In this paper, we obtain a non-asymptotic bound for Gaussian approximations for centred high-dimensional non-degenerate U-statistics over the class of hyper-rectangles via exchangeable pairs and Stein’s method. We improve the upper bound of the convergence rate from n^(-1/6) in Chen (2018) to n^(-1/4) up to a polynomial factor of log(d) under the same conditions, where n is the sample size and d is the dimension of the U-statistic. Convergence to zero of the bound requires log(d)=o(n^(1/7)) in Chen (2018), this requirement on d is weaken in this paper by allowing log(d)=o(n^(1/5)).