Globally, the life expectancy of the population has been continuously improving over the years due to healthcare and socioeconomic advancements. The rapid increase in life expectancy over the last few decades leads to an ageing population as the population lives longer. With the rise of elderly population in the society, insurance companies and pension funds need to deal with longevity risk, which is the risk of incurring greater pay-out ratios than projected as life expectancies exceed pricing assumptions. Hence, accurate mortality modelling and projection are of key interest to insurance companies, pension providers and government to minimize such risks. This study will focus on the modelling of mortality rates in Malaysia based on three main ethnic groups, namely Malay, Chinese and Indian using data from Abridged Life Tables for a 20-year period (2001-2020) obtained from Department of Statistics Malaysia. Mortality rates for six subpopulations (Malay male, Malay female, Chinese male, Chinese female, Indian male and Indian female) under 18 age groups will be modelled using three stochastic mortality models, i.e. Lee-Carter model, Hyndman-Ullah model and Augmented Common Factor model. We conclude that Hyndman-Ullah model has the best fit for past mortality rates with the lowest values of goodness-of-fit using Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Future research can be conducted by using Hyndman-Ullah model to forecast mortality rates in Malaysia based on age, gender and ethnic groups, which can be then applied in updating pension and annuities calculations on the existing and new contracts to minimize financial losses arising from longevity risk.