GPT-Rosalind is OpenAI's specialist model for the life sciences. On October 15, 2026, OpenAI launched the Rosalind Academic Plan: eligible academic labs at US universities can claim up to 30 complimentary ChatGPT Business Standard seats for as long as 12 months. The seats include OpenAI's commercially available frontier models, while GPT-Rosalind itself sits behind a separate trusted-access review.
What is free is a set of seats, not API credit
The scope matters. A principal investigator or lab lead applies on the lab's behalf, and organizational approval is required. Each workspace is capped at 30 seats, and free does not mean unlimited: the usual usage limits still apply. The plan explicitly excludes the API. There are no API credits, seat holders cannot call GPT-Rosalind through an interface, and the model cannot be wired into a lab's own pipeline. Whether a member can use GPT-Rosalind at all is switched on individually by the workspace owner. On data, OpenAI says research data submitted through the plan is not used for model training by default, a key point for labs holding unpublished, sensitive results, although standard data retention applies and zero data retention is not supported.
Why a second layer of approval exists
GPT-Rosalind debuted as a research preview in April, built to reason across biological entities, synthesize evidence from the literature, plan experiments and connect with scientific databases. Figures OpenAI published at the time showed clear gains over baselines on GeneBench, Medchem Bench and Labworkbench. Capabilities like these are inherently dual-use: they can speed up legitimate screening and design work, and they can be misused. So this free rollout does not work on a sign-up-and-go basis. Institutional review, trusted access, a named safety contact and ongoing monitoring are all part of the package. The price is waived; the gatekeeping is not.
Which labs should apply now
US university labs with real biology or chemistry projects and no budget for commercial subscriptions benefit most: use the seats to run everyday work such as literature synthesis and experiment planning, then decide whether a paid plan is worth it. Teams that need API integration or bulk data runs, or that sit outside the US university system, will get nothing from this plan. Specialist research models are becoming a new competitive variable between labs. The physics model built by CERN with Mistral follows the same logic: model vendors are competing not just for users, but to become the default entry point of a discipline's workflow.
