OpenAI has introduced GPT-Rosalind, a frontier reasoning model built specifically for life science research. The model is named after chemist Rosalind Franklin, whose X-ray diffraction images from the 1950s became a key part of the evidence behind the discovery of DNA’s double-helix structure.
According to OpenAI, GPT-Rosalind is designed around three core domains: genomics, protein engineering, and chemistry. The company says the model can connect to scientific databases on its own, read recent journal papers, use scientific tools, and propose new experimental designs based on existing evidence. Those are some of the most time-consuming parts of research work, including literature review, hypothesis generation, and experiment planning. In the source material, OpenAI notes that a postdoctoral researcher may spend weeks reviewing enough papers to build a testable hypothesis, while AI could reduce that process to hours.
Beats GPT-5.4 in 6 of 11 LABBench2 tasks
OpenAI shared benchmark results showing GPT-Rosalind outperforming its own GPT-5.4 in 6 out of 11 tasks on LABBench2, a benchmark commonly used in the industry. The biggest improvement came in CloningQA, a task tied to molecular cloning protocols and the steps involved in copying and manipulating specific gene fragments.
The company’s framing is clear. GPT-Rosalind is not presented as a replacement for scientists, but as a system aimed at taking over the repetitive and labor-intensive parts of the workflow. The practical value lies in shortening the path from reviewing evidence to deciding what experiment should be run next.
Access restricted to qualified U.S. enterprise customers
For now, GPT-Rosalind is being released as a research preview, with access gated through OpenAI’s Trusted Access program. Eligibility is limited to qualified enterprise customers in the United States. To apply, organizations must be conducting scientific research tied to the public interest and must also have strong governance and safety oversight in place. Approved users can access the model through ChatGPT, Codex, and the API.
The first group of partners includes pharmaceutical company Amgen, mRNA vaccine developer Moderna, nonprofit research organization Allen Institute, and scientific instruments and reagents company Thermo Fisher Scientific. Moderna CEO Stéphane Bancel said the model can reason across complex biological evidence and help teams turn insights into experimental workflows.
Competition in life science AI is getting sharper
The release also adds to a broader race in life science AI. The source material says Anthropic acquired AI biotech startup Coefficient Bio for $400 million a few weeks ago, moving directly into the field, while Google is also building in the same area. Set against OpenAI’s earlier efforts to push for broader access to medical data and to frame compute as a national research resource in the U.S., GPT-Rosalind looks like a direct product step in that strategy.
Drug development can stretch across a decade and cost billions of dollars. Tools that reduce the time between hypothesis formation and experimental design are likely to attract serious attention from research organizations. GPT-Rosalind is entering at that exact point in the workflow.