I want to make a winning decision with OpenGL.

In the case of 2D games, we define squares that cover characters, and I think we determine roughly whether or not the squares overlap each other.
After that, I think we will make a detailed collision decision using the decision that matches the character's shape.

There is a way to define a cube AABB (Axis-Aligned Bounding Box) that encloses characters in 3D, and to determine if the cubes overlap each other.
After that, AABB repeatedly determines the collision for each part of the character, and eventually drops the collision determination in polygons, but depending on the accuracy of the collision determination, it will change how far you go.

I can't explain the detailed formula here, so please consider purchasing a book.
What I use for studying and developing is
Ms. O'Reilly's https://www.oreilly.co.jp/books/9784873113777/
That's it.

"By the way, the reason why square judgment is often used is ""the least computational cost."""

Is that the answer you are looking for?