Gokturk-2 is the third Earth observation (EO) satellite of Turkey and acquires high-resolution panchromatic (2.5 m) and multispectral images (5 m) from its near-polar orbit at a nominal altitude of 685 km from the Earth. The push broom scanner mounted on the satellite can obtain both mono and stereo images. Images of Gokturk-2 can be used in many application fields both for defense and civilian use, such as environment and city planning, geology, agriculture, and forestry. The purpose of this study is to evaluate the radiometric and geometric quality of Gokturk-2 stereo images acquired from panchromatic band and the digital surface model (DSM) generation capability from these images. The Gokturk-2 stereo images used in this study have been acquired over a test area located near Kesan town of Edirne Province in Turkey in May 2017. Initially, the radiometric quality has been investigated by statistical analysis of the histogram, noise estimation on homogeneous surfaces, and visual checks. The image orientation has been performed by polynomial modeling of the image trajectories using a number of ground control points (GCPs) extracted from road intersections in cadastral maps. The image orientation accuracy obtained using the GCPs is around 5.4 pixels in planimetry and 4.5 pixels in height for this stereopair. The DSM has been generated with dense feature point matching using normalized cross-correlation. The quality of the generated DSM has been assessed by comparison with a reference digital terrain model (DTM) data in open terrain areas. The results show that despite the inferior image orientation accuracy and radiometric problems, DSMs can be generated with sufficient accuracy especially in open terrain. Major problems occur in forests and also in areas with rugged topography.