Findings of neurocognitive dysfunction in subjects at Clinical High Risk for Psychosis (CHR-P) have been controversial. This meta-analysis systematically examined studies of neurocognitive functions using the MATRICS Consensus Cognitive Battery (MCCB) in CHR-P. An independent literature search of both English and Chinese databases was conducted by two reviewers. Standardized mean difference (SMD) was calculated using a random effects model to evaluate the effect size of the meta-analytic results. Six case-control studies (n = 396) comparing neurocognitive functions between CHR-P subjects (n = 197) and healthy controls (n = 199) using the MCCB were identified; 4 (66.7%) studies were rated as "high quality". Compared to healthy controls, CHR-P subjects showed impairment with large effect size in overall cognition (n = 128, SMD = -1.00, 95%CI: -1.38, -0.63, P < 0.00001; I = 2%), processing speed (SMD = -1.21) and attention/vigilance (SMD = -0.83), and with medium effect size in working memory (SMD = -0.76), reasoning and problem solving (SMD = -0.71), visual (SMD = -0.68) and verbal learning (SMD = -0.67). No significant difference between CHR-P subjects and controls was found regarding social cognition (SMD = -0.33, 95%CI: -0.76, 0.10, P = 0.14; I = 70%) with small effect size. Apart from social cognition, CHR-P subjects performed worse than healthy control in all MCCB cognitive domains, particularly in processing speed, attention/vigilance and working memory.