Gestational Diabetes Mellitus (GDM) is a serious condition that can develop during pregnancy. It can introduce complications that will threaten both mother and child. Detecting it in earlier stages is key to preventing serious health risks that can occur in later stages of the pregnancy cycle. Although there have been advances in healthcare, current methods often fail to capture glucose fluctuations in real time. Our contribution in this study is an AI-driven model that aims to predict GDM in the early stages of pregnancy using real clinical data gathered from the University of Sharjah Center for Diabetes. The private dataset we received consisted of demographic, biochemical, and pregnancy-related information that, after cleaning, were reduced to five features: Age, Body Mass Index (BMI), Oral Glucose Tolerance Test (OGTT) Fasting, OGTT 1-h (OGTT_1h), and OGTT 2-h (OGTT_2h). Among the tested models, Extreme Gradient Boosting (XGB) achieved the highest classification performance with an ROC-AUC score of 0.87 and an F1-score of 0.90, along with a recall percentage of 90%, which outperformed other models. The recall is especially important, as this study is for identifying true GDM cases. Although precision may drop slightly when the recall is high, this trade-off is acceptable if the clinical cost of missing true GDM cases outweighs the cost of false alarms. Correlation Analysis identified OGTT and BMI as the most influential biomarkers associated with GDM risk. This study aims to showcase the performance ability of AI in reducing pregnancy complications in relation to Gestational Diabetes.