This research is concerned with the application and comparison of mathematical models for predicting deaths due to fire accident in Taraba State. The reason for this research is because it was observed that whenever fire accident occurs, it claims lives and properties. Data were collected for the period of ten (10) years from 2005 to 2014 from Fire Service Board Jalingo. Taraba State, Nigeria. Various models were applied, such as Poisson distribution, Exponential distribution and Negative Binomial distribution. The aim of the application of these models was to determine the best model for predicting deaths due to fire accident. It was observed that the Poisson distribution result summed to approximately 0.6 with 0.4 deviation from the exact probability of 1.The Exponential distribution summed to approximately 0.9, with 0.1 deviation from the exact probability 1. The Negative Binomial distribution summed to approximately 0.8 with 0.2 deviation from the exact probability 1. The results showed that all the models are appropriate for modeling death due to fire accident. However, exponential distribution seems to be the best model in comparison with Poisson and Negative Binomial distributions in this work.