Synapses of retinal rod photoreceptors involve deep invaginations occupied by second-order rod bipolar cell (RBP) and horizontal cell (HC) dendrites. Synaptic vesicles are released into this invagination at multiple sites beneath an elongated presynaptic ribbon. To study the impact of this architecture on glutamate diffusion and receptor activity, we reconstructed four rod terminals and their postsynaptic dendrites from serial electron micrographs of the mouse retina. We incorporated these structures into anatomically realistic Monte Carlo simulations of neurotransmitter diffusion and receptor activation. By comparing passive diffusion of glutamate in realistic structures with geometrically simplified models, we found that glutamate exits anatomically realistic synapses 10-fold more slowly than previously predicted. Constraining simulations with physiological data, we modeled activity of EAAT5 glutamate transporters in rods, AMPA receptors on HC dendrites, and metabotropic glutamate receptors (mGluR6) on RBP dendrites. Simulations suggested that ∼3,000 EAAT5 populate rod membranes. While uptake by surrounding glial Müller cells retrieves most glutamate released by rods, binding and uptake by EAAT5 influence RBP kinetics. Glutamate persistence allows mGluR6 on RBP dendrites to integrate the stream of vesicles released by rods in darkness. Glutamate's tortuous diffusional path confers quantal variability, as release from nearby ribbon sites exerts larger effects on RBP and HC receptors than release from more distant sites. Temporal integration supports slower sustained release rates, but additional quantal variability can impede postsynaptic detection of changes in release produced by rod light responses. These results show an example of the profound impact that synaptic architecture can have on postsynaptic responses.