Abstract: High-tech enterprises are the important subject to promote new industrialization at a high level in the new era, and the main driver of new quality productive forces. Their innovation level is an important issue of concern to people in all walks of life. Based on the micro-data of enterprises in the national high-tech zones from 2013 to 2019, the paper constructs a non-parametric production function model, relaxes the assumption that input factors have a linear impact on output, and considers the endogenous problems that may exist when high-tech enterprises make decisions on the input factors. In addition, it uses the two-stage nonparametric method to estimate the production function of high-tech enterprises and calculates the total factor productivity at the enterprise level, thus analyzing their innovation level from the perspective of total factor productivity. The results show that the total factor productivity of start-up high-tech enterprises is relatively low. This means that the improvement of total output is still in the early stage of factor accumulation and mainly dependent on the input of various input factors. What’s more, the total output is more sensitive to the input of capital factors, and there is heterogeneity in the total factor productivity among enterprises in different technical fields. The study finds that the total factor productivity of manufacturing export enterprises is significantly higher than that of non-export enterprises, and there is no significant difference between the total factor productivity of service export enterprises and non-export enterprises. This study provides policy recommendations for the government to promote the development of high-tech enterprises in China.