A study by seven law professors at the University of Maryland Francis King Carey School of Law found that OpenAI's latest generative AI model, o3, earned grades from A+ to B on eight spring law school final exams. The results were published on SSRN. The AI model received an A+ in Constitutional Law, Professional Responsibility, and Property; an A in Income Taxation; an A- in Criminal Procedure; a B+ in Secured Transactions and Torts; and a B in Administrative Law. The program performed well on both essays and multiple choice questions. These results show a marked improvement over earlier versions of ChatGPT, which previously earned grades of B, C, and D on similar exams in 2022 and 2023. The o3 model is described as a 'reasoning model' that internally evaluates and revises its responses before generating final answers. The study identified limitations, including the model's lower grade in Administrative Law due to its lack of knowledge about the 2024 U.S. Supreme Court decision Loper Bright Enterprises v. Raimondo. The model's performance also declined when given access to professors' notes, possibly due to information overload. OpenAI did not comment on the study.