Increasingly stringent regulations on NOx emissions are enforced by governments owing to their contribution in the formation of ozone, smog, fine aerosols, acid rains, and nutrient pollution of surface water, which affect human health and the environment. The design of high-efficiency, low-emission combustors achieving these ever-decreasing emission standards requires thermochemical mechanisms of sufficiently high accuracy. Recently, a comprehensive set of experimental data, collected through laser-based diagnostics in atmospheric, jet-wall, stagnation, premixed flames, was published for all isomers of C1–C4 alkane and alcohol fuels. The rapid formation of NO through the flame front via the prompt (Fenimore) route was shown to be strongly coupled to the maximum concentration of the methylidyne radical, [CH]peak, and the flow residence time within the CH layer. A proper description of CH formation is then a prerequisite for accurate predictions of NO concentrations in hydrocarbon–air flames. However, a comparison against the Laser-induced fluorescence (LIF) experimental data of Versailles, P., et al. (2016, “Quantitative CH Measurements in Atmospheric-Pressure, Premixed Flames of C1–C4 Alkanes,” Combust. Flame, 165, pp. 109--124) revealed that (1) modern thermochemical mechanisms are unable to accurately capture the stoichiometric dependence of [CH]peak, and (2) for a given equivalence ratio, the predictions of different mechanisms span over more than an order of magnitude. This paper presents an optimization of the specific rate of a selection of nine elementary reactions included in the San Diego combustion mechanism. A quasi-Newton algorithm is used to minimize an objective function defined as the sum of squares of the relative difference between the numerical and experimental CH–LIF data of Versailles, P., et al. (2016, “Quantitative CH Measurements in Atmospheric-Pressure, Premixed Flames of C1–C4 Alkanes,” Combust. Flame, 165, pp. 109--124), while constraining the specific rates to physically reasonable values. A mechanism properly describing CH formation for lean to rich, C1–C3 alkane–air flames is obtained. This optimized mechanism will enable accurate predictions of prompt-NO formation over a wide range of equivalence ratios and alkane fuels. Suggestions regarding which reactions require further investigations, either through experimental or theoretical assessments of the individual specific rates, are also provided.