<p>The paper presents a new investigation on the bearing capacity of rectangular foundations resting on a cohesive-frictional slopes. It is also the first time the 3D finite element limit analysis combined with the hybrid machine learning models has been suggested to analyze the bearing capacity of rectangular foundation rest on cohesive-frictional slopes. The combined effect of six design parameters, including slope angle, shape ratio, depth ratio, setback ratio of the footing, internal friction angle, and cohesion ratio of the soil mass on bearing capacity and failure mechanism, is implemented. The discussion is implemented based on the design chart showing the relationship between the bearing capacity factor and design parameters. The impact of design parameters on failure mechanism also provides several insights into the bearing capacity analysis of rectangular rest on the slope for practical engineering. Finally, the eXtreme Gradient Boosting (XGBoost) technique is used to analyze the correlation between the investigated design parameters and the bearing capacity factor. A novel hybrid machine learning framework is proposed, which integrates an optimization algorithm, specifically Ant Colony Optimization (ACO), with the XGBoost algorithm to improve its performance. The accuracy of the ACO-XGBoost model (R² = 97.84%) is proposed as the most optimal hybrid XGBoost model for predicting the bearing capacity of a rectangular footing on a cohesive-frictional slopes. Furthermore, the importance analysis results show that friction angle is the most critical design parameter, followed by cohesion ratio, setback ratio, shape ratio, slope angle, and depth ratio.</p>