As global warming and urban overheating continue to intensify, accurate urban microclimate modeling has become critical for sustainable urban planning. While the Single-layer urban canopy model (SLUCM, a reduced-order surface energy balance model) and ENVI-met (a computational fluid dynamic model) are the two most widely used models, a direct comparison of their performance is missing. This study aims to examine potential biases between SLUCM and ENVI-met using Hong Kong as a case study and provide guidance on model selection for different purposes. Evaluated against pedestrian-level observational data, the results show that both SLUCM and ENVI-met simulate air temperatures reasonably well, with mean absolute errors less than 1.5 °C. However, SLUCM outperforms ENVI-met in simulating relative humidity, which is partially caused by the insufficient representation of sea breeze by both models. To extrapolate SLUCM output to different heights, Monin-Obukhov similarity theory is applied. This leads to large gradient of temperature and humidity in the vertical direction, while ENVI-met simulations yield homogeneous profiles due to explicit modeling of the turbulent mixing. Findings suggest that ENVI-met suits heterogeneous neighborhoods where turbulent mixing is largely regulated by urban morphology, but its accuracy on humidity simulation needs special attention. SLUCM performs reasonably well in simulating air temperature, but it tends to yield large bias in the vertical direction. Based on the findings, we recommend development of enhanced turbulence parameterization for SLUCM, and coupling both models with mesoscale models to better account for the effect of land/sea breeze on urban microclimate in coastal cities.