The adjustment of interest rate by central bank brings the benchmark interest rate risk． The change of benchmark interest rate certainly will affect the financial asset pricing and risk premium． In this paper，au-toregressive model ( AＲ) is used to measure the time-varying frequency of interest rate jump and determine the probability of the benchmark interest rate jump． Time and amplitude of benchmark interest rate jump are cal-culated on two conditions respectively obeying the gamma distribution and the normal distribution． The risk premium of benchmark interest rate jump is calculated according to probability，time and amplitude． Then we establish the risk premium model based on time-varying benchmark interest rate jump and empirical study is carry out based on the 7-day repo rate data from Shanghai Stock Exchange． Innovation and contributions of this paper: First，using autoregressive model to calculate the frequency，probability and risk premium of time-var-ying interest rate jump，we reveal the law of dynamic changing jump frequency and the impact of historical in-terest rate jump on future interest rate jump． In the existing research，probability of interest rate jump is calcu-lated based on constant jump frequency which can not truly reflect the jump frequency，leading to inaccurate calculation of jump probability and risk premium． Our research makes up for the deficiency． Second，this pa-per demonstrates that the constant jump frequency in the existing study is only a special case of our model when the parameter both ρ and γ equal 0． Third，using the probability，time and amplitude of interest rate jump to determine risk premium，we solve the problem of the risk compensation calculation of benchmark in-terest rate jump．