Empowering Education At the Icahn School of Medicine With AI

Active medical and graduate students at the Icahn School of Medicine at Mount Sinai, along with some teaching and research faculty, now have comprehensive access to ChatGPT through the School. Compared to free users, they’ll be able to make more complex queries with fewer limits, especially regarding science, technology, engineering, and mathematics (STEM) fields, and benefit from enhanced privacy protections.

The Icahn School announced, in May, that it has inked an agreement with OpenAI for an educational license—dubbed ChatGPT Edu—of OpenAI’s products. While a number of universities are using ChatGPT Edu, this agreement makes the Icahn School the first medical school partner in the nation, according to OpenAI.

“At Mount Sinai, we believe it’s our responsibility not just to adopt emerging technologies, but to do so with care, purpose, and a strong commitment to equity and academic integrity,” said David C. Thomas, MD, MHPE, Dean for Medical Education at the Icahn School of Medicine.

“This initiative is the result of a close collaboration between the Scholarly and Research Technologies team, the Department of Medical Education, Graduate School leadership, and OpenAI,” says Marta Filizola, PhD, Dean of the Graduate School of Biomedical Sciences. She added that the Levy Library team has developed educational content and workshops to support students, faculty, and staff in effectively using ChatGPT Edu and integrating AI into academic learning, research, and scholarly activities.

What’s contained in this license, and what do those features spell for our users? Read on to learn more.

Access to advanced models

OpenAI offers a range of models to handle different types of tasks. Through the educational license, users can access them via ChatGPT.com. The following bullets describe the models available to Icahn School of Medicine users, and the table below summarizes specifications at a glance. The following information is current as of June 2025, but may change in the future as OpenAI models can change rapidly.

• GPT-4o: This is OpenAI’s flagship model, the one most consumers are familiar with, and starts up by default on the ChatGPT Edu license. This model of reasoning is grounded in unsupervised learning, and is capable of processing text, images, file uploads, and voice for input and output. Tasks this model is good at include summarizing meetings into actionable notes, drafting copy, proofreading, and coming up with new angles and ideas.
• GPT-4.1, GPT-4.1-mini: A specialized model for coding and tasks that need precise, step-by-step instructions. This model is helpful for front-end coding, such as developing web applications, debugging, explaining, and improving code, or comparing long reports while focusing on specific factors. The mini version trades off processing capability for speed, and is ideal for people who need quicker answers for simpler problems about coding.
• The o-series: These are advanced models of reasoning, designed to deliberate and produce a “chain of thought” before generating answers. Its deliberation process is supposed to provide better-quality output, with lower levels of hallucinations. It is described as being “ideal for complex queries requiring multifaceted analysis and whose answers may not be immediately obvious.”
  • o3: With a knowledge base specialization in STEM, this model is suited for advanced math, science, and coding tasks, detailed analysis, devising strategy, and visual reasoning.
  • o4-mini, o4-mini-high: The model to use if one needs the STEM expertise and deliberative reasoning, but quicker answers. The o4-mini model can provide quick summaries of scientific articles or extract key points from an Excel spreadsheet, and the o4-mini-high model is capable of more complex tasks, albeit more narrowly than the o3 model in terms of subject expertise.
• GPT-4.5: Intended to serve as a bridge between GPT-4o and its eventual successor, GPT-5, this model draws from both unsupervised learning and the “chain of thought” reasoning that the o-series models use. It was touted as being fine-tuned to understand human needs and intents better, and for better collaboration with the user. This model will be phased out later this year, in favor of GPT-4.1, as announced by OpenAI.

Privacy, data safety, and ethics guardrails

Personal health information, sensitive information, and student data are safeguarded via a data privacy and business associate agreement between the Icahn School and OpenAI.

This means that no data, prompts, or responses will be used to train OpenAI’s models when using a ChatGPT Edu account through the Icahn School. The guardrails implementation was guided by Research and Education leadership, Cybersecurity, Compliance, and Legal teams, alongside the AI Steering Committee on Teaching, Learning, and Discoveries.

Community sharing of custom-build applications

Users can create customized resources for specific tasks—such as a flash card study guide for biochemistry—known as GPTs. Icahn School of Medicine ChatGPT Edu users can share the GPTs they’ve built on the Explore GPTs community page with other users, as well as access OpenAI’s official GPTs. However, they can’t access GPTs built by public users. This ensures a safe and secure space that supports collaboration while following Mount Sinai’s security standards.

“ChatGPT is just a great way to establish the baseline for what I want to say how I want to approach the patient,” said Joy Jiang, MD/PhD student, about how she used the AI tool during a mock patient simulation exam, in an interview segment on CBS Saturday Morning.

Watch how various students from the Icahn School of Medicine are using ChatGPT to bring their education to the next level in the segment below.