OpenAI on Thursday unveiled GPT-Rosalind, an artificial intelligence model developed to aid research efforts in biology, drug discovery and translational medicine.
The company says the model is optimized for scientific workflows, combining the ability to use external tools with domain knowledge that covers chemistry, protein engineering and genomics. According to OpenAI, the typical timeline from target discovery to regulatory approval for a new drug in the United States is roughly 10 to 15 years.
GPT-Rosalind is being offered as a research preview across ChatGPT, Codex, and via the API to qualified customers participating in OpenAI’s trusted access program. In parallel, OpenAI is releasing a Life Sciences research plugin for Codex that connects models to more than 50 scientific tools and data sources.
OpenAI is working with a set of industry and research partners to test GPT-Rosalind across laboratory and computational workflows. Named after Rosalind Franklin, whose research contributed to understanding DNA structure, the model is being applied in collaborations with Amgen (NASDAQ:AMGN), Moderna (NASDAQ:MRNA), the Allen Institute, and Thermo Fisher Scientific.
In benchmark evaluations, GPT-Rosalind recorded a 0.751 pass rate on BixBench, a test suite focused on bioinformatics and data analysis. On LABBench2, which assesses performance on tasks such as literature retrieval and protocol design, the model outperformed GPT-5.4 on six out of eleven tasks.
OpenAI also described a joint evaluation with Dyno Therapeutics in which the model was assessed on RNA sequence-to-function prediction using unpublished sequences. In that exercise, best-of-ten model submissions ranked above the 95th percentile of human experts on the prediction task and around the 84th percentile on the sequence generation task.
The rollout is structured as a trusted-access deployment for qualified Enterprise customers in the United States. Eligible organizations must be conducting scientific research that provides public benefit and must maintain governance and safety oversight controls as part of their use.
OpenAI’s announcement positions GPT-Rosalind as a tool intended to integrate into established research workflows, with early partners and benchmark results serving as initial indicators of capability. Availability and use will be limited to entities that meet the company’s trusted-access criteria.
Key points
- OpenAI introduced GPT-Rosalind to support biology, drug discovery and translational medicine research.
- The model is available as a research preview in ChatGPT, Codex and the API for qualified customers through a trusted-access program; a Codex Life Sciences plugin links to over 50 scientific tools and data sources.
- Early evaluations show a 0.751 pass rate on BixBench and better performance than GPT-5.4 on six of eleven LABBench2 tasks; partners include Amgen and Moderna.
Risks and uncertainties
- Access is limited to qualified Enterprise customers under a trusted-access structure, which constrains immediate broad adoption in research and commercial settings - this affects the biotech and pharmaceutical sectors.
- Benchmarks show mixed results across tasks, with GPT-Rosalind outperforming prior models on some but not all LABBench2 tasks, indicating varying strengths across different research activities - this affects computational biology and lab automation users.
- Deployment requires organizations to maintain governance and safety oversight controls, creating procedural and compliance hurdles for institutions seeking to integrate the model into workflows - this impacts research institutions and corporate R&D teams.