We have been developing the PPP algorithms based on our GR (GNSS Regresion) models with applying CLAS (Centimeter Level Augmentation Service) data from Japan Quasi-Zenith-Satellite-System (QZSS)[1]- [7]. In this paper, we examine the accuracy and convergence of positioning estimators applying the fixed-point Kalman smoother. Throughout the numerical experiments the total 3D-RMS error by applying CLAS and fixed-point Kalman smoother is approximately 0.15 [m].