[edit: I didn't make this thread, looks like the OP got reported, deleted his post, had is account deactivated or deactivated it himself]
They're the same as any other library. Ultimately, the video driver is doing all the work, interfacing with the hardware directly; the library just hides calls to the video driver under a layer of abstraction. It has imaginary conceptions of a "desktop" and "windows" and "widgets". All it really does is tell the graphics driver where to put pixels.
That said, AFAIK the library doesn't usually, or at least, always, do the driver-interfacing itself. On UNIX OSes (Linux, OS X, BSD) and any others that use the X Window System, the programs (X clients) use the library (XLib) to send commands via the network to a server (XServer), and the server does the drawing. This is useful because you can have the server and clients running on different computers, with the hard work being done on one computer (or several) and the graphical rendering on the other. Graphics libraries like Gtk and Qt just wrap XLib with a further layer of abstraction.
I don't know how Windows does it; probably the simpler way of just having the library call the graphics driver directly. On DOS, of course, you would just access the video card yourself or use a library to do so, or maybe a dedicated driver for more complex cards.