Recent advancements in MRI hardware, including ultra-high magnetic field scanners (≥7T) and MR data acquisition methods, have enhanced functional imaging techniques, allowing for the detailed study of brain function, particularly at the mesoscopic level of cortical organization. This has enabled the measurement of blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) signal changes across cortical depth in the human brain, facilitating the study of neuronal activity at laminar level. In order to better understand the generation of cortical depth-dependent BOLD signals, biophysical modeling and computational simulations permit the characterization of the impact of vascular architecture, as well as the biophysical and hemodynamic effects at the mesoscopic level. In this study, we employed four realistic 3D vascular models that mimic the human cortical vascular architecture and simulated various vessel-dependent oxygen saturation and hematocrit states, aiming to characterize the intravascular and extravascular contributions to gradient-echo (GE) and spin-echo (SE) BOLD signal changes across human cortical depth at 7T. We found that differences in the local vascular architecture between the four models, away from the pial surface, do not significantly influence the shape and amplitude of BOLD profiles. This implies that signal profiles within a cortical region of a given angioarchitecture can be averaged within a given layer without introducing substantial errors in the results. The findings futher reveal that in deeper laminae, relative relaxation rates for both GE and SE decrease linearly with increasing oxygen saturation levels, with GE showing a stronger effect. In contrast, the top lamina shows a non-linear behavior due to large vessel contributions, particularly venous, with GE displaying higher relaxation rates (4-8 times larger dependent on oxygen saturation levels) than SE. Relative BOLD signal changes also follow linear trends in deeper layers, with GE peaking at ~8% and SE at ~4%, reflecting the higher microvascular specificity of SE. However, SE does not fully eliminate large vessel contributions at the pial surface, where diffusion effects and vessel architecture play a role. Hematocrit levels linearly change the BOLD signal amplitude and significantly influence laminar contributions across cortical depth and imaging techniques. While GE signals are dominated by extravascular effects, SE retains notable intravascular venous contributions at high oxygen saturation levels, which is particularly relevant in experiments involving controlled vascular oxygenation, that is, gas challenges. These results underscore how vascular features, hematocrit, and biophysical interactions shape cortical depth-dependent BOLD signals and their specificity in ultra-high field imaging.