MIN_VALUE

Smallest positive non-zero value a half-precision float may have.