Considering the extensive use of concrete in construction over the last few decades, a robust and reliable method is required to estimate its strength based on mix parameters. The reason behind this is the complex and nonlinear connection between the strength of concrete and its constituent components. Artificial intelligence has emerged as a powerful tool for simulating the nonlinear behavior of complex systems and predicting the response of systems. This study developed and evaluated AI models for predicting the shear strength of steel fiber reinforcement concrete beams, focusing on a metaheuristics-trained TSK fuzzy inference system. In this regard, an enhanced version of GOA is proposed for accurate training of the TSK model. The developed EGOA-TSK model is evaluated in two ways, with some other metaheuristics and AI methods such as ANFIS and MLP. To implement the proposed method in the prediction of shear strength for SFRCB, a dataset with 243 samples is used. The results show that the TSK-CGO emerged as the superior method, achieving the lowest RMSE (0.0791 for training and 0.1195 for testing) and highest R² (0.9977 for training and 0.9933 for testing) due to its hybrid fuzzy-optimization architecture. Meanwhile, the MT-TSK methods demonstrated accurate resilience to predict the shear strength of SFRCB, whereas ANFIS and MLP faltered due to overfitting and structural limitations. This is whileو EGOA-TSK's ability to exploit nonlinear relationships positions it as a reliable tool for strength prediction, even in data-constrained scenarios.