OpenGL ES is a much smaller API in terms of entry points (functions) and functionality. OpenGL ES 2.0 is based on OpenGL 2.1 but with all legacy functions removed (i.e. no immediate mode rendering, no display lists, no fixed-function functionality.) OpenGL ES 3.0 will bring newer functionality equivalent to OpenGL 3.x (FBOs by default, more flexible shaders), with the possible exception of geometry shaders (which are much less useful than they sound, esp. in the mobile space.)
Apart from these differences, there are OpenGL ES versions for devices without floating-point hardware and for devices with increased stability requirements.
In short, OpenGL ES is closer to what OpenGL could have been if it wasn't burdened by backwards compatibility.
So they're pretty much same powerful and flexible?
OpenGL ES is easier to learn and develop with?