Synthetic Aperture Radar (SAR) target detection suffers from complex environments and small target sizes. The number of baseline model RT-DETR parameters is high. Aiming at the above problems, this paper proposes a lightweight small target detection in SAR images based on improved RT-DETR (LSTRT-DETR). Firstly, a lightweight additive convolutional GLU backbone network (LACGBNet) is designed to capture remote dependencies and optimize the feature extraction process to cope with complex scenes while achieving model lightweighting. Secondly, the algorithm proposes a small target feature fusion network (STFFN), which introduces SPD-Conv to process the P2 feature layer to enrich the small target feature information. And the shallow feature fusion module (SDFM) is introduced to enhance the detection capability of small targets. Experimental results on the joint dataset consisting of SSDD, HRSID, and SARAIRcraft show that the mAP50 and mAP50-95 of LSTRT-DETR reach 92.5% and 63.8%, respectively, which are improved by 0.4% and 3%, respectively, and the number of parameters is reduced by 4.1 M compared to the original model RT-DETR. This sufficiently demonstrates its effectiveness and efficiency in the SAR image target detection task.