A stable distribution is a probability distribution that remains unchanged under convolution and is characterized by heavy tails and skewness. It is a generalization of the normal distribution and is used to model data with large outliers or extreme events that deviate from the norm, such as financial returns and natural phenomena.