BK7 glass is an important material commonly used is high end optics. However, due to its brittle and hard nature, it is regarded as a difficult to cut material. Rotary ultrasonic machining (RUM) has high potential to machine brittle materials like BK7 glass. However, reliable models are requited for predicting the machining performance of RUM of BK7 glass. In this paper, adaptive neuro-fuzzy inference system (ANFIS) models have been used for modeling and predicting the surface roughness and exit chipping size of the holes drilled in BK7 glass by employing RUM. This has been achieved by studying the effect of major RUM input variables such as spindle speed, feed rate, and ultrasonic vibration on the output responses including surface roughness and exit chipping size which are directly linked with the quality of the drilled hole. Full factorial design of experiment approach has been used to generate the experimental data. The results of the ANFIS models were validated with the experimental data and were also compared with the regression based models results. The results showed that the ANFIS model can achieve more accurate results with lower mean absolute percentage error.