Abstract: To optimize the process of preparing cobalt powder by hydrogen reduction of cobalt oxide and improve product quality, key influencing factors (boat loading capacity, reduction temperature, reduction duration) are first screened via single-factor experiments. Then, the response surface Box-Behnken design is used to explore process conditions, establish a cobalt powder quality prediction model, and verify it. Results show that the model has excellent fitting performance (R2=0.9918), with predicted values consistent with experimental values, enabling accurate prediction of cobalt powder quality. The optimal process conditions are: boat loading capacity 68 g, reduction temperature 398 ℃, reduction duration 186 min. Under these conditions, the cobalt powder has a particle size of 0.98 μm, oxygen content of 0.38%, carbon content of 0.011%, span of 1.54, quality score of 83, and optimal comprehensive performance. Analysis of variance (ANOVA) shows the significance order of factors affecting quality is reduction temperature (B)＞reduction duration (C)＞boat loading capacity (A), with significant interactions among factors. The process parameters and model have guiding significance for the industrialization of this technology and provide support for enterprises to improve quality and efficiency.