Rett syndrome (RTT) is a severe neurodevelopmental disorder caused by a mutation in the X-linked methyl-CpG-binding protein 2 (MECP2). There is currently no disease-specific treatment, but MECP2 restoration through reactivation of the inactive X (Xi) has been of considerable interest. Progress toward an Xi-reactivation therapy has been hampered by a lack of suitable female mouse models. Because of cellular mosaicism due to random X-chromosome inactivation (XCI), Mecp2+/- heterozygous females develop only mild RTT. Here, we create an improved female mouse model by introducing a mutation in Tsix, the antisense regulator of XCI allelic choice. Tsix-Mecp2 mice show reduced MECP2 mosaicism and closely phenocopy the severely affected Mecp2-null males. Tsix-Mecp2 females demonstrate shortened lifespan, motor weakness, tremors, and gait disturbance. Intriguingly, they also exhibit repetitive behaviors, as is often seen in human RTT, including excessive grooming and biting that result in self-injury. With a Tsix allelic series, we vary MECP2 levels in brain and demonstrate a direct, but nonlinear correlation between MECP2 levels and phenotypic improvement. As little as 5-10% MECP2 restoration improves neuromotor function and extends lifespan five- to eightfold. Our study thus guides future pharmacological strategies and suggests that partial MECP2 restoration could have disproportionate therapeutic benefit.