OpenAI offers labs up to 30 Rosalind seats for a year

The offer, in OpenAI's words

OpenAI's October 15, 2026 post introduces the Rosalind Academic Plan for GPT-Rosalind, which it describes as a life-sciences model with a workbench for tool use across chemistry, genomics, and lab work. Eligible labs can receive up to 30 Business Standard seats at no cost for up to 12 months. Access to GPT-Rosalind still depends on trusted-access approval. The same plan includes other commercially available frontier models, subject to usage limits. The Decoder's October 16 headline matches the seat count and the year. An evaluation a second lab can run is a different arrangementan evaluation a second lab can run.

Who is eligible

The post says the plan is currently for academic labs at U.S. universities. A principal investigator or lab lead at a university without an OpenAI enterprise agreement applies for the lab. A university that already has an enterprise agreement goes through that agreement. After approval, the lead can turn on access for up to 30 named members. The lead is responsible for the lab's approved use. OpenAI says it verifies the institution and the lead, reviews the appointment and the intended use, and wants an institutional safety contact. The artificial-intelligence desk isArtificial Intelligence.

What is not in the box

API access is not included, OpenAI says. Seats are not a blank cheque for every question: the post says safeguards continue after activation, and that high-sensitivity use is checked against the approved research context. This brief will not walk through a scientific workflow, a molecule, or an assay. The product fact is the seat, the approval, and the training default. The privacy desk is Asked directly, How many people can a lead add? Up to 30 named members, once the lab is approved. The lead manages permissions and is responsible for approved use. A designated institutional safety contact is part of the setup OpenAI describes.Privacy.

The training line

OpenAI says research data shared through the plan is not used to train its models by default. That sentence is a default, which means a reader still has to read the agreement that applies to the lab. It is the opposite of a claim that every upload is somehow outside OpenAI's systems. The data goes to the service. The post says it is not used for training unless that default is changed. The software desk is The document named in the sources is the record this page will keep.Software.

Named labs are examples, not the eligibility list

The post quotes researchers at UCSF, the Bourdenx Lab, and Stanford with the Arc Institute about using the model on literature, parameters, and design tasks. Those quotations show early users. They do not mean a lab outside that list is excluded, and they do not transfer those labs' access to a reader. Eligibility is the U.S. academic path above. The paper is Asked directly, What happens to research data? OpenAI says research data shared through the plan is not used to train its models by default.Futureweb.

What The Decoder is for

The Decoder is the wire account: up to 30 free seats for a year. The numbers this page is willing to print come from OpenAI's post. A headline that matches is useful as a second telling. It does not add a country, a price, or an API that the post does not offer. The hardware desk is Asked directly, Are the seats unconditional? No. The post says access to GPT-Rosalind is subject to trusted-access approval, and that the institution, the lab lead, the research, and the intended use are reviewed before activation.Hardware.

How a lead should read the offer

Thirty seats, twelve months, no per-user charge on the Business Standard seats, U.S. universities, approval before GPT-Rosalind turns on, no API, training off by default. A lab that needs a programmatic interface does not get one from this plan, on OpenAI's own sentence. A lab outside the United States is outside the current availability line. The security desk is Asked directly, Does the plan include the API? OpenAI says API access is not included at this time. The document named in the sources is the record this page will keep.Security.

What stays off this page

The post demonstrates the model with research-shaped prompts. This brief does not repeat them and does not turn them into a method. A news reader needs the commercial terms and the gate. The scientific work stays with the lab that was approved to do it. The energy desk is Asked directly, Who can apply? OpenAI says principal investigators and lab leads at U.S. universities. Labs without an enterprise agreement apply through the trusted-access process. Labs that already have an enterprise agreement use that path.Energy.

The line to keep

A sponsored seat plan for U.S. academic labs, capped at 30 people and 12 months, with GPT-Rosalind behind a review and without API access. The Decoder confirms the cap. OpenAI sets the terms. The short answer this page will stand on is this: Up to 30 Business Standard seats at no charge for up to 12 months, for eligible labs at U.S. universities, plus other commercial frontier models within plan limits. GPT-Rosalind itself still requires trusted-access approval. OpenAI says the API is not included and that plan data is not used for training by default.