Fuzzy systems have recently attracted the attention of researchers and various methods based on fuzzy logic systems have been presented. This paper presents a novel adaptive neuro fuzzy inference system based on interval Gaussian type-2 fuzzy sets in the antecedent part and Gaussian type-1 fuzzy sets as coefficients of linear combination of input variables in the consequent part. The capability of the proposed ANFIS2 for function approximation and dynamical system identification is remarkable. The structure of ANFIS2 is very similar to ANFIS but in ANFIS2 a layer is added for purpose to type reduction. An adaptive learning rate-based backpropagation with convergence guaranteed is used for parameter learning. Finally, the proposed ANFIS2 are used to control of a flexible link robot arm. Simulation results shows the proposed ANFIS2 with Gaussian type-1 fuzzy set as coefficients of linear combination of input variables in the consequent part has good performance and high accuracy but more training time.