This research examines the pull-out characteristics of twisted twin fibers within concrete employing advanced soft computing methods. The study highlights the necessity for precise predictive models in fiber-reinforced concrete scenarios, considering the intricate interactions between fibers and their surrounding matrix. Artificial Neural Networks (ANN) and Gene Expression Programming (GEP), were created to forecast the pull-out energy needed for fiber extraction. A detailed dataset comprising 228 experimental samples was used, and various models were trained, including 51 ANN designs and 10 GEP configurations. For the first time, a mathematical formula was established using GEP to estimate pull-out energy, showcasing high accuracy with minimal error margins. The ANN model, especially the one utilizing a log-sigmoid activation function, achieved the highest correlation coefficient (0.995), surpassing the GEP model, which also demonstrated a robust correlation (0.98). Sensitivity analysis indicated that compressive strength had the most substantial effect on pull-out energy, accounting for 18.5% of the observed variance. The results offer a new and precise method for predicting fiber pull-out energy, improving the comprehension of fiber-matrix interactions in cement-based materials. Future investigations should aim to broaden the dataset and examine additional fiber shapes to enhance predictive accuracy.