Ahlswede and Dueck identification has the potential of exponentially reducing traffic or exponentially increasing rates in applications where a full decoding of the message is not necessary and, instead, a simple verification of the message of interest suffices. However, the proposed constructions can suffer from exponential increase in the computational load at the sender and receiver, rendering these advantages unusable. This has been shown in particular to be the case for a construction achieving identification capacity based on concatenated Reed-Solomon codes. Here, we consider the natural generalization of identification based on Reed-Muller codes and we show that, although without achieving identification capacity, they allow to achieve the exponentially large rates mentioned above without the computational penalty increasing too much the latency with respect to transmission.