OpenAI o3 or DeepSeek-R1: Which Is the Better Reasoning Model?

In previous lessons, we compared various aspects of DeepSeek models against other competitors, including OpenAI, Gemini, Llama, and Mistral models. In this lesson, we will conduct our own experiments, testing DeepSeek’s R1 and OpenAI’s o3-mini (high)—currently among the best models for coding and reasoning, as shown in our comparisons in the previous lessons.

We will run multiple experiments to evaluate both models in coding, logical reasoning, and STEM-based problem-solving. For each task, we will provide the same prompt to both models and analyze their responses.

Coding

Let’s start with a coding example. We want to create an interactive physics-based animation using JavaScript. The animation will simulate a galaxy of stars moving under the influence of gravity while incorporating dynamic behaviors such as merging, color blending, and supernova explosions.

The prompt is given below:

Prompt:

Generate a JavaScript animation that should simulate a galaxy of stars moving in a gravitational field inside a container with the following features:

• Randomly placed stars with different masses and colors (white, blue, yellow, green, and red)

• Gravity simulation: Stars attract each other based on a simple Newtonian gravity model

• Star merging: If two stars get close enough, they merge into a larger star, blending their colors using additive color mixing

• Supernova effect: When a star reaches a certain mass threshold, it explodes into multiple smaller stars

• Smooth physics updates with realistic-looking gravitational motion

First of all, in terms of time, o3-mini-high took around 30 seconds to generate a response, whereas DeepSeek-R1 took almost 6 minutes. R1 kept on thinking and rethinking about the prompt. The slow response might frustrate some users.

As can be seen by running the code, the generated code does almost exactly what was asked in the prompt. This generated JavaScript code creates an interactive simulation of a galaxy where stars move under gravitational forces, merge when they collide, and explode into supernovae when they become too massive. The stars are also sometimes seen revolving around each other due to gravitational force and change paths when two stars come close, but not close enough to merge.

The code generated by DeepSeek takes a different approach. As observed, the stars are much larger, and seem to move much slowly, and when merged, they all seem to converge to color white. This is due to the different approaches taken by both codes.

In the code generated by o3-mini-high, when two stars merge, their color is blended using a mass-weighted average rather than a simple addition. This ensures that the resulting color realistically represents ...