This collection consists of SpatialThinker 3B and 7B model checkpoints, and STVQA-7K, a Spatial VQA dataset used for training the models.