While the cortex is undoubtedly the most sophisticated complex network, surprisingly little is known about its topological organization. Performing brain-wide retrograde tracing experiments in the macaque, we generated a consistent database of interareal connections with weights (projection densities), and distance information. This network's dense connectivity (66%) shows that it is neither a sparse small-world, graph nor scale-free. Local connectivity accounts for 79% of labeled neurons, subcortical afferents for 1.3%. Link weights, highly characteristic across animals, follow a heavy-tailed lognormal distribution over 5 orders of magnitude and decay exponentially with distance. Weighted network analysis reveals a high-bandwidth functionally clustered backbone (according to cortical lobes) within the sea of weak connections ensuring maximum local processing efficiency. These findings underline the importance of connection strengths and geometry in cortical architecture and processing.