In this paper, a novel Takagi-Sugeno-Kang fuzzy system is implemented through the introduction of a Broad Learning System. This system not only enhances the accuracy and interpretability of neuro-fuzzy models but also retains the fast computation characteristics of the Broad Learning System. Initially, the traditional TSK fuzzy system is used to partition the input evenly, resulting in fuzzy sets with appropriate labels. In the consequent part of the fuzzy system’s If-Then rules, the Broad Learning System is used in place of the traditional first-order polynomial. The pseudoinverse is employed to solve the optimal weights for the output layer of the BLS, ensuring rapid and efficient computation without the need for iterative optimization. The proposed BLS-based fuzzy inference system was tested on standard benchmark datasets and its performance was compared against various advanced non-fuzzy and neuro-fuzzy approaches. The results demonstrate that the proposed model outperforms other related models, effectively handling high-dimensional data while maintaining the interpretability of fuzzy rules, thus avoiding the computational complexity associated with rule explosion, and showing promising potential for real-world applications.