Soil CO2 efflux (Fsoil) represents a significant source of ecosystem CO2 emissions that is rarely quantified with high-temporal-resolution data in carbon flux studies. Fsoil estimates can be obtained by the low-cost gradient method (GM), but the utility of the method is hindered by uncertainties in the application of published models for the diffusion coefficient. Therefore, to address and resolve these uncertainties, we compared Fsoil measured by 2 soil CO2 efflux chambers and Fsoil estimated by 16 gas transport models using the GM across 1 year. We used 14 published empirical gas diffusion models and 2 in situ models: (1) a gas transfer model called “Chamber model” obtained using a calibration between the chamber and the gradient method and (2) a diffusion model called “SF6 model” obtained through an interwell conservative tracer experiment. Most of the published models using the GM underestimated cumulative annual Fsoil by 55% to 361%, while the Chamber model closely approximated cumulative Fsoil (0.6% error). Surprisingly, the SF6 model combined with the GM underestimated Fsoil by 32%. Differences between in situ models could stem from the Chamber model implicitly accounting for production of soil CO2, while the conservative tracer model does not. Therefore, we recommend using the GM only after calibration with chamber measurements to generate reliable long-term ecosystem Fsoil measurements. Accurate estimates of Fsoil will improve our understanding of soil respiration's contribution to ecosystem fluxes.