Overview
This is an enterprise offering for life sciences R&D. It helps research teams synthesize evidence, reason across biological data, and coordinate workflows across scientific tools, literature, databases, internal systems, and specialist models.
GPT-Rosalind, OpenAI’s most capable model for life sciences research, is designed for advanced biological reasoning, scientific tool use, and multi-step workflows such as target discovery, target validation, genomics interpretation, pathway analysis, literature synthesis, and hypothesis generation.
The model is currently most useful for early discovery work, especially target biology, mechanism understanding, literature synthesis, and omics interpretation.
Organizations can request responsible access, or contact sales for more information.
Availability
GPT-Rosalind is currently available to eligible U.S. customers with Enterprise agreements. Organizations must have legitimate biology research use cases and meet safety and compliance requirements.
During the research preview, teams can access GPT-Rosalind in ChatGPT Enterprise and Codex through the model picker. Eligible organizations can also access the model through the OpenAI API for internal research tools, workflows, and applications. It is not available for customer-facing products or external commercial applications at this time.
Add users to the ChatGPT workspace
Before enabling GPT-Rosalind access, workspace owners or admins should make sure each approved user has been added to the GPT-Rosalind-enabled ChatGPT workspace.
To add users:
Switch to the GPT-Rosalind-enabled workspace and select Workspace settings.
In the left panel, select Members.
Select Invite member.
Enter the users’ email addresses or upload a CSV.
Choose the appropriate role and seat type, then send the invitations.
Once users have joined the workspace, continue with the steps below to enable GPT-Rosalind access through groups and custom roles.
Use GPT-Rosalind in ChatGPT
After access is enabled, users should make sure they are working in the GPT-Rosalind-enabled workspace.
To get started:
In the bottom-left corner of ChatGPT, open the workspace switcher and select the GPT-Rosalind workspace.
Select Skills and confirm that the provisioned skills are available.
Turn on the skills you plan to use.
Start a new chat.
Open the model picker, go to Legacy models, and select GPT-Rosalind.
GPT-Rosalind must be selected explicitly each time you want to use it.
Use GPT-Rosalind in Codex
To get started in Codex:
Sign in to Codex with the GPT-Rosalind workspace.
In the left navigation, select Plugins.
Install the Life Sciences Research Plugin.
Start a new thread.
Open the model picker and select GPT-Rosalind.
Enable GPT-Rosalind in ChatGPT
Workspace admins can enable GPT-Rosalind for approved users by using groups and custom roles.
To enable access:
Switch to your GPT-Rosalind-enabled workspace and select Workspace settings.
In the left panel, select Groups and create a group for GPT-Rosalind access.
Add the users who should have access to GPT-Rosalind.
In the left panel, select Permissions & roles and create a custom role.
Add the GPT-Rosalind group to the custom role.
In the role permissions, search for “model,” “skills,” and “Codex,” or scroll through the available permissions. Turn on Life Sciences and enable any required Skills and Codex permissions for GPT-Rosalind access.
Turn on Life Sciences.
Users in the group should now see GPT-Rosalind in the ChatGPT model picker.
Changes to role permissions may take a few minutes to apply.
Add users for API access
Eligible organizations can also access GPT-Rosalind through the OpenAI API for approved internal research tools, workflows, and applications.
To add members to your API organization:
Go to platform.openai.com.
Select the Settings icon in the top-right corner.
In the left panel, select People.
Add the member who needs API access.
For more information, see our guide to setting up your API organization.
GPT-Rosalind API usage will be free during the beta testing period. GPT-Rosalind API access is not available for customer-facing products or external commercial applications at this time.
Working with large research datasets in Codex
GPT-Rosalind does not need raw, large-scale datasets added directly into context to support in-depth analysis.
For omics data or other large datasets, point GPT-Rosalind to the files in a local directory, database, or approved storage system. You can then ask a specific research question, and GPT-Rosalind can help choose the right analysis, run targeted steps over the data, interpret the outputs, and turn useful workflows into repeatable skills.
This approach keeps raw files in storage and lets GPT-Rosalind perform focused, tool-driven analysis at query time. It is usually faster and more effective than trying to process the full dataset directly in context.
Data and security
GPT-Rosalind is delivered through ChatGPT Enterprise, Codex, and the API with enterprise security and governance controls.
The platform supports enterprise deployments with Regulated Workspaces and BAAs, along with SOC 2 Type 2 and HIPAA-aligned standards. OpenAI does not train on customer data.
The model includes role-based access controls for ChatGPT, a security and permissions model used to control access to systems or resources based on a user’s role assignments.
FAQ
What is GPT-Rosalind?
GPT-Rosalind is OpenAI’s most capable model for life sciences research. It is designed for deep scientific reasoning, stronger tool and database use, and safe access to multi-step workflows across areas such as protein understanding, genomics analysis, and biochemistry reasoning.
Who is GPT-Rosalind best for?
The model is best suited for bioinformaticians, computational biologists, and early discovery biologists.
What can GPT-Rosalind do in the research preview?
In the research preview, the model can help scientists reason across biological data, use tools and synthesize evidence from multiple sources, generate and refine hypotheses, and support multi-step workflows such as target discovery, target validation, and experiment planning.
What stage of research is it best for?
The model is currently most valuable for early discovery and hypothesis generation, especially for target biology, mechanism understanding, literature synthesis, and omics interpretation.
How is GPT-Rosalind different from GPT-5.4 in ChatGPT and Codex?
GPT-Rosalind is built for scientific research workflows rather than general productivity use cases. It offers deeper biological reasoning, literature synthesis workflows, longer reasoning for complex tasks, and access to scientific tools and databases for research tasks.
Can customers use their own internal knowledge and data?
Yes, this is directionally in scope, especially through Codex and custom workflows. It is not fully available out of the box today. What is possible depends on how data is structured, and some workflows may require custom setup, skills, or automation.
What does research preview participation involve?
Beta participants get access to GPT-Rosalind in their ChatGPT Enterprise and Codex workspaces. Teams can start using it for research workflows and are encouraged to share feedback to help shape the product.
Can eligible customers access GPT-Rosalind in the API?
Yes. Eligible organizations can access GPT-Rosalind through the OpenAI API for internal research tools, workflows, and applications. It is not available for customer-facing products or external commercial applications at this time.
Are Life Sciences plugins available in ChatGPT?
Not at launch. Life Sciences plugins are currently designed for Codex, where they can support more complex, agentic research workflows. We may expand plugin support to ChatGPT in the future.
How much does GPT-Rosalind cost?
During the research preview, use of GPT-Rosalind will not consume existing enterprise credits or paid tokens. Usage limits may apply. We’ll share more details on access, pricing, and availability as the program expands.
What data protections are in place?
The model is available inside ChatGPT Enterprise and Codex with enterprise-grade security and role-based access controls. OpenAI does not train on customer data.