful, consider becoming a paid subscriber to AI Made Simple. You'll get more high-quality posts, and your support will help me create even better content.
With that said, let’s get into it.
1. Solving a word puzzle
First test — can these LLMs solve a simple word puzzle from an image?
For that, I took a screenshot from a word puzzle game and asked both ChatGPT and DeepSeek to find the correct word.
Here’s the output from ChatGPT:
And here’s the output from DeepSeek:
You see, DeepSeek wasn’t able to take images, and when I uploaded by writing a simpler prompt, it wasn’t able to find the answer.
The result? ChatGPT crushed it.
Since it can analyze images, it scanned the puzzle and gave me the right answer within seconds.
DeepSeek? Different story.
It can’t process images, so I reworded the prompt to see if it could still figure it out logically. Instead, it went down a rabbit hole of random guesses, overcomplicating things. Even after minutes of processing, it couldn’t find the correct word.
The winner is obvious — ChatGPT.
2. Solving a simple logic problem
ChatGPT had a weird issue before — it struggled with counting letters in a word.
Ask it how many “r’s” are in strawberry, and it would fumble.
But with the latest updates? It thinks now.
So I tested it again, even adding an extra “r” to see if it would catch the difference.
This time, it got it right — no hesitation.
Now, what about DeepSeek? I ran the same test.
And surprisingly, it nailed it too.
Final verdict? Both passed. No clear winner here.
3. Building a game
Now, let’s put these LLMs to the real test — can they actually build something useful?
Since I’m a web developer, I asked both ChatGPT and DeepSeek to write a snake game using HTML, CSS, and JavaScript.
ChatGPT’s Code:
I ran it. No issues. The game worked smoothly.
The best part? No debugging needed from my side.
And now, here’s the DeepSeek’s Code:
I ran it, here’s the output:
And honestly? I liked this version even more. The gameplay felt better, and overall, it just looked cleaner.
Final Verdict? DeepSeek wins this round.
4. Fact check
Here’s where things get interesting.
Unlike coding or logic tests, this is about bias — and DeepSeek has a clear issue.
Since it’s built by China, I wanted to see if it would dodge politically sensitive topics.
So, I asked about Arunachal Pradesh, an Indian state that China falsely claims as its own.
You see, DeepSeek gave me details about other Indian places but completely avoided answering about Arunachal Pradesh.
Now, let’s ask to ChatGPT:
As you can see, there is no bias. It provided a clear, factual answer.
Final Verdict? DeepSeek’s censorship is a serious limitation, and being open-source makes no sense.
And so, ChatGPT is the clear winner.
5. Aptitude test
Now, let’s put these LLMs to the test with real aptitude problems!
I’m talking about real-world math, logic, and reasoning — the kind of challenges you need to crack competitive exams.
First up, I asked ChatGPT to solve a time and work problem.
Within seconds, it gave the correct answer. Impressive.
Now, let’s see if DeepSeek can do the same. I asked it the exact question — and as expected, it nailed it too.
Well, I can’t just trust DeepSeek by asking a simple question. So I threw a different problem at both LLMs.
Here’s ChatGPT’s response:
And here’s DeepSeek’s response:
Final verdict? Both models are smart enough and well-trained to handle aptitude problems with ease.
6. Book knowledge test
Now, let’s test how well these LLMs know about books!
Since I’ve read Rich Dad, Poor Dad, I’ll ask them a question about it and see which gives the best response.
First, I asked ChatGPT to list the six best lessons from the book in bullet points.
Here’s the output:
Pretty solid, right?
Now, it’s DeepSeek’s turn to answer the same question.
After reviewing both responses, I have to say — I like them both!
Let’s wrap up
Look, the internet is full of creators saying that DeepSeek is way better than ChatGPT.
But after running both side by side for a few days, here’s my take:
DeepSeek is better than I first thought. I’ll give it that.
But when they claim they trained their model at a fraction of ChatGPT’s cost? That sounds like complete BS.
If that were true, it would be game-changing.
But I think DeepSeek has just copied a lot from ChatGPT.
And even after using it for a couple of days, I don’t think DeepSeek is better than ChatGPT.
Sure, it’s open-source, but the biases are glaring, and it’s just not trained well enough to compete.
DeepSeek could improve. Maybe. But right now? ChatGPT is still the king.
Hope you like it.
That’s it — thanks.
If you found this post helpful, consider becoming a paid subscriber to AI Made Simple. You'll get more high-quality posts, and your support will help me create even better content.