Previous engine data suggest that slower flame propagation in lean-burn engines could be due to slower flame expansion velocity at lean conditions than at stoichiometric combustion. Two classes of model, a quasi-dimensional engine-simulation program and a multidimensional engine-flow and combustion code, were used to study this effect in detail and to assess the capabilities of the models to resolve combustion details. The computed flame-speed data from each program differed somewhat in magnitude, but the predicted trends at various equivalence ratios were quite similar. The trends include: (1) The peak in-cylinder burned-gas temperature decreases by about 300 K as the equivalence ratio is decreased from 0.98 to 0.70. (2) Both the laminar flame speed and the flame-propagation speed, the latter computed from the time derivative of flame radius, decrease with decreasing equivalence ratio. (3) The turbulent burning speed, defined as the ratio of specific fuel-burning rate to the product of the flame frontal area and unburned-mixture density, is relatively insensitive to equivalence-ratio variations at the same flame-radius position. The previous experimental finding that the reduction in flame-propagation speed with decreasing equivalence ratio is caused mainly by the lower thermal-expansion speed, calculated by subtracting the turbulent burning speed from the flame-propagation speed, was confirmed. This is a consequence of lower burned-gas temperature for the lean case. Regarding the reliability of the models to resolve the combustion details, limitations of the models are identified and discussed in detail.