<p>High-performance concrete (HPC) outperforms ordinary concrete in compressive strength, durability, and resistance, making it essential in major infrastructure projects. Accurate prediction of its compressive strength (CS) is critical for structural safety, but traditional methods are often slow and unreliable. This study introduces a hybrid machine learning (ML) framework to improve prediction accuracy. It combines RBF and stacking regression (STACKR) with advanced optimization techniques—the Equilibrium Slime Mould Algorithm (ESMA) and the Population-based Vortex Search Algorithm (PVSA). The approach yields four hybrid models—RBPV, RBES, STPV, and STES—designed to integrate ML with metaheuristic optimization for adaptive learning and stronger predictive performance. The whole 460 HPC mix designs into training, validation, and testing phases comprise a whole dataset of experiments. Numerical data from the testing phase depict that STES, the best performing model, resulted in excellent performance with RMSE of 4.435, R2 of 0.969, MSE of 19.673, MDAPE of 6.041, and NSE of 0.967. Secondly, the RBES model presented good performance with RMSE equal to 5.512, R² equal to 0.956, MSE equal to 30.385, Median MAPE equal to 6.056, and NSE equal to 0.949. Thus, the many promises that ML-integrated hybrid modeling has to offer in high-fidelity CS predictions in HPC will certainly impress. It enhances fidelity in predictions and hastens evaluation, hence making it practical yet scalable for real-world construction situations.</p>