Code and data for: Language models are equally competitive but unequally cooperative

Code and data for the paper "Language models are equally competitive but unequally cooperative" (Felipe M. Affonso, Spears School of Business, Oklahoma State University). Twenty-five language models from seven developers played 38 repeated economic games against programmed opponents, copies of themselves and other models, in 51,906 trials. llm-strategic-personalities.zip is the code repository (https://github.com/FelipeMAffonso/llm-strategic-personalities): the game engine and the 38 games, the code that ran the models, the analysis scripts, the supplementary tables, the figure scripts and figures, the trial-level dataset and the analysis outputs, and sample trial files. The complete raw corpus of 51,906 trial files is archived separately at https://doi.org/10.5281/zenodo.19896196; unpack it into a folder named raw_data at the top of the repository to rerun the analyses that read individual rounds. Version 3 accompanies the manuscript named above. It adds the re-parsing check and corrects the figures, tables and documentation after an audit of the parsing and the design coverage. Version 1 is the April 2026 replication package of an earlier version of the study.