Abstract: Objective To establish an improved mouse model of valproic acid (VPA)-induced autism that better mimics human autism. Methods We established mouse models of autism in female C57 mice by intraperitoneal injection of sodium valproate either at a single dose (600 mg/kg) on day 12.5 after conception (conventional group) or in two doses of 300 mg/kg each on days 10 and 12 after conception (modified group), and the control mice were injected with saline only on day 12.5. The responses of the mice to VPA injection, the uterus, mortality rate, and abortion rate were compared among the 3 groups. The morphology and development of the offspring mice were assessed, and their behavioral ontogeny was evaluated using 3- chambered social test, social test, juvenil play test, and open field test. Results The mortality and abortion rates were significantly lower in the modified model group than in the conventional group (P<0.01). Compared with those in the control group, the offspring mice in both the conventional group and the modified group showed developmental disorders (P<0.05). The mortality rate of the newborn mice was significantly lower in the modified group than in the conventional group with a rate of curvy tail of up to 100% (P<0.001). The offspring mice in both the modified group and conventional group exhibited autism-like behavioral abnormalities, including social disorder and repetitive stereotyped behavior (P<0.05). Conclusion The mouse model of autism established using the modified method better mimics human autism with reduced mortality and abortion rates of the pregnant mice and also decreased mortality rate of the newborn mice.