OpenAI’s new AI model, Rosalind, could shave off years of drug discovery that you otherwise might not otherwise be able to use

OpenAI has just named its first domain-specific AI model after itself Rosalind Franklin– The British chemist whose work in X-ray crystallography helped uncover the DNA double helix, and who was famously denied credit for it during her lifetime.

GPT-Rosalind, revealed Thursdayis a logic model specifically designed for biology, drug discovery, and translational medicine. It’s the first in what OpenAI calls its Life Sciences Model Series, a direct play on a market where many specialized labs from universities to Google DeepMind are vying for position.

It takes a drug from target discovery to regulatory approval in the United States From 10 to 15 years On average, according to experts, most of that time is spent not in moments of discovery, but in the grinding process: analyzing thousands of papers, querying databases, designing reagents, and interpreting ambiguous results. This is what GPT-Rosaling is trying to address.

OpenAI argues that the model can compress this early-stage work. As the company said, GPT-Rosalind is designed to help scientists “explore more possibilities, superficial connections that might be missed, and come up with better hypotheses sooner.”

The standards support at least some of this ambition. In BixBench — a benchmark built on real-world bioinformatics tasks — GPT-Rosalind recorded a success rate of 0.751, the highest score among models with published results. In LABBench2, it outperformed its GPT-5.4 predecessor in six out of eleven tasks.

GPT-Rosalind outperforms GPT 5.4 in every life sciences situation, but it is a very specific model that will perform poorly in anything other than that.

As announced by OpenAI Dyno treatments It will help test and evaluate his model based on unpublished RNA sequences to rule out conservation. The best submissions from GPT-Rosalind were ranked above the 95th percentile of human experts in sequence prediction tasks, and around the 84th percentile in generation.

However, Joy Jiao, life sciences research lead at OpenAI, was measured about what the model could actually do. She explained that the company does not see Rosalind as a model capable of creating new treatments independently, but she told reporters that she could be a great help in accelerating research. “We believe there is a real opportunity to help researchers move faster through some of the most complex and time-consuming parts of the scientific process,” Jiao said at a press conference. Los Angeles Times.

The ecosystem surrounding the model can be as important as the model itself. OpenAI is also releasing a free life sciences research plug-in for Codex that connects to more than 50 scientific databases and tools — protein structure searches, sequence searches, literature reviews, and genomics pipelines. Enterprise users with access to GPT-Rosalind get the reasoning layer on top. Everyone gets the plugin with standard templates.

OpenAI has lined up a list of pharma and biotechnology clients for the launch, including Amgen, Moderna and Thermo Fisher Scientific. Separately, it runs a research collaboration with Los Alamos National Laboratory on AI-guided protein and catalyst design.

“The life sciences field requires precision at every step. The questions are very complex, the data are very unique, and the stakes are incredibly high,” said Sean Bruitsch, Amgen’s senior vice president of artificial intelligence and data, in the official announcement.

Access to Rosalind is deliberately restricted. This form is intended for US institutions only, and is subject to a qualifications and safety review. The concern is not abstract: an international coalition of more than 100 scientists has already called for stricter controls on biological data used to train artificial intelligence, citing the risks of designing pathogens. OpenAI’s restricted rollout is a direct response. During search preview, usage will not consume existing API credits.

This is also not OpenAI’s first step into scientific workflow. the Prism scientific writing workspace Its launch in January was a first step. GPT-Rosalind is the most precise and specialized follow-up – a sign that domain-specific models are becoming a serious competitive frontier.

No drug discovery is entirely done by AI It has been erased Phase 3 trials. This number is still zero. But if GPT-Rosalind helps a researcher design a better trial six months faster across thousands of laboratories, the compounding effect on what is discovered and when could be the whole game. That’s the actual thesis here, and it’s worth watching closely.