GPT-Rosalind
Last reviewed
Jun 7, 2026
Sources
16 citations
Review status
Source-backed
Revision
v1 · 1,661 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 7, 2026
Sources
16 citations
Review status
Source-backed
Revision
v1 · 1,661 words
Add missing citations, update stale details, or suggest a clearer explanation.
GPT-Rosalind is a frontier reasoning model purpose-built for the life sciences, developed by OpenAI and first released in mid-April 2026. Named after the British chemist and X-ray crystallographer Rosalind Franklin, whose diffraction work helped reveal the structure of DNA, the model is designed to reason across biology, scientific evidence, data, and tools to accelerate early-stage research in drug discovery, genomics, and translational medicine. It is the first domain-specific entry in OpenAI's Life Sciences model series and is distributed not as a general consumer product but through a gated "trusted access" program available to qualified organizations with legitimate biology research use cases. GPT-Rosalind is offered across ChatGPT, the OpenAI API, and the company's Codex agentic coding environment, and is paired with a freely available Life Sciences research plugin that connects models to more than 50 scientific tools and databases. The launch positioned OpenAI to compete in scientific AI alongside efforts such as Google DeepMind's AlphaFold, and reflects a broader 2026 push by the company into AI for science and AI in healthcare.
GPT-Rosalind is a specialized large language model optimized for scientific workflows rather than open-ended chat. OpenAI describes it as built "for modern scientific work across published evidence, data, tools, and experiments," with strength in reasoning over molecules, proteins, genes, pathways, and disease-relevant biology. The model targets multi-step research tasks such as target discovery and validation, genomics interpretation, pathway analysis, literature synthesis, sequence-to-function interpretation, experimental planning, and data analysis. OpenAI frames the system as a tool to help scientists "explore more possibilities, surface connections that might otherwise be missed, and arrive at better hypotheses sooner," while emphasizing that humans remain in the loop and that the model is meant to synthesize evidence and generate hypotheses rather than replace expert judgment or real-world experimental validation.
The model name honors Rosalind Franklin (1920-1958), whose rigorous X-ray crystallography helped establish the molecular structures of DNA, RNA, and viruses and laid foundations for modern molecular biology. The choice reflects OpenAI's positioning of the model as a tool for rigorous, evidence-driven discovery. GPT-Rosalind followed earlier OpenAI scientific efforts, including the Prism scientific writing platform launched in January 2026, and arrived amid a wave of generative AI products aimed at the research enterprise.
| Attribute | Detail |
|---|---|
| Developer | OpenAI |
| Initial release | Mid-April 2026 (research preview) |
| Capabilities update | June 3, 2026 (GPT-5.5-based) |
| Model type | Frontier reasoning model for life sciences |
| Namesake | Rosalind Franklin, DNA-structure crystallographer |
| Domains | Drug discovery, genomics, protein engineering, translational medicine |
| Availability | ChatGPT, API, and Codex via trusted-access program |
| Companion tool | Free Life Sciences research plugin for Codex (50+ tools and databases) |
| Launch partners | Amgen, Moderna, the Allen Institute, Thermo Fisher Scientific |
| Additional collaborators | Dyno Therapeutics, Los Alamos National Laboratory, Novo Nordisk |
| Research lead | Joy Jiao (Life Sciences Research Lead) |
| Product lead | Yunyun Wang (Life Sciences Product Lead) |
| Biodefense program | Rosalind Biodefense (launched May 2026) |
OpenAI introduced GPT-Rosalind in mid-April 2026 (Axios reported the launch on April 16, with several trade outlets covering it April 17-20), publishing a dedicated announcement and an OpenAI Help Center article describing the model and its access program. The company named launch customers Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific, with reporting also citing collaborations with Dyno Therapeutics and ongoing work at Los Alamos National Laboratory on AI-guided protein and catalyst design. The launch came the same week OpenAI announced a separate strategic alliance with Novo Nordisk, underscoring a concerted move into pharmaceutical and biotech partnerships.
Industry partners emphasized the model's fit for high-stakes, data-intensive research. Sean Bruich, Amgen's head of AI and data, said the life sciences field "demands precision at every step," noting that "the questions are highly complex, the data is highly unique, and the stakes are incredibly high." OpenAI's life sciences leads, research lead Joy Jiao and product lead Yunyun Wang, framed the system as a way to help researchers "move faster through some of the most complex and time-intensive parts of the scientific process." OpenAI argued that gains made at the earliest stages of discovery compound downstream in better target selection and more efficient research programs, a core part of the value proposition for AI drug discovery.
Access is deliberately restricted. At launch GPT-Rosalind was limited to qualified, largely U.S. enterprise customers behind a qualification and safety review, with the research preview not consuming standard API credits. By the June 2026 update, OpenAI said the model was available in research preview to eligible organizations globally through its trusted-access deployment structure, while still requiring legitimate biology research use cases and compliance with safety requirements. Alongside the model, OpenAI released a free Life Sciences research plugin for Codex that connects to more than 50 scientific databases and tools, available more broadly than the gated model itself.
OpenAI reported that GPT-Rosalind delivered the best performance among published models on a range of biology and chemistry reasoning tasks at launch. On BixBench, a benchmark for biology data-analysis reasoning, the model posted a 0.751 pass rate, described as the top score among published models. On the LAB-Bench family of evaluations it was reported to outperform the prior GPT-5.4 model on 6 of 11 tasks. In an evaluation using unpublished RNA sequences, GPT-Rosalind's best-of-ten submissions ranked above the 95th percentile of human experts on prediction tasks and reached the 84th percentile for sequence generation, figures OpenAI used to illustrate expert-level performance on novel data.
On June 3, 2026, OpenAI announced new capabilities that rebuilt GPT-Rosalind on top of GPT-5.5, combining that model's agentic coding and tool-use abilities with deeper domain intelligence in core drug-discovery areas such as medicinal chemistry and genomics. OpenAI introduced new evaluations alongside the update, including GeneBench (an agentic evaluation of long-horizon, end-to-end genomics and quantitative-biology analysis), MedChemBench, LifeSciBench (also referred to as Life Sciences Bench), and a wet-lab troubleshooting evaluation reported as LabWorkBench. On GeneBench, the updated model was reported to use 31 percent fewer tokens than GPT-5.5 while achieving higher accuracy of 21.6 percent versus 20.4 percent, and OpenAI said it led GPT-5.5, Grok 4.3, and Gemini 3.1 Pro on overall LifeSciBench scores. The update claimed broad gains across biology-expert research tasks, complex medicinal-chemistry queries, quantitative biology, and wet-lab troubleshooting.
The model's intended workflow is agentic and tool-augmented: rather than answering single questions, it is designed to plan and execute multi-step research, calling external scientific tools and databases through the Codex plugin to interpret sequences, analyze pathways, review literature, and propose experiments. This positions GPT-Rosalind as a reasoning layer over the scientific software stack rather than a replacement for specialized structure-prediction or simulation systems.
GPT-Rosalind marked one of the most concrete steps by a leading foundation-model developer into specialized scientific AI, extending the reach of general-purpose generative AI into the regulated, high-stakes world of biomedical research. By gating the model behind a trusted-access program and pairing it with vetted enterprise partners, OpenAI signaled an approach that treats frontier biological capability as dual-use, requiring qualification and safety review rather than open release. Commentators noted that the model could compress timelines in early drug discovery, though access restrictions meant most researchers could not use it directly at launch.
The biosecurity dimension became explicit in May 2026, when OpenAI launched Rosalind Biodefense, an initiative for defensive applications of AI in the life sciences. The program operates two tracks: one sponsoring vetted outside developers building pandemic-preparedness and biosecurity tools (in areas such as epidemiological modeling, early detection, screening, and non-pharmaceutical interventions), and another providing access to select U.S. government and allied partners. OpenAI said it briefed the White House and several federal agencies on the approach, and named collaborators including Lawrence Livermore National Laboratory (pairing the model with supercomputing to design and evaluate medical countermeasures), the Johns Hopkins Applied Physics Laboratory (integrating GPT-Rosalind into protein-engineering platforms to screen mutant enzymes), and the Coalition for Epidemic Preparedness Innovations (applying it to vaccine development, including work related to a Bundibugyo Ebola outbreak in the Democratic Republic of the Congo and Uganda). OpenAI framed the effort as "defensive acceleration," arguing that frontier AI "should meaningfully advantage those defenders." The announcement drew attention in part because it came shortly after the U.S. administration postponed an executive order that would have created government review processes for powerful AI models before release.
Taken together, the launch and rapid follow-up positioned GPT-Rosalind as a flagship of OpenAI's 2026 AI for science strategy, a competitor in scientific AI alongside structure-prediction systems, and a test case for how frontier AI in healthcare and biology can be deployed responsibly under access controls.