OpenAI has opened a free route into its life sciences model. On October 15, 2026 the company announced the Rosalind Academic Plan, which lets eligible academic labs get "up to 30 Business Standard seats at no cost for up to 12 months," with access to GPT-Rosalind "subject to trusted-access approval," according to OpenAI's announcement. It is a plan and access change that matters to anyone who pays for ChatGPT seats in a research setting.
This post covers what the plan includes, what it leaves out, how the approval process works, and how it fits with the earlier Rosalind launches we have covered.
TL;DR: the Rosalind Academic Plan in one table
| Question | Answer (per OpenAI) |
|---|---|
| What do labs get? | Up to 30 Business Standard seats at $0 per user per month for up to 12 months |
| Which models? | All commercially available frontier OpenAI models, subject to plan usage limits |
| Is GPT-Rosalind included? | Yes, but only after trusted-access approval |
| Who can apply? | Academic labs at U.S. universities, for now |
| Is the API included? | No, not at this time |
| Is lab data used for training? | Not by default |
| Who manages seats? | Verified PIs or lab leads, who are responsible for their lab's approved use |
Every figure in this post comes from OpenAI's own page. There is no independent measurement of how useful GPT-Rosalind is for any specific lab, and OpenAI's own sample assessments carry a disclaimer that they do not verify scientific correctness, feasibility, or whether a method works.
What exactly does the plan include?
OpenAI lists three concrete benefits for eligible labs:
- Up to 30 Business Standard seats at $0 per user per month for up to 12 months. That is a real subsidy: a 30-person lab would otherwise pay per-seat business pricing for every student, postdoc and technician who uses the tool.
- Access to all commercially available frontier OpenAI models, subject to plan usage limits. OpenAI does not publish the limits in the announcement, so labs should expect the usual caps on heavy usage rather than unlimited access.
- Access to GPT-Rosalind following trusted-access approval. This is the specialized life sciences model, paired with the Rosalind workbench that coordinates tools across medicinal chemistry, genomics, wet-lab assistance and other scientific applications.
Two details are easy to miss. First, the free seats are an account arrangement, not a credit pool: a PI activates free access for "up to 30 named members." Second, the plan excludes the API, so labs that want to script GPT-Rosalind into a pipeline cannot do that through this program yet.
Why is this a plan change and not just a press release?
Because it changes what a customer segment can get and pay. Research groups are a natural buyer for a specialized biology model, and they are also the group least able to absorb per-seat pricing. A year of subsidized seats removes the budget conversation for a lab that was waiting on a grant line.
It also fits a pattern in how OpenAI has rolled out Rosalind. The model arrived with a benchmark (LifeSciBench, 750 tasks for biotech work), then a product surface in the Rosalind Workbench research preview, and now a pricing and access program aimed at universities. OpenAI also published a computational biology benchmark with an earlier model generation, covered in our GeneBench Pro post.
Which research areas does OpenAI point to?
The announcement lists seven areas where it expects academic labs to apply the model:
| Area | What OpenAI says it covers |
|---|---|
| Drug and protein design | Candidate drugs, therapeutic proteins, peptides and enzymes, and optimizing stability and suitability for development |
| Disease detection and monitoring | Molecular signatures of disease, liquid biopsies and less invasive diagnostics |
| Vaccines and immune responses | Antibodies and vaccines against emerging variants, host-pathogen interactions, durability of immunity |
| Genetic and RNA therapies | Correcting genetic defects, RNA therapeutics, safe delivery to the intended tissues |
| Agriculture and animal health | Crop resilience, plant disease, animal health, veterinary care |
| Ecology and conservation | Biodiversity, adaptation, species interactions, bioremediation |
| Antimicrobial resistance | How microorganisms resist treatment and new treatments for resistant infections |
OpenAI also quotes three researchers. The Zhao Lab at UCSF says GPT-Rosalind helped it gather drug physicochemical parameters faster across ten drugs using PubChem and NCBI tools. Dr. Mathieu Bourdenx says GPT models help him iterate on ideas faster. Dr. Brian Hie of Stanford and the Arc Institute says Rosalind should help apply discovery and design tasks "at larger scales than previously possible." These are testimonials chosen by OpenAI, not evaluations.
How does the trusted-access review work?
GPT-Rosalind is not a model you simply switch on. OpenAI says that before activating access it will:
- Verify the institution's legitimacy and the PI or lab lead's identity.
- Review the lead's professional background, institutional appointment and authority to oversee lab access.
- Assess the lab's institutional affiliation, research activities and intended use.
- Establish clear responsibilities for the lead and a designated institutional safety contact.
Safeguards continue after activation through monitoring and enforcement. For registered high-sensitivity use cases, OpenAI says its safety checks account for the approved research context, and it may follow up with researchers to confirm the work and their authorization.
Two separate gates a key passes through, illustrating trusted access approval for Rosalind Academic Plan labs
This is a gated-access model: approved context is meant to reduce unnecessary interruptions for legitimate researchers, while monitoring is meant to catch misuse. The trade-off is friction. Labs should expect an application with credentials and a stated research purpose, not instant activation.
Who is eligible, and who is not?
OpenAI draws two lines. Academic labs at U.S. universities without an OpenAI Enterprise agreement can apply through the trusted-access process, with the PI or lead applying on behalf of the lab. Labs at universities that already have an agreement must go through their university's Enterprise access process, which has a separate request form for GPT-Rosalind.
Outside the plan: labs outside the United States, companies, hospitals and non-university research institutes are not covered by this announcement, and the API is excluded. If you sit in one of those groups, the existing Rosalind Workbench preview remains the reference point, and OpenAI has said nothing about extending the academic plan.
How does this compare with what other labs offer scientists?
The race to put frontier models in scientists' hands is crowded. Anthropic has its Claude for science workbench and committed credits to federal science agencies through its Genesis Mission program. Google DeepMind has put its co-scientist into real-world labs and published a survey of AI in science in its discovery curve report. Claude has also been credited with surfacing a novel enzyme system.
What distinguishes OpenAI's offer is the combination of a dedicated domain model, a named seat count and a fixed 12-month window. What it does not offer is published usage limits or API access, so a direct cost comparison with other programs is not possible from public information.
What should a PI do before applying?
- Check your university's status. If your institution has an OpenAI Enterprise agreement, use its internal route rather than the lab application.
- Name your safety contact early. OpenAI requires a designated institutional safety contact, so loop in your biosafety or research compliance office before you submit.
- Write a precise research description. The review assesses affiliation, activities and intended use, so vague applications are the likeliest to stall.
- List your 30 members. Seats are for named members, and the PI is responsible for their approved use.
- Plan for the end date. Free seats last up to 12 months. Decide now what happens to your workflows when the subsidy ends, particularly if you build habits around features that only exist on paid plans.
- Keep sensitive data decisions explicit. OpenAI says shared research data is not used for training by default, but your institution's own data-handling rules still apply.
What we do not know yet
- The exact usage limits for the free seats.
- Whether the plan will expand beyond the United States or to API access.
- Whether the trusted-access queue will be fast enough for a semester timeline.
- Any independent evidence that GPT-Rosalind beats general frontier models on a lab's own tasks.
Those gaps matter more than the headline. A free seat is only worth what you can do with it inside the limits and the review process.
The bottom line
The Rosalind Academic Plan is a meaningful subsidy for US university labs, with a gated model behind it. If your lab does life sciences work and qualifies, applying costs little. If you are outside the United States, in industry, or need the API, nothing here changes your options yet. We will update this post if OpenAI publishes limits or widens eligibility.
Related reading
- OpenAI Rosalind Workbench: a life sciences workspace in research preview
- LifeSciBench: OpenAI's 750-task benchmark for GPT-Rosalind
- GeneBench Pro and computational biology
- Claude for science: an AI workbench for scientists
- Anthropic's Genesis Mission credits for federal science agencies
- Google DeepMind co-scientist in real-world labs
- GPT-6 and Intelligent UI in ChatGPT, explained
Sources: OpenAI: Bringing GPT-Rosalind to academic research labs (October 15, 2026). OpenAI frames the plan with historical examples of publicly funded basic research, such as the Human Genome Project, which is the kind of foundational academic work it says it wants to support.
Details on eligibility and limits are accurate as of October 16, 2026 and may change.