AI MatchMaker ✨ Chat GPT vs DeepSeek

Written By M K

This week I compared ChatGPT & DeepSeek for ✨'AI MatchMaker'✨

 

Key considerations:

First off, I would like to point out few, key difference between the two softwares:

  • I have paid version of Chat that I have been using for a while now, therefore, it knows more about me compared to DeepSeek

  • I will be asking exact same questions, provide the same datasets for both platforms

  • I will be using newest newest models for both (4o & deepthinkR1)

  • Images provided represent responds from Chat & DeepSeek in this order

  • no confidential / sensitive data was shared, but the dataset I am using for my data analysis side project, which is avaiable here


Task: Help with data analysis of my current dataset 💯


Questions I asked:

  • " Tell me its strenghts and areas for improvement for this data set "

  • " Any initial conclusions you can draw from the dataset? "

  • " What do you think about my favorite character Sampo? "

With this task, I had to upload the data manualy (copy + paste) as DeepSeek doesn't read files, so to make it more "fair" I followed the same procedure for both softwares.


 

Question 1: " Tell me its strengths and areas for improvement for this data set " 🥊

Both tools pointed out similar strenghts & areas of improvement, however DeepSeek gave more contextual anserws to my question like mentioning what dataset is about - I did not provide it with that information. But, it insists on me, misspleing certain entries (which maybe I am but I geniuenly cannot see the similarity between Rappa & Robin 👀)

Chat called for more categorization, possible for more context and understanding the data itself. However, it seemed to be more general rather than percise with its conclusions. Small, yet important comment the Chat gave me was that the dataset can be used for trend analysis, which the dataset was made for.



 

Question 2: " Any initial conclusions you can draw from the dataset? " 🧐


As DeepSeek is newer model, it new more neuances related to my dataset, its characters and drew conclusions based on what it knew for example: certain characters being newer vs being from the start of the game. It resulted in me getting very nice comments alongside the data like:

"Bronya - Free starter character? High ownership but niche usage."

For ChatGPT, it created categories for characters bsaed on the dataset and trends found within it, similarly to DeepSeek + it reminded me to improve the data accuracy 😭



 

Question 3: " What do you think about my favorite character Sampo? " ✨

Why did I ask this - well, to see more than just the analytical side of the model! Would it be creative, consider how I feel about character and combine it with data? (which shows he is VERY underutilized character)

Chat GPT pulled the information about the character and presented it in clear and straightforward. It started off nicely by remarking character's great personality traits and showcased that despite its great personality, the data does not support him being the MOST popular or as loved. I also really appreciate links to source information.

DeepSeek similarly listed both benefits & disadvantages of investing into my favorite character, nevertheless I have to appreciate the cheeky comment at the end of the analysis summary 🧡

 

Finals thoughts

I'm not gonna lie, I use Chat more often + I utilize paid version which in itself already brings more to the table than its free version counterpart. I assume that DeepSeek would evolve into subscription model soon enough and basic version will not be enough to use on more regular basis just like ChatGPT at its own beginning.

Maybe it is the recency & novelty bias, but I have fun seeing the thought process of DeepSeek and how it tries to figure out who I am, maybe to adapt to me and my preferences. Nevertheless I would try to use DeepSeek a bit more to truly spot the more neuances but so far I had nice experience using it.

What I would do, is to combine both softwares. Chat seems to be neutral and “fair” in assesing data, while DeepSeek adds extra contextual information about the dataset if its necessary and accesible on the internet. As I ran the test on my ‘homebrewed’ dataset on Honkai Star Rail (all publicly accesible), I wasn’t afraid of sharing the information to the ML model.

I will defenitely see which tool is more creative and can pitch the dataset and its information better to fully understand which one use for what, but so far I am positively surprised by the possibilities of the free, DeepSeek model. Let’s see if it will change or if certain features will be limited with introduction of subscription wall or more paid services.

creepy? fun? - you decide! 👀







M K
Next

AI MatchMaker ✨ Notebook LM