Studying cognitive load (CL) presents several challenges, including the need for accurate and generic CL classification models. Previous studies have solely relied on physiological data to measure CL, as in Appel et al. (2021). However, what about subjective CL measure? Thus, we proposed a three-class classification model of subjective CL through five cognitive tasks: N-back, Corsi, Go/No-Go, WCST, and Dual task (Louis et al. 2023). For this conference, we examined whether a model trained on a Task A could accurately predict the subjective CL classes of a Task B. Firstly, the results showed that Corsi was the most effective task for classifying subjective CL based only on performance and complexity levels, achieving 80% accuracy. Moreover, a classification model trained on N-back, WCST, and Dual task could predict the subjective CL classes of Corsi with over 70% accuracy. This perspective would save time in setting up and training a classification model.