Evidence regarding the effect of physical effort on cognitive performance is mixed, with some studies suggesting improvements, particularly at low to moderate intensities. The variability in findings is partly due to a lack of standardized protocols and the limitations of behavioral measures in capturing cognitive functions at the process level. In this study, we aimed to investigate how physical effort affects control parameters by combining a conflict task with cognitive modeling, administered at established moderate and heavy intensity domains. To this end, thirty-two participants completed the Eriksen flanker task while cycling at 50, 80(90), and 110(130) Watts for females (males), respectively. During the exercise, we measured heart rate, oxygen uptake, blood lactate, and subjective ratings of physical and cognitive fatigue. Mixed-effects models revealed a significant decrease in reaction times with increasing intensity but no effect on accuracy. Interestingly, fitting the diffusion model of conflict (Ulrich et al., 2015) using the DMCfun package (Mackenzie & Dudschig, 2021), indicated this was driven by a linear decrease in boundary separation, an inverted U-shape in the drift rate parameter but not by peak latency or amplitude. This suggests that participants exhibited reduced response caution at higher intensities, an initial improvement in attentional capture followed by a decline, consistent with the Yerkes-Dodson law (1908), but no differences in automatic activation by the distractors across intensities. Together, our findings offer new insights into the exercise-cognition relationship, highlighting the need for more sensitive indicators obtained through the modeling of behavior.