A floating-point type is a data type in computing that approximates real numbers within a finite number of bits. It stores values using a significand and an exponent, enabling a vast range but with inherent limitations in precision. This design balances the representation of very large and very small fractional numbers.