<p>This paper develops an energy-aware mobile edge computing (MEC) framework integrated with cell-free massive MIMO (CF-mMIMO) for dense-user scenarios. We formulate a joint resource-allocation problem that targets worst-user latency (min-max) under power and computation constraints, and solve it via an iterative convex approximation (ICA) method to obtain high-quality labels. Leveraging these ICA-derived solutions, we train two data-driven edge controllers-a convolutional neural network (CNN) and a spiking neural network (SNN)-that replace online iterative optimization with a single forward pass, thereby reducing decision complexity and supporting near real-time operation at the edge. Simulations show that the CNN closely approximates the ICA baseline (low MSE), while the SNN achieves comparable accuracy with substantially lower inference energy (about one-third of the CNN), and both preserve fairness across users and resource blocks. Overall, the proposed approach balances accuracy and energy efficiency and is thus relevant for sustainable, scalable edge intelligence in MEC-enabled CF-mMIMO networks.</p>