In IEEE arithmetic, a nonzero floating point number with a biased exponent of zero. The subnormal numbers are those between zero and the smallest normal number.