GPT-Rosalind: Accelerating Life Sciences & Drug Discovery with AI

GPT-Rosalind: A New Era for Life Sciences & Drug Discovery with AI

Today marks a pivotal moment for scientific innovation as OpenAI introduces GPT-Rosalind, its groundbreaking frontier reasoning model engineered specifically for life sciences research. This purpose-built AI is poised to revolutionize fields spanning biology, drug discovery, and translational medicine, promising to dramatically accelerate the pace of scientific advancement. Named in honor of Rosalind Franklin, whose pioneering work illuminated the structure of DNA, GPT-Rosalind embodies a commitment to foundational scientific inquiry, now supercharged by advanced artificial intelligence.

The journey from target discovery to regulatory approval for a new drug is notoriously arduous, typically consuming 10 to 15 years in the United States. This protracted timeline is a testament not only to the inherent difficulty of the science but also to the complex, often fragmented nature of research workflows. Scientists must meticulously navigate vast volumes of literature, specialized databases, experimental data, and evolving hypotheses. GPT-Rosalind is designed to be a catalyst in this intricate process, providing a powerful assistant that can synthesize evidence, generate novel hypotheses, and plan experiments with unprecedented efficiency and depth. By streamlining these early, critical stages of discovery, the model aims to compound downstream gains, leading to better target selection, stronger biological hypotheses, and higher-quality experiments, ultimately fostering breakthroughs that would otherwise remain out of reach.

GPT-Rosalind is now available as a research preview within ChatGPT, Codex, and the API, accessible to qualified customers through a trusted access program. Further democratizing access to AI-powered research, OpenAI is also releasing a freely accessible Life Sciences research plugin for Codex, enabling scientists to connect the models to over 50 scientific tools and data sources. This dual approach ensures both specialized, secure deployment for advanced research organizations and broader utility for the wider scientific community, propelling us towards a future where AI is an indispensable partner in the quest for human health.

Engineered for Advanced Scientific Workflows

The GPT-Rosalind life sciences model series represents a paradigm shift in how AI can integrate with modern scientific work, seamlessly operating across published evidence, complex data sets, diverse tools, and ongoing experiments. OpenAI's robust compute infrastructure underpins this capability, enabling continuous training and refinement of increasingly sophisticated domain models against real-world scientific tasks. This ensures GPT-Rosalind remains at the cutting edge as scientific workflows themselves evolve in complexity.

In rigorous evaluations, GPT-Rosalind has demonstrated best-in-class performance on tasks demanding deep reasoning over molecules, proteins, genes, pathways, and disease-relevant biology. Its efficacy extends to the practical application of scientific tools and databases within multi-step workflows, including comprehensive literature review, intricate sequence-to-function interpretation, strategic experimental planning, and nuanced data analysis. This initial release in the GPT-Rosalind series marks the beginning of a long-term commitment to enhancing the model's biochemical reasoning capabilities across even more tool-heavy and long-horizon scientific endeavors. OpenAI is actively collaborating with leading organizations such as Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific to embed GPT-Rosalind into workflows that drive transformative discovery.

Unprecedented Performance in Benchmarks and Real-World Applications

The capabilities of GPT-Rosalind have been thoroughly evaluated across a spectrum of challenges fundamental to scientific discovery and industry research. These assessments measure core reasoning across diverse scientific subdomains, including the intricacies of chemical reaction mechanisms, the understanding of protein structure, mutation effects, and interactions, and the phylogenetic interpretation of DNA sequences. Beyond theoretical reasoning, the evaluations also gauge the model's ability to support real-world research by interpreting experimental outputs, identifying expert-relevant patterns, and synthesizing external information to design subsequent experiments. Crucially, GPT-Rosalind's proficiency in selecting and utilizing the appropriate computational tools, databases, and domain-specific capabilities to augment its reasoning has been a key focus, demonstrating its practical utility throughout the end-to-end scientific research process.

In public benchmarks, GPT-Rosalind has consistently demonstrated superior performance. On BixBench, a benchmark specifically designed around real-world bioinformatics and data analysis challenges, GPT-Rosalind achieved leading performance among models with published scores.

Performance evaluated against other models with available access.

Furthermore, on LABBench2, which assesses a range of research tasks such as literature retrieval, database access, sequence manipulation, and protocol design, GPT-Rosalind outperformed GPT-5.4 on 6 out of 11 tasks. A particularly notable improvement was observed in CloningQA, a task demanding end-to-end design of DNA and enzyme reagents for molecular cloning protocols. The model's real-world impact was further validated through a partnership with Dyno Therapeutics, a company pioneering AI-designed gene therapies. In an evaluation using unpublished, uncontaminated RNA sequences, GPT-Rosalind's best-of-ten model submissions, when assessed directly in the Codex app, ranked above the 95th percentile of human experts on the prediction task and around the 84th percentile on the sequence generation task. These comprehensive evaluations underscore GPT-Rosalind's robust capability to generate evidence, analyze complex data, and drive defensible biological conclusions in the hands of scientists. For advanced usage with Codex, researchers might find the Codex prompting guide helpful in maximizing GPT-Rosalind's potential.

Bridging AI with Existing Scientific Tools: The Life Sciences Plugin

A cornerstone of GPT-Rosalind's utility is its seamless integration with the existing ecosystem of scientific tools. OpenAI has launched a new Life Sciences research plugin for Codex, now available on GitHub. This comprehensive package comprises a broad set of modular skills meticulously designed for the most common research workflows across various disciplines, including human genetics, functional genomics, protein structure, biochemistry, clinical evidence, and public study discovery.

This plugin acts as a crucial orchestration layer, empowering scientists to tackle broad, ambiguous, and multi-step questions more effectively. It provides direct access to over 50 public multi-omics databases, a wealth of literature sources, and numerous biology tools. This rich integration offers a flexible starting point for common, repeatable workflows such as protein structure lookup, sequence search, extensive literature review, and public dataset discovery. While eligible Enterprise users can leverage this plugin in research workflows with GPT-Rosalind for deeper biological reasoning, all users can utilize the plugin package with OpenAI's mainline models, democratizing access to AI-enhanced life sciences research. This flexibility ensures that a wide array of researchers can benefit from AI's power, whether using specialized models or more general-purpose AI. Learn more about maximizing your AI tools with guides like using Codex with your ChatGPT plan.

Secured Access for Responsible Innovation

Recognizing the profound implications of advanced AI in life sciences, OpenAI has implemented a stringent trusted-access deployment structure for GPT-Rosalind. This program is initially available for qualified Enterprise customers in the U.S., featuring robust controls around eligibility, access management, and organizational governance. This cautious approach ensures that these powerful capabilities are made available to scientists and research organizations best positioned to advance human health, while simultaneously maintaining strong safeguards against potential biological misuse.

The Life Sciences model has been developed with heightened enterprise-grade security controls and strengthened access management, making it suitable for professional scientific use in governed research environments. OpenAI evaluates access based on three core principles: ensuring beneficial use in legitimate scientific research with clear public benefit; mandating appropriate governance, compliance, and misuse-prevention controls; and guaranteeing controlled access within secure, well-managed environments for approved users. Organizations must also comply with OpenAI’s usage policies and the specific life sciences research preview terms. During this research preview phase, use of GPT-Rosalind will not consume existing credits or tokens, subject to abuse guardrails. For organizations prioritizing data security, understanding concepts like enterprise privacy is crucial when integrating advanced AI models.

To facilitate seamless integration and maximize impact, OpenAI's dedicated Life Sciences team, supported by advisory partners including McKinsey & Company, Boston Consulting Group (BCG), and Bain & Company, assists organizations in identifying high-impact use cases, integrating the model into enterprise environments, and driving measurable outcomes.

The Future of AI in Biological Discovery

The introduction of GPT-Rosalind is just the first release in OpenAI's ambitious Life Sciences model series. This launch signifies the beginning of a long-term commitment to building advanced AI that can profoundly accelerate scientific discovery in areas of critical importance to society, from human health to broader biological research. OpenAI is dedicated to continuously improving the model's biological reasoning capabilities, further expanding its support for tool-heavy and long-horizon scientific workflows.

As AI models continue to evolve, their ability to transform complex scientific challenges will only grow. GPT-Rosalind represents a significant leap forward, offering scientists a powerful new ally in their quest to unravel nature's mysteries and develop life-saving innovations. The era where AI acts not just as a text generator but as a true execution interface, capable of driving tangible research outcomes, is truly upon us. This journey underscores OpenAI’s vision for a future where AI empowers humanity to achieve scientific milestones that once seemed impossible.