OpenAI's Rosalind Academic Plan gives US university labs 30 free seats for a year

OpenAI is giving US university research labs free access to its life-sciences model. The Rosalind Academic Plan, announced this week, gives eligible labs up to 30 free seats on the Business Standard plan for up to twelve months. API access is not included. And the seats come with vetting: before granting access, OpenAI reviews the institution, the lab's lead researcher, and the planned use, and it monitors how approved labs use the model.

What the plan covers

The plan centers on GPT-Rosalind, the life-sciences model series OpenAI released in April. The company says it supports deeper reasoning on complex biological questions and that it leads BixBench, a benchmark built around real-world bioinformatics tasks, among models with published scores. Through the Rosalind Workbench, the model connects to specialized tools for areas like medicinal chemistry, genomics and lab experiments. The workbench extends the life-sciences research plugin for Codex that OpenAI introduced at launch, which connects models to more than 50 scientific tools and data sources.

The academic seats are a separate channel from OpenAI's existing customer track. GPT-Rosalind is already available in ChatGPT, Codex and the API through its trusted access program, with customers including Amgen, Moderna, the Allen Institute and Thermo Fisher Scientific.

The vetting side, in detail

This is the part worth a closer look. OpenAI vets each institution, lab lead and intended use before access, then monitors model usage afterwards. Approved researchers can work on sensitive scientific questions inside a set of safeguards. For registered, highly sensitive applications, the safety checks factor in the approved research context, and OpenAI follows up with researchers when something is unclear.

One line matters for any academic lab weighing an application: OpenAI says it does not train on research data by default.

What this means for a lab

Two practical points. First, eligibility: the plan is currently open only to research labs at US universities, and application details live in OpenAI's help center. If your lab fits, thirty free seats for a year puts the cost of trying a frontier model on your own workflows near zero. Second, the vetting model itself: checking the institution and the planned use before access is a concrete signal of how a frontier lab intends to distribute a model that touches sensitive science, and the monitoring afterwards tells you that access comes with conditions.

If you run an academic lab, the arithmetic is simple, and the main non-zero cost is the vetting process. The practical question is whether the workbench tooling matches your actual workflow. The help-center application is the first step to finding out.