OpenAI unveiled its new artificial intelligence model on Thursday named GPT-Rosalind, an advanced reasoning model designed and built specifically to support biosciences research. This historic step represents the company’s first strategic and dedicated entry into the field of scientific AI tools, with a primary focus on critical sectors such as genomics, drug discovery processes, and complex protein engineering technologies. This prominent launch coincided with the announcement of a major and pivotal expansion of the company’s programming agent known as Codex, which has been enhanced with over 90 new plugins in a separate announcement released on the same day.
A scientific model inspired by a pioneering legacy
The company named its new model after the prominent British scientist and chemist Rosalind Franklin, whose groundbreaking work in X-ray crystallography played a pivotal and essential role in discovering the double-helix structure of DNA. GPT-Rosalind was specifically designed to assist researchers and scientists in navigating vast amounts of scientific data, breaking down barriers between isolated disciplines that have long slowed the pace of progress in modern biology and medicine. According to reports, this advanced model aims to help the scientific community extract precise insights from massive datasets and translate theoretical studies and research into tangible practical applications in the healthcare sector.
Accelerating the pace of drug discovery
Drug discovery and development is one of the most complex, time-consuming, and resource-intensive processes in the medical sector. In this context, the OpenAI team explained current challenges during a press conference, stating: “On average, it takes between 10 to 15 years to move from the target discovery phase to obtaining regulatory approval for new drugs in the United States, and only one out of every 10 drugs that enter clinical trials ultimately receives final approval.” The team added: “We believe there is a real and exceptional opportunity to help researchers move faster to bypass some of the most complex and time-consuming parts of the scientific process pathway.”
Remarkable superiority over human experts
GPT-Rosalind demonstrated superior capabilities during benchmark tests measuring comprehension levels in protein, DNA, and chemical interaction fields, significantly outperforming the company’s existing models. In an independent evaluation involving practicing scientists, the model performed better than 95 percent of specialized human experts when predicting the functions of specific RNA sequences. Alongside this standalone model, the company is also launching a biosciences research assistant plugin that connects the model to more than 50 trusted scientific databases, allowing researchers to search protein structures, review DNA sequences, and examine research papers, all through a single simplified user interface.
Strict restrictions and biosafety concerns
Unlike the vast majority of products offered by OpenAI to the general public, GPT-Rosalind will not be available for widespread use. The company has decided to limit access to it, restricting it to pre-vetted, trusted research organizations due to concerns regarding potential biosafety risks. The initial partner list includes prominent institutions such as Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific. This restricted rollout places the new model alongside other recent models launched by the company with limited permissions, such as the GPT-5.4-Cyber model, which is available only to authorized security researchers, reflecting a clear shift in company policy toward subjecting powerful AI technologies to rigorous approval processes.
Massive investments and future vision
This launch comes just days after reports circulated regarding a new policy paper prepared by the company, which calls for a broader integration of artificial intelligence technologies in the biosciences sector, demands greater access to medical data, and stresses the need to recognize advanced AI as a national research asset. In a related context, OpenAI separately announced its ambitious plans to spend at least $1 billion this year to support AI-related causes, with biosciences and finding cures for intractable diseases topping its list of primary priorities and interests.
Codex transforms into a comprehensive engineering agent
On another front, and on the same day filled with technical announcements, the company revealed a wide-ranging update to its cloud-based programming agent known as Codex. This update included the launch of more than 90 new plugins, featuring programming connectors for popular working environments such as CircleCI, GitLab, and the Microsoft Suite. This expansion represents an important step that elevates Codex from a simple code completion tool into a versatile, comprehensive software engineering agent capable of gathering complex contexts and taking appropriate actions across various developer workflows smoothly and professionally.
Frequently asked questions
What is the GPT-Rosalind model?
It is an advanced AI reasoning model developed by OpenAI, designed specifically to assist researchers and scientists in biosciences fields such as genomics, drug discovery, and protein engineering.
Why did the company restrict access to this advanced model?
The company restricted access to the model, making it exclusive to certified research institutions, due to biosafety concerns and to ensure the technology is used in safe and reliable environments.
How can the new model accelerate the drug discovery process?
The model can process massive scientific data at lightning speed and connect it to more than 50 specialized databases, cutting down time and effort in the early stages of drug discovery that typically take 10 to 15 years.
What are the most prominent updates received by the Codex programming tool?
The Codex tool received a massive update including over 90 new plugins, allowing it to integrate with workflow platforms like GitLab and CircleCI, transforming it from a simple code completion tool into an integrated software engineering agent.
