Neurons in the brain exhibit significant heterogeneity in their dynamics. Whether this diversity plays any computational role and if it is exploited to perform efficient computation is not well understood. We used the reservoir computing paradigm to answer this question. Balanced networks with different levels of heterogeneity in their neuronal time constants were set up and benchmarked against various working-memory-related tasks, including stimulus recall, forecasting, and nonlinear transformation. Our results suggest that neuronal heterogeneity, even without any recurrent communication, significantly enhances the performance of a wide range of temporal tasks. Moreover, this computational benefit is highly robust to the hyperparameters of our model: the average neuronal time constant and global synaptic strength. Furthermore, this benefit is already present in networks of very modest size, with hundreds of neurons. We empirically show that this computational benefit stems from dimension inflation that renders the neuronal activity repertoire richer in heterogeneous networks. We also verified our results in networks with spiking neurons and different types of inputs (chaotic and periodic) with noise to demonstrate that our results are independent of the dynamics and the inputs' statistics and predictability. In summary, our results provide firm evidence for the benefit of neuronal diversity in brain regions involved in information processing such as the hippocampus, prefrontal cortex, or entorhinal cortex. A corollary of our results is that regions with a minimal role in neuronal processing, and especially the "actuator" regions of the brain dedicated to motor generation, must be homogeneous to ensure the reliability of movement.