Wireless Local Area Networks (WLANs, popularly known as Wi-Fi) are widely adopted for their ability to provide high data rates at affordable costs, making them the most popular connectivity solution worldwide. However, the achievable MAC-layer throughput in WLANs is highly dynamic, depending on factors such as received signal strength, modulation and coding scheme (MCS), channel width, transmit power, interference, and retry rate. This complexity often makes analytical methods for throughput estimation inaccurate. On the other hand, machine learning algorithms learn patterns from data of complex environments and can make better predictions. Accurate prediction of WLAN speeds helps network administrators better plan and tune WLAN networks. So in this work, we evaluate machine learning models Random Forest (RF), XGBoost, Multi-Layer Perceptron (MLP), TabNet, and their stacked meta-models. Hyperparameters of these models are tuned for such dynamic data and evaluated for the metrics \(R^2\) score and Mean Average Error(MAE). The RF-based stacked meta-model(MLP+TabNet) achieved the best performance with an \(R^2\) score of 98%) and a mean absolute error (MAE) of 6 Mbps, outperforming other approaches.