The efficiency of C₄ photosynthesis in Zea mays, Miscanthus x giganteus and Flaveria bidentis in response to light was determined using measurements of gas exchange, ¹³CO₂ photosynthetic discrimination, metabolite pools and spectroscopic assays, with models of C₄ photosynthesis and leaf ¹³CO₂ discrimination. Spectroscopic and metabolite assays suggested constant energy partitioning between the C₄ and C₃ cycles across photosynthetically active radiation (PAR). Leakiness (φ), modelled using C₄ light‐limited photosynthesis equations (φ_mod), matched values from the isotope method without simplifications (φ_is) and increased slightly from high to low PAR in all species. However, simplifications of bundle‐sheath [CO₂] and respiratory fractionation lead to large overestimations of φ at low PAR with the isotope method. These species used different strategies to maintain similar φ. For example, Z. mays had large rates of the C₄ cycle and low bundle‐sheath cells CO₂ conductance (g_bs). While F. bidentis had larger g_bs but lower respiration rates and M. giganteus had less C₄ cycle capacity but low g_bs, which resulted in similar φ. This demonstrates that low g_bs is important for efficient C₄ photosynthesis but it is not the only factor determining φ. Additionally, these C₄ species are able to optimize photosynthesis and minimize φ over a range of PARs, including low light.