Experiment. Which Model Can Better Create a Simple Web Page – GPT-4o, DeepSeek-R1, or Grok 3?

Let's conduct an interesting experiment to determine which LLM can better create a simple frontend page with a 2D solar system simulation.

For this test, we will compare GPT-4o, DeepSeek-R1, and Grok 3 (the new model from Elon Musk).

I plan to use the following prompt for this task

Task:

Create a simple webpage using HTML, CSS, and JavaScript that simulates the movement of the solar system. The page should be responsive.

Acceptance Criteria:

? The Sun and all planets should be represented as simple colored circles.

• The Sun should be yellow.
• Earth should be blue.
• Mars should be brown.
• The colors of other planets are up to you.

? The Sun should be positioned at the center of the page.

? Planet Sizes:

• The diameter of Earth should be 20px.
• The diameter of other planets should follow the real-world ratio compared to Earth.

? Planet Distances:

• The distance between the Sun and the first planet should be 30px.
• The distances of other planets should follow real-world ratios.

? Orbital Movement:

• The planets should orbit the Sun.
• Earth's full orbit (one year) should take 1 minute.
• Other planets should move at speeds proportional to their real-world orbital periods.

? Animation Behavior:

• The movement should start automatically when the page loads.
• The animation should run continuously.

Results & Analysis

In the video below, you can see the results I obtained ??