ChatGPT vs Gemini vs Grok vs Deepseek vs Claude

In this post we will compare ChatGPT vs Gemini vs Grok vs Deepseek vs Claude for several uses cases to understand which one suits best for your needs using the available free models.

What we will test are the following things:

• Code generation
• Content generation
• Problem solving

What we will compare are the following aspects:

• Generation Speed
• Code/Content Quality/Plagiarism
• Limitations
• Robustness
• Readability
• Bugs/Issues

As a reminder, all the tests are executed using free models only.

AI Code Generation

I decided to use a short and simple, yet challenging prompt for code generation using python as the desired language since it’s quite popular.

The prompt is asking for a CSV parser script without using any external libraries, so it should be straightforward.

Prompt for all models:

1Create a python script that can parse CSV's, without using an external library.

Results/Comparison

Some models executed and returned a script very fast whilst some models took even 5 minutes to “think” 🥴 on it. All scripts however executed without any errors, which is nice.

Overall Best

In my opinion, the winner between these models is ChatGPT-4o (Think). A balance between decent generation time and output.

• Most practical, especially for “normal” CSV files (no weird newlines inside cells).
• Offers both dictionary (column names) and list output.
• Customizable delimiter, good error handling, and clear feedback.
• Very easy to read, extend, and integrate.
• Works from command line or as an imported function.

If you want to see beautiful code, learn, or use it in your own scripts: ChatGPT-4o (Think) is the best. For most people and most files, ChatGPT-4o (Think) is the winner!

The scripts are available to download and view on my repository here .

AI Content generation

Code generation is by far the only thing AI’s are used for nowadays, as more and more people use AI to generate content. This ranges from blog posts, documents, emails and more.

Below we will test two categories: Email and Academic Style Writing.

Email Generation

For our Email content generation test, I will ask for a simple email pitch with the following prompt:

1Create an email pitch about my new floral shop. I sell cut flowers, often arranged into bouquets or floral designs. I also offer custom arrangements, provides daily or weekly flower deliveries, and may offer services like wedding or events styling.

Results/Comparison

Every AI/model executed ridiculously fast in generating the content so I will not include the Generation Speed column below in the table.

Overall Best

In my opinion ChatGPT-4o (both versions) wins because of:

• Producing a single, ready to use, natural email per file.
• No formatting oddities, no AI artifacts, no excessive length, and high readability.

It wrote the most natural, easy to read, and professional-sounding email. It followed my instructions exactly (one email per file, no extra formatting or AI mistakes), so you can use it’s email pitch right away with just a few personal details added.

The contents generated are available to download and view on my repository here .

Academic Style Writing

It’s no surprise that more and more people are using AI’s to generate academic style writing, however most of the content generated will fail due to plagiarism checks because of either the AI not generating unique content, either because of the prompt input.

For our second test, I will ask for a short essay with the following prompt:

1Create a short essay (max 1000 words) about the evolution of CPU's, using academic style writing and unique content. Do not use content from already existing essays or sources. Include references where appropriate.

Results/Comparison

Overall Best

The winner is Grok 3 in my opinion for academic style writing.

• Uses storytelling and accessible metaphors (“technological odyssey”), making it pleasant to read for non-specialists.
• Walks the reader chronologically through CPU history while covering modern themes (multi-core, specialization, the future).
• Explains key concepts without overwhelming the reader with jargon or dry technicalities.
• You don’t need a deep technical background to follow and enjoy it.

Plagiarism Check

I used the Plagiarism Checker from Grammarly to check the content and see where each AI stands.

You can see the results in the following table:

The winner is clearly Grok 3 even though it’s not quite perfect; you can fix the small issues yourself and have an “award winning” essay 😊 (lol)

AI Problem Solving

A + B Integral Problem

It’s well know already that AI’s have wast amount of computing power and knowledge, but how do they compare to one another?

Let’s use this popular math quiz provided in some high schools:

1A = Integral from 0 to 1 of ex^2
2B = Integral from 1 to e2 of ln(√x)
3Find A + B

Results/Comparison

This is where every AI’s issues will start to shine. Every single AI struggled to provide a ready copy/paste solution, which is the simplest part of the “problem”. The math was done, but it wasn’t actually copiable, so I had to ask several times in several formats until I could somehow save it as .txt files for you.

The one that stands out here is Claude Sonnet 4, which is the winner in my opinion. For a quick numeric result, Gemini is fastest, but for complete clarity and reusability, stick with Claude, DeepSeek, or ChatGPT-4o.

Broken Code Problem

Let’s kick this up a notch and see how smart are AI’s between them, by asking to fix a piece of broken plain C code:

 1#include <stdio.h>
 2#include <stdlib.h>
 3#include <string.h>
 4
 5char* copy_string(const char* src) {
 6char\* dest;
 7strcpy(dest, src);
 8return dest;
 9}
10
11int main() {
12char* original = "Hello, world!";
13char* copy = copy_string(original);
14
15    printf("Copied string: %s\n", copy);
16
17    return 0;
18
19}

What’s wrong with the above code? Let me explain:

• dest in copy_string is used uninitialized, no memory allocated.
• Using strcpy(dest, src) with an uninitialized pointer causes undefined behavior and likely a crash.
• The memory for the copy (if it had been allocated) is never freed, potential memory leak.
• The code prints the copy without checking for success.

With the above code, we formulate the prompt as follows:

 1Fix the following code for me and provide summary of fixes:
 2
 3```
 4#include <stdio.h>
 5#include <stdlib.h>
 6#include <string.h>
 7
 8char* copy_string(const char* src) {
 9    char* dest;
10    strcpy(dest, src);
11    return dest;
12}
13
14int main() {
15    char* original = "Hello, world!";
16    char* copy = copy_string(original);
17
18    printf("Copied string: %s\n", copy);
19
20    return 0;
21}
22```

Results/Comparison

Honestly, every model produced a correct and professional fix.

All models produced a correct and professional fix. Claude 4 Sonnet went above and beyond with edge-case handling and explanation, but all answers are solid and suitable for copy/paste into a C project. No model introduced any new errors.

The codes generated are available to download and view on my repository here .

Conclusion

After testing the latest generation of AI models on three very different tasks like academic essay writing, business email marketing, and hands-on coding, I found that no single AI rules them all. Instead, each model brings its own strengths, quirks, and ideal use cases.

But, Who Wins Overall?

It depends on what you need:

• Ready to automate or build something serious? Go with ChatGPT-4o or Claude 4 Sonnet for code.
• Need friendly, customer focused communication? Grok-3.
• Want to inform and delight readers? Grok-3 is your best friend.

There is no single “best” AI, just the right tool for the right job.

The smartest way to use AI is to match the model to your mission, because as this experiment shows, even the most advanced bots have their own personalities and strengths.

References/Links

• ChatGPT
• Claude
• Deepseek
• Gemini
• Grok

Thank you for taking the time to read my article and please feel free to share it with friends.