This function returns the smallest “model number” for the type and kind of X.

This is an inquiry function, so X need not be defined; indeed, it is permitted to be disassociated, unallocated, or absent.

The model for real type is that it may be zero or a signed fraction between 1 and 1/RADIX(X) consisting of DIGITS(X) digits, multiplied by an exponent e with MINEXPONENT(X)≤e≤MAXEXPONENT(X). Thus the smallest model number is the value obtained for the fraction being equal to 1/RADIX(X) (i.e. the first digit 1 and the remaining digits zero) and with e equal to MINEXPONENT(X).

Note that TINY(X) is not necessarily the smallest possible value for X; for example, with IEEE floating-point, gradual underflow provides many smaller values (with less precision). It is, however, likely to be the smallest value for which the full precision of DIGITS(X) of radix RADIX(X) is available.