Brain tumors are a significant health threat to humans. Magnetic resonance imaging (MRI) holds considerable promise for the automatic delineation of brain tumors in clinical evaluations. Adopting more accurate image segmentation technology is essential in brain tumor diagnosis. This article develops an innovative automated three-dimensional glioma automatic segmentation network based on attention mechanism. This model mainly uses a sampling module combining contextual and channel attention mechanism. The GC-CA block, which integrates both contextual and channel attention mechanisms, allows the network to concentrate on crucial channel characteristics while capturing global information. Meanwhile, the position pixel attention block (PA) enhances the model’s ability to better focus on essential features during energy processing, and improve the attention and effectiveness of the model to the input data. In addition, the upper sampling also combines the residual connection to help the model better use the underlying feature information to make the network more effectively learn the characteristics. In short, with the automatic segmentation of this model, it provides an efficient and accurate solution with substantial potential for practical clinical use.