Personally, I would go with whatever dimension works best for you and/or you are most comfortable with, especially if you are not under a time constraint.
That said, having done several games in 2D and many in 3D, there is definitely differences in the creation process.
I find that making 3D game assets generally take longer to create than 2D sprites. For 3D models you need to create the 3D model (or edit a reference/base-mesh), UV wrap it, texture it, create an armature and rig it, and then make any animations you might need.
There is a lot more detail you can put into a 3D model than you can fit into a 2D sprite, especially if your 2D sprites are lower resolution. I find animating 3D models easier than making 2D animation, but that might just be me.
2D game assets are faster to make, but depending on the resolution and how detailed/complicated you want to make them, it can take just as long. 2D game assets generally require making a sprite sheet if you want to do animation, or cutting the image up into separate parts for bone based animation. This can make things interesting depending on the art style.
Overall, I think 2D game assets can have a wider variety of really cool art styles, because it is easier to control how things look and move within the game. Also, depending on which perspective you take (like isometric vs. top-down, for example), you can create the illusion of having a third dimension in your games and do some cool things that would be very hard and time consuming to replicate in 3D.
I personally don't have much skill in the 2D arts, so for me even simple sprites can take a long time. Because of this, I generally don't do 2D game assets, as I find I personally like the 3D creation process a little better.
One great advantage with 2D sprites is that it is much easier to refine and tweak the sprites as you go through the development process in comparison to 3D art. With my own game assets, I find I will tweak things here and there as I am actively developing mechanics because it generally only requires moving a few pixels around to get the desired result.
But as I said above, I would go with whatever works best for you and your project.
As @Megalomaniak mentioned, it is highly recommenced to make prototype/placeholder art first so you can work on the gameplay, and then swap out the prototype/placeholder art once you are further in development.
I rarely make prototype art if I am working under a tight time limit (like a game jam), but for longer projects I almost always end up changing the art at least one or twice, generally more for important elements. If you can take the time, it can really help to just make enough art to prototype your gameplay and get that going, so then you can tailor the finished art so it fits with both the visual and gameplay requirements of your game.
I should mention that everything I have written above is based on my own experience! Others may experience things completely differently based on how they do things, what projects they work on, their skills, and so on.
Hopefully this helps a bit :smile:
