We use six years (2003–2008) of Optical Gravitational Lensing Experiment-III microlensing observations to derive the survey detection efficiency for a range of planetary masses and projected distances from the host star. We perform an independent analysis of the microlensing light curves to extract the event parameters and compute the planet detection probability given the data. 2433 light curves satisfy our quality selection criteria and are retained for further processing. The aggregate of the detection probabilities over the range explored yields the expected number of microlensing planet detections. We employ a Galactic model to convert this distribution from dimensionless to physical units, α/au and M⊕. The survey sensitivity to small planets is highest in the range 1–4 au, shifting to slightly larger separations for more massive ones.