OpenAI’s New AI Model Rosalind Could Shave Years Drug-Free You may not use it

OpenAI has just named its first AI platform after all Rosalind Franklin-British scientist whose X-ray work helped reveal DNA’s double helix, and who was denied credit for the rest of his life.

GPT-Rosalind, to be revealed on Thursdayand a conceptual framework for design biology, drug discovery, and translational medicine. It’s the first of what OpenAI is calling a series of Life Sciences models—a direct play on the market where many specialized labs from universities to Google DeepMind are all vying for titles.

Getting a drug from research to regulatory approval in the US takes time 10 to 15 years on average according to experts .. It often disappears not in the eureka moment, but in grinding: sorting thousands of papers, querying databases, making reagents, and interpreting ambiguous results. This is what GPT-Rosaling is trying to address.

OpenAI claims that this model can force the initialization process. As the company says, GPT-Rosalind is designed to help scientists “explore more possibilities, surface connections that might be missed, and arrive at better ideas sooner.”

The signs confirm some of the claims. On BixBench—a benchmark built around real-world bioinformatics applications—GPT-Rosalind achieved a score of 0.751, the highest score among models with published results. On LABBench2, it outperformed the default GPT-5.4 on six out of eleven tasks.

GPT-Rosalind Beats GPT 5.4 in almost every aspect of life science, but it’s a unique model that can’t handle anything other than that.

OpenAI also announced Dyno Therapeutics will help test and test his model based on unpublished RNA sequences to prevent memorization. GPT-Rosalind’s top ten contributions are above the 95th percentile of humanists on sequential prediction tasks, and around the 84th percentile on generation.

That said, the OpenAI life science research led by Joy Jiao was tested on what the model could do. He explained that the company does not see Rosalind as a model that can create a new random treatment, but told reporters that it could be useful in accelerating research. “We think there is a real opportunity to help researchers move faster through the most complex and time-consuming parts of science,” Jiao said in a press release. LA Times.

The environment surrounding the brand can be as important as the brand itself. OpenAI is also releasing a free Life Sciences extension for Codex linking scientific databases with more than 50 tools, protein analysis, sequence searching, literature review, genomics pipelines. Business users with GPT-Rosalind access get a high-quality view. Everyone else gets a plugin with default colors.

OpenAI has enlisted a list of pharma and biotech customers to launch, including Amgen, Moderna, and Thermo Fisher Scientific. Separately, it is developing a research partnership with Los Alamos National Laboratory on AI-guided proteins and artificial intelligence.

“Life science requires precision at every step. The questions are very complex, the data is very unique, and the stakes are very complex,” said Sean Bruich, senior VP of AI and Data at Amgen in an official announcement.

Rosalind’s access is deliberately restricted. This brand is exclusive to US companies, behind due diligence and security. The concern is not clear: an international coalition of more than 100 scientists has already called for stricter controls on the natural resources used to train AI, citing the dangers of the design of pathogens. OpenAI’s restricted release is a direct response. In preview research, the application will not affect existing APIs.

This isn’t OpenAI’s first scientific move either. The Prism science writing for the workplace launched in January was the first step. GPT-Rosalind is a sharp follow-up, very special – and a sign that well-known brands are staying ahead of the competition.

No cure has been found by AI it has been fixed Tests for phase 3. That number is still zero. But if GPT-Rosalind enables a researcher to develop a better test six months faster in thousands of labs, then the integration of the findings, and when, could be the whole ball. That’s the real idea here, and it’s worth a closer look.