Epidemiological studies have proven that extreme temperatures have a significant threat to public health. This study aimed to investigate the association between extreme temperatures and circulatory mortality from January 1, 2014, to December 31, 2016, in Harbin, a city with a cold climate in Northeast China. We set a maximum lag of 27 days to evaluate the hysteresis effects of different temperatures on circulatory mortality using a distributed lag nonlinear model (DLNM). Results indicated that daily mean temperature and circulatory mortality presented approximately an L-shaped, and the cumulative relative risks (RRs) decreased continuously as the temperature increased in both low and high temperatures. Extremely low temperature showed a hysteresis and durability on circulatory mortality, with the largest RR of 1.023 (95%CI: 1.001-1.046) at lag 26, and RR of the cumulative cold effect of 0-27 days was 1.302 (95%CI: 1.160-1.462). The effect of extremely high temperatures presented more acute and intense, with the largest RR of 1.033 (95%CI: 1.004-1.063) at lag 0. RR of the cumulative hot effect of 0-3 days was 1.056 (1.008-1.106). In addition, females were more susceptible to extremely low temperatures, while males were more vulnerable to extremely high temperatures. This study demonstrated that extremely low temperatures have a stronger effect on circulatory mortality than extremely high temperatures in Harbin.