Abstract: ⁹⁰Rb is a high-yield short-lived fission product, a highly sensitive detector for low fission burnup, and its half-life is an important nuclear parameter. At present, the ENSDF (Evaluated Nuclear Structure Data File) evaluated value for the half-life of ⁹⁰Rb is (158±5) s (±3.2%). Due to the low energy resolution of early detectors such as Geiger counter, proportional counter and Ge(Li) detector and significant interference of impurity nuclides during the measurement process, the evaluated value is resulted in a high degree of uncertainty, which can’t meet the requirements of low fission burnup analysis. Therefore, it is of practical application value to determine the half-life of ⁹⁰Rb. In this work, the half-life of ⁹⁰Rb was tracked and measured by the reference source method with a better energy resolution coaxial HPGe γ spectrometer. During the measurement, characteristic γ radiations of ⁶⁰Co reference source (1 173.23 keV and 1 332.50 keV) and ⁹⁰Rb determine source (3 534.24, 4 135.51 and 4 365.9 keV) were tracked. In order to avoid the influence of counting correction caused by dead time and pulse pile-up during the measurement, the ratio R between the counting of ⁹⁰Rb γ radiations (3 534.24, 4 135.51 and 4 365.9 keV) and the counting of the two characteristic energy (1 173.23 keV and 1 332.50 keV) of ⁶⁰Co was calculated. By means of weighted linear fitting method, the slope of the ln R-t fitting curve was obtained, and the experimental values of ⁹⁰Rb half-life were calculated. The uncertainty of the experimental values was comprehensively considered, taking into account the uncertainty introduced by linear fitting and statistical counting. Finally, the weighted average of the experimental values was calculated and the final half-life value with a combined uncertainty was obtained as (160.50±0.47) s (±0.29%). In order to test the reliability of the final half-life value, the relative uncertainty of the data points in each fitting curve was obtained by amplifying the difference between the data points in each fitting curve. It can be found that data points with a relative uncertainty within 1σ account for about 70% of the total data points, 2σ for about 95% and 3σ for 100%, proving the reliability of the given half-life value. The final half-life value has lower uncertainties than that of literature values and agrees with the ENSDF values (158±5) s (±3.2%). The results of this work will provide data support for rapid and accurate measurement of low fission burnup in the future.