Using an. i - z dropout criterion, we determine the space density of z ∼ 6 galaxies from two deep ACS GTO fields with deep optical-IR imaging. A total of 23 objects are found over 46 arcmin2, or ∼0.5 ± 0.1 objects arcmin-2 down to ZAB ∼ 27.3 (6 σ), or a completeness-corrected ∼0.5 ± 0.2 objects arcmin-2 down to ZAB ∼ 26.5 (including one probable z ∼ 6 active galactic nucleus). Combining deep ISAAC data for our RDCS 1252-2927 field (JAB ∼ 25.7 and KS,AB ∼ 25.0; 5 σ) and NICMOS data for the Hubble Deep Field-North (J110,AB and H 160,AB ∼ 27.3, 5 σ), we verify that these dropouts have relatively flat spectral slopes, as one would expect for star-forming objects at z ∼ 6. Compared with the average-color (β = -1.3) U-dropout in the Steidel et al. z ∼ 3 sample, i-dropouts in our sample range in luminosity from ∼1.5L* (zAB ∼ 25.6) to ∼0.3L* (z AB ∼ 27.3) with the exception of one very bright candidate at z850,AB ∼ 24.2. The half-light radii vary from 0.″09 to 0.″21, or 0.5 kpc to 1.3 kpc. We derive the z ∼ 6 rest-frame UV luminosity density (or star formation rate density) by using three different procedures. All three procedures use simulations based on a slightly lower redshift (z ∼ 5) V606-dropout sample from Chandra Deep Field-South ACS images. First, we make a direct comparison of our findings with a no-evolution projection of this V-dropout sample, allowing us to automatically correct for the light lost at faint magnitudes or lower surface brightnesses. We find 23% ± 25% more i-dropouts than we predict, consistent with no strong evolution over this redshift range. Adopting previous results to z ∼ 5, this works out to a mere 20% ± 29% drop in the luminosity density from z ∼ 3 to z ∼ 6. Second, we use the same V-dropout simulations to derive a detailed selection function for our i-dropout sample and compute the UV-luminosity density [(7.2 ± 2.5) × 10 25 ergs s-1 Hz-1 Mpc-3 down to zAB ∼ 27]. We find a 39% ± 21% drop over the same redshift range (z ∼ 3-6), consistent with the first estimate. This is our preferred value and suggests a star formation rate of 0.0090 ± 0.0031 M⊙ yr-1 Mpc-3 to zAB ∼ 27, or ∼0.036 ± 0.012 M⊙ yr-1 Mpc -3 by extrapolating the luminosity function to the faint limit, assuming α = - 1.6, Third, we follow a very similar procedure, except that we assume no incompleteness, and find a rest-frame continuum luminosity that is ∼2-3 times lower than our other two determinations. This final estimate is to be taken as a lower limit and is important if there are modest changes in the colors or surface brightnesses from z ∼ 5 to z ∼ 6 (the other estimates assume no large changes in the intrinsic select-ability of objects). We note that all three estimates are well within the canonical range of luminosity densities necessary for reionization of the universe at this epoch by star-forming galaxies.