Named after the crystallographer who helped reveal the structure of DNA, GPT-Rosalind is OpenAI’s inaugural domain-specific model series, tailored for biochemistry, genomics, and protein engineering. Access is limited to a trusted-access program for vetted enterprise clients including Amgen, Moderna, and Thermo Fisher Scientific.
OpenAI announced the launch of GPT-Rosalind, a cutting-edge reasoning model created for life sciences research, on Thursday.
The model aids in evidence synthesis, hypothesis generation, experimental planning, and complex scientific workflows in biochemistry, genomics, and protein engineering, marking OpenAI’s first domain-specific model series.
It is available as a research preview through ChatGPT, Codex, and the OpenAI API, although access is limited to a trusted-access program for approved enterprise clients in the United States.
The model bears the name of Rosalind Franklin, the British chemist and X-ray crystallographer whose DNA diffraction imaging was key to unveiling the double helix structure, a contribution overlooked in the 1962 Nobel Prize awarded to Watson, Crick, and Wilkins.
This naming serves as recognition: Franklin’s work is now seen as foundational to modern molecular biology and a reference point in discussions about the erasure of women from scientific history.
OpenAI presents GPT-Rosalind as a tool to reduce the timeline from scientific ideas to clinical evidence. Currently, it takes about 10 to 15 years to progress a drug from target discovery to regulatory approval in the United States.
GPT-Rosalind is positioned for early-stage assistance: querying specialized databases, parsing scientific literature, interacting with computational tools, and suggesting new experimental pathways within one interface.
In addition to the model, OpenAI is launching a Life Sciences research plugin for Codex, linking models to over 50 scientific tools and data sources, providing researchers with programmatic access to biological databases and computational pipelines.
Launch partners include Amgen, Moderna, Thermo Fisher Scientific, and the Allen Institute. OpenAI collaborates with Los Alamos National Laboratory on AI-guided protein and catalyst design.
Benchmark performance from OpenAI shows GPT-Rosalind achieving a 0.751 pass rate on BixBench, a bioinformatics benchmark from Edison Scientific that tests models on real-world computational biology tasks.
On LABBench2, a broader research task benchmark, the model surpassed GPT-5.4 on six of eleven tasks, with a notable advantage in CloningQA, which involves the end-to-end design of reagents for molecular cloning protocols.
A prominent performance signal came from a third-party evaluation with Dyno Therapeutics, a gene therapy company focused on designing AAV capsid proteins.
GPT-Rosalind was tested on sequence-to-function prediction and sequence generation tasks using unpublished, previously unseen RNA sequences to prevent benchmark contamination.
According to OpenAI, the top model submissions ranked above the 95th percentile of human experts on the prediction task and around the 84th percentile on sequence generation, confirmed by multiple outlets covering the launch.
The launch has significant dual-use implications, which OpenAI has addressed through its access model.
Researchers have warned that AI models trained on biological data could be misused to design dangerous pathogens.
OpenAI’s choice to restrict access to a vetted trusted-access program, requiring organizations to show they work toward improving human health outcomes and maintain strong security and governance controls, directly addresses that risk. During the research preview phase, usage will not affect existing API credits.