A weight is a numerical value in a neural network that is learned during training and determines how much influence one neuron's output has on the next. During forward propagation, each input to a neuron is multiplied by its corresponding weight before being summed. Training algorithms like backpropagation adjust weights iteratively to minimize the loss function. The collection of all weights in a model constitutes the bulk of its learnable parameters, and their final values encode the patterns the model has extracted from its training data.