The repeated emergence of NADP–malic enzyme (ME), NAD–ME and phosphoenolpyruvate carboxykinase (PEPCK) subtypes of C4 photosynthesis are iconic examples of convergent evolution, which suggests that these biochemistries do not randomly assemble, but are instead specific adaptations resulting from unknown evolutionary drivers. Theoretical studies that are based on the classic biochemical understanding have repeatedly proposed light‐use efficiency as a possible benefit of the PEPCK subtype. However, quantum yield measurements do not support this idea. We explore this inconsistency here via an analytical model that features explicit descriptions across a seamless gradient between C4 biochemistries to analyse light harvesting and dark photosynthetic metabolism. Our simulations show that the NADP–ME subtype, operated by the most productive crops, is the most efficient. The NAD–ME subtype has lower efficiency, but has greater light harvesting plasticity (the capacity to assimilate CO2 in the broadest combination of light intensity and spectral qualities). In both NADP–ME and NAD–ME backgrounds, increasing PEPCK activity corresponds to greater light harvesting plasticity but likely imposed a reduction in photosynthetic efficiency. We draw the first mechanistic links between light harvesting and C4 subtypes, providing the theoretical basis for future investigation.