OpenAI Rebuilds GPT-Rosalind on GPT-5.5 and Widens Access — Its Science Model Now Beats the General One on Biology

OpenAI has shipped a major upgrade to GPT-Rosalind, the specialized life-sciences model it first introduced in April, and is widening access to more research organizations. The new version is built on the reasoning and coding backbone of GPT-5.5 but layers on deep expertise in drug discovery, medicinal chemistry, genomics, and biological data analysis — and, OpenAI says, now outperforms the general-purpose GPT-5.5 on scientific work.

The model is named after Rosalind Franklin, the British scientist whose X-ray diffraction images were pivotal to decoding the structure of DNA. It is pitched not as a chatbot but as a research collaborator: GPT-Rosalind reasons across molecules, proteins, genes, pathways, and disease biology, and chains multi-step workflows such as literature review, sequence-to-function interpretation, experimental design, lab troubleshooting, and analysis of large biological datasets.

To put numbers behind the claims, OpenAI introduced a new evaluation it calls LifeSciBench, which scores models across areas including evidence evaluation, scientific reasoning, data analysis, experimental validation, and communication. The company says GPT-Rosalind beat both GPT-5.5 and competing models on the benchmark. It reports stronger reasoning around drug optimization and toxicity prediction in medicinal chemistry, and higher accuracy on genomics and quantitative biology tasks while using fewer computational resources than its predecessor.

Access remains deliberately gated. GPT-Rosalind runs through a research-preview program under OpenAI's controlled-access framework, open to organizations doing legitimate scientific work rather than the general public — a posture that reflects the dual-use sensitivity of frontier biology models. The launch cohort already includes heavyweight partners such as Moderna, Amgen, Thermo Fisher Scientific, and the Allen Institute, with Danish pharmaceutical giant Novo Nordisk exploring the model to accelerate analysis of complex scientific data across R&D.

The upgrade lands amid a broader pivot toward "science-first" AI models tuned for specific domains rather than general assistance. Moderna's chief executive framed GPT-Rosalind as "an important step in helping scientific teams use advanced AI to reason across complex biological evidence," and OpenAI describes the work as the start of a long-term commitment to AI that can shorten the years-long, billion-dollar slog of drug discovery. The open question is whether faster reasoning translates into validated results at the bench — the only proof that counts in biology.