Immersive and non-immersive virtual reality (NIVR) technology can supplement and improve standard physiotherapy and neurorehabilitation in post-stroke patients. We aimed to use MIRA software to investigate the efficiency of specific NIVR therapy as a standalone intervention, versus standardized physiotherapy for upper extremity rehabilitation in patients post-stroke. Fifty-five inpatients were randomized to control groups (applying standard physiotherapy and dexterity exercises) and experimental groups (applying NIVR and dexterity exercises). The two groups were subdivided into subacute (<six months post-stroke) and chronic (>six months to four years post-stroke survival patients). The following standardized tests were applied at baseline and after two weeks post-therapy: Fugl–Meyer Assessment for Upper Extremity (FMUE), the Modified Rankin Scale (MRS), Functional Independence Measure (FIM), Active Range of Motion (AROM), Manual Muscle Testing (MMT), Modified Ashworth Scale (MAS), and Functional Reach Test (FRT). The Kruskal–Wallis test was used to determine if there were significant differences between the groups, followed with pairwise comparisons. The Wilcoxon Signed-Rank test was used to determine the significance of pre to post-therapy changes. The Wilcoxon Signed-Rank test showed significant differences in all four groups regarding MMT, FMUE, and FIM assessments pre- and post-therapy, while for AROM, only experimental groups registered significant differences. Independent Kruskal–Wallis results showed that the subacute experimental group outcomes were statistically significant regarding the assessments, especially in comparison with the control groups. The results suggest that NIVR rehabilitation is efficient to be administered to post-stroke patients, and the study design can be used for a further trial, in the perspective that NIVR therapy can be more efficient than standard physiotherapy within the first six months post-stroke.