A new revision of GLM is available. The version improves the support of half-precision floating-point arithmetic for consistency with the overall library.
It's generally a very bad practice to use half-precision based types (half, hvec*, hmat*) because x86 CPU hasn't a native half float support. GPU doesn't support them, the purpose of half float is to store data for the GPUs and reduce memory bandwise.
To get started with GLM, have a look on the OpenGL 3 Samples Pack.