The function of Fintech is preventing and resolving financial risks, and the ability of small and medium-sized commercial banks to apply Fintech to resolve non-performing loan risks deserves attention. This paper uses as a sample the micro-survey data of 432 branches of the city commercial banks in Beijing from 2005 to 2022 and constructs an econometric model of the risk reduction effect of Fintech on non-performing loans.The findings are as follows.Firstly, Fintech inputs can significantly reduce the risk of non-performing loans. For every 1% increase in IT personnel inputs, software inputs, and hardware inputs, the non-performing loan ratio will reduce by 0.091%, 0.055%, and 0.024%, respectively. In other words, IT personnel inputs contribute the most to the reduction of non-performing loan risk, followed by software inputs.Fintech has a certain lag effect in alleviating the risk of non-performing loans in banks and tends to strengthen over time.Secondly, Fintech inputs has an indirect effect on banks' performance by reducing non-performing loan risk, And there is a positive feedback loop between Fintech inputs, non-performing loan risk mitigation, and improving performance. This effect is not significant in the current period but significant when lagged by three periods.Thirdly, Fintech inputs has a significant inhibitory effect on non-performing loans loans of concern and subprime. For every 1% increase in inputs in IT personnel, software, and hardware, concern will decrease by 0.263%, 0.358%, and 0.107%, respectively, while subprime will decrease by 0.115%, 0.216%, and 0.057%, respectively. Software and personnel inputs have a greater inhibitory effect on the above two types of loans than hardware inputs. Fourthly,mechanism analysis shows that data governance application, compliance enforcement, and internal control enforcement have a moderating effect on Fintech input's impact on banks' non-performing loan risk.