We present the first galaxy counts at 18{mu}m using the Japanese AKARI satellite's survey at the North Ecliptic Pole (NEP), produced from the images from the NEP-Deep and NEP-Wide surveys covering 0.6 and 5.8deg^2^, respectively. We describe a procedure using a point source filtering algorithm to remove background structure and a minimum variance method for our source extraction and photometry that delivers the optimum signal to noise for our extracted sources, confirming this by comparison with standard photometry methods. The final source counts are complete and reliable over three orders of magnitude in flux density, resulting in sensitivities (80 per cent completeness) of 0.15 and 0.3mJy for the NEP-Deep and NEP-Wide surveys, respectively, a factor of 1.3 deeper than previous catalogues constructed from this field. The differential source counts exhibit a characteristic upturn from Euclidean expectations at around a milliJansky and a corresponding evolutionary bump between 0.2-0.4mJy consistent with previous mid-infrared surveys with ISO and Spitzer at 15 and 24{mu}m. We compare our results with galaxy evolution models confirming the striking divergence from the non-evolving scenario. The models and observations are in broad agreement implying that the source counts are consistent with a strongly evolving population of luminous infrared galaxies at redshifts higher than unity. Integrating our source counts down to the limit of the NEP survey at the 150{mu}Jy level we calculate that AKARI has resolved approximately 55 per cent of the 18{mu}m cosmic infrared background relative to the predictions of contemporary source count models.
