The gloss_tinynetworks dataset is a collection of behavioral and computational model data created for the study “Human gloss perception reproduced by tiny neural networks.” It contains raw and cleaned MATLAB data files, including online behavioral response data and model outputs used to analyze how humans perceive surface gloss and how tiny neural networks predict those judgments. The dataset supports data preprocessing, participant exclusion, summary statistic computation, correlation and kernel-fitting analyses, t-SNE visualization, and comparison of model predictions with human perceptual responses. Its primary purpose is to reproduce the manuscript’s main and supplementary figures and to enable further psychophysical and model-based analyses of gloss perception.