OpenAI Launches GPT-Rosalind for Life Sciences and Drug Discovery

OpenAI has officially introduced GPT-Rosalind, its first specialized AI model designed specifically for the life sciences. Named in honor of the pioneering chemist Rosalind Franklin, the model is built to provide foundational reasoning in fields such as biochemistry, genomics, and protein engineering. By moving beyond the capabilities of general-purpose language models, GPT-Rosalind aims to accelerate the time-intensive stages of drug discovery and biological research.

Enhancing Scientific Workflows

Drug discovery is a historically slow and expensive process, often requiring 10 to 15 years to move from initial target discovery to regulatory approval. Much of this duration is consumed by analytical tasks, including literature reviews, reagent design, and the interpretation of complex biological data. GPT-Rosalind is engineered to assist researchers with these multi-step workflows by supporting hypothesis generation, experimental planning, and evidence synthesis.
The model functions as a specialized research assistant capable of querying databases, parsing scientific literature, and interacting with computational tools within a single interface. To further support these tasks, OpenAI is launching a Life Sciences research plugin for Codex, which provides programmatic access to over 50 scientific tools and data sources, allowing researchers to integrate the model directly into their existing computational pipelines.

Benchmark Performance and Real-World Application

OpenAI has validated the model’s capabilities through rigorous benchmarking. GPT-Rosalind achieved a 0.751 pass rate on BixBench, a standard for evaluating bioinformatics and data analysis tasks such as genomic output interpretation and sequencing data processing. Furthermore, on the LABBench2 evaluation, the model outperformed GPT-5.4 on six out of eleven tasks, demonstrating significant proficiency in the end-to-end design of reagents for molecular cloning.
In a practical application with Dyno Therapeutics, the model was tested on RNA sequence-to-function prediction using unpublished sequences. The results were notable: the model’s best-of-ten submissions ranked above the 95th percentile of human experts for prediction tasks and reached the 84th percentile for sequence generation, proving its effectiveness on novel biological data.

A Controlled Launch for Research

Access to GPT-Rosalind is currently restricted through a trusted-access program for qualified enterprise customers in the United States. OpenAI has implemented technical safeguards and usage limits to ensure the model is used for legitimate life sciences research and the improvement of human health outcomes.
The company is already collaborating with several prominent organizations, including Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific, to integrate the model into their research workflows. Additionally, OpenAI is partnering with the Los Alamos National Laboratory to utilize the model for the AI-guided design of proteins and catalysts. This launch represents a strategic shift toward domain-specific AI, applying specialized training strategies to address the high-dimensional data and complex search spaces inherent in the life sciences.