The relative timing of different sensory input signals is an important factor in multisensory integration and perception. When multiple forms of sensory stimuli arise from a single source, the degree of synchrony in the arrival of these stimuli can provide distance cues. The difference between the velocities of light and sound introduces a distance dependent lag for auditory signals with respect to visual information. Audio-visual delays therefore carry information about the distance to the stimulus source. Here we investigated the representation of audio-visual delays in the mouse superior colliculus (SC), a midbrain area that represents the location of visual and auditory targets topographically. Neuronal activity was recorded with Neuropixels probes in awake animals presented with visual and auditory stimuli with staggered onset times (ranging from 0 to 100ms). We found that 30% of the recorded neurons are modulated by both visual and auditory stimuli. These neurons exhibit a broad range of audio-visual delay preferences and exhibit nonlinear multisensory interactions. We used a random forest classifier to decode audio visual delays from population responses. We found differences in decoding performance across the anterior-posterior and the medio-lateral axes of the SC, with the highest accuracy in the posterior-medial region. This result suggests that there is a functional specialization of multisensory integration across anatomical regions of the SC. While separability of audio-visual delay representations is enhanced in the upper visual field, perceptual binding is favoured in the lower visual field.