OpenAI unveils GPT-Rosalind, an AI model for biology, genomics and drug discovery

On April 16, OpenAI introduced GPT-Rosalind, the company's first specialized model for life sciences, meaning tasks at the intersection of biology, pharma and applied medicine. The model is named after Rosalind Franklin and is designed not to replace scientists but to speed up the hardest parts of scientific work: reading literature, cross-checking data, forming hypotheses and designing next steps. For context, the path from finding a new target to getting a drug approved in the US typically takes 10 to 15 years.

OpenAI is betting that AI can shorten at least the early stages of that cycle, where researchers spend huge amounts of time on analysis, comparing facts and working with scattered sources. According to the company, GPT-Rosalind is optimized for scientific workflows that require reasoning about chemistry, proteins, genes, biological pathways and experimental protocols. Unlike general-purpose chatbots that do a little of everything, the focus here is on long research chains: collecting and comparing dozens of papers, pulling data from specialized databases, interpreting results, proposing a working hypothesis and helping plan the next experiment.

OpenAI stresses that this is not just text generation. The model is meant to help with summarizing the evidence base, generating hypotheses, planning experiments and analyzing data, that is, to work inside real research routines rather than only answer questions in a chat window. OpenAI also released a Life Sciences research plugin for Codex.

The plugin provides access to more than 50 scientific tools, literature sources, multi-omics databases and biological services. In practice, a researcher can run sequence searches, view protein structures, collect publications, find open datasets and tie all of it to the model's reasoning from a single interface. For labs and pharma companies this matters more than yet another LLM: the value appears where the model not only writes an answer but can reach the right data and fit into an existing computational pipeline.

By OpenAI's own public assessments, GPT-Rosalind posts the best result on BixBench among models with published metrics. On LABBench2, the model beat GPT-5.4 on six of eleven tasks; the company saw the biggest gain in CloningQA, where the model must design reagents and a complete plan for molecular cloning.

OpenAI also tested the model with Dyno Therapeutics on predicting and generating RNA sequences by function, using unpublished data that could not have been in the training set. In the Codex environment, the model's best of ten attempts ranked above the 95th percentile of human experts on the prediction task and at roughly the 84th percentile on sequence generation. This does not prove that AI is ready to make discoveries on its own, but it shows that in narrow bioinformatics scenarios it is starting to compete with strong specialists.

The launch is cautious. GPT-Rosalind is currently available as a research preview through ChatGPT, Codex and the API, but only to qualified enterprise customers in the US under a trusted access program. OpenAI cites heightened safety measures, access controls and governance and internal-control requirements: an organization must conduct legitimate research with a clear benefit to human health, restrict access to the model and follow rules on preventing misuse.

The company names Amgen, Moderna, Thermo Fisher Scientific, the Allen Institute and other organizations from the life sciences ecosystem among its partners and early users. In parallel, OpenAI is working with Los Alamos National Laboratory on AI support for designing proteins and catalysts. For the market, this is an important shift.

OpenAI is showing that the next big bet is not only more general and larger models but narrowly specialized systems built for specific professional workflows. In life sciences the cost of error is high, the data is complex, and the payoff is measured not by the number of polished answers but by the quality of hypotheses and the speed of experiments. If GPT-Rosalind really takes root in lab workflows, it will strengthen the trend toward vertical AI models for industries where value comes from deep domain understanding, access to tools and operation in a tightly controlled environment.