Fluid intelligence is one of the general intelligence types originally proposed by Cattell (1963), which refers to the innate ability of analytically solving novel problems and logically identifying new patterns and relationships. Increasing evidence has shown that cognitive training, especially those aiming at enhancing working memory, can significantly improve fluid intelligence. Moreover, central executive functions, especially working memory updating, were reported to positively correlate with individual intelligence. Thus, it is of great theoretical and practical significance to investigate whether working memory training can improve fluid intelligence. Recent studies have shown that, after short-term working memory training, performance in fluid intelligence tests, including running memory task and n-back paradigm, was improved both in adults and school-aged children. It is therefore suggested that the transfer of an improved working memory updating ability contributed to the reported training effects. However, there remain 2 major unsolved problems. On the one hand, although training has been reported effective for adults and school-aged children, few studies have focused on pre-school children. On the other hand, the demonstration of long-term effects was unreliable, because the reported studies either failed to examine lasting effects or lacked time for confirmation. In our study, 96 children aged 4-5 from one kindergarten participated in the experiment. They were randomly selected from 3 classes and assigned to 3 groups - experimental group, control group 1, and control group 2. Participants in experimental group were trained for 15 minutes per day for a period of 14 days using an n-back working memory training program displayed on tablet computers. In this program, participants were shown a series of stimuli and asked to judge whether the current stimulus is the same as the one displayed before the previous n items. As n increased, the task difficulty increased, resulting in higher demands of working memory. Each participant started training at 1-back level with n being continuously adjusted based on their performance. To account for the possibility that better fluid IQ performance is essentially due to attention improvement, participants in control group 1 played the Fruit Ninja game for the same period of time, which was designed to yield comparable attention improvements. For control group 2, children received no training. Parallel versions of Raven’s Standard Progressive Matrices (SPM) and the Test of Nonverbal Intelligence (TONI-4) were used to assess participants’ fluid intelligence in pretest, posttest and a follow-up test 6 months after training respectively. All experimental manipulations and the data analysis were designed and performed double-blinded. Results indicated that, after 14-day training, participants’ task performance in the training task was significantly improved. Children in experimental group significantly outperformed 2 other control groups in the posttest, which remained equally significant 6 months after training. As task motivation and attitude were balanced, our results were convincing in demonstrating training effects. In conclusion, our results proved that the n-back working memory task can efficiently improve kindergarten children’s fluid intelligence, whose effect is sustainable in the long term.