The Diffusion Images for Quality Assessment dataset consists of 1,000 images generated by 10 open-source diffusion models using 100 diverse text prompts. These prompts are grouped into five categories: simple, realistic, fairy-tale, complex, and artistic. The dataset was created to evaluate the quality of text-to-image generation models using Mean Opinion Score (MOS) predictions. Each model was tasked with generating 512x512 images, ensuring variety and balance across different styles and complexity levels. A detailed description of the prompts used to generate each image is provided in the file prompts.json, while the measured MOS for each image is recorded in the file diffusion_dataset_results_512.csv. The data was generated using 10 open-source models, and the names of these models correspond to the folder names where the images are stored. This dataset serves as a valuable resource for researchers and developers aiming to benchmark and improve image quality assessment algorithms in generative AI applications.