Firefly algorithm is a bionic random algorithm for solving complex optimisation problems. Any firefly will be attracted to other better fireflies to complete the population evolution. In this method, the better fireflies only show the advantages of them, but do not represent that of swarm. In order to enhance information exchange between the swarms, different learning time of the general central particle is embedded into the particle update phase. So, we propose the general central firefly algorithm based on different learning time. Correspondingly, three variants of general central FA are generated, namely, the algorithms based on one-to-one (OO) learning time, one-to-all (OA) learning time and all-to-all (AA) learning time. Experiments are tested on 12 benchmark functions. The results show that the optimisation performance of three algorithms are better than that of the standard FA. OOFA algorithm has the best optimisation performance.