MIN_NORMAL

Smallest positive normal value a half-precision float may have.