OpenAI Introduces GPT-Rosalind for Life Science Research
The new model outperforms GPT-5.4 in chemistry and experimental design, scoring 0.751 on the BixBench bioinformatics test.
OpenAI has introduced GPT-Rosalind, a specialized frontier model named for biologist Rosalind Franklin, designed to accelerate research in biology, drug discovery, and translational medicine. The model is built to handle complex, multi-stage scientific workflows that typically bottleneck early-stage research, such as synthesizing information from vast literature, generating testable hypotheses, and designing experiments. In benchmark tests, GPT-Rosalind outperformed even OpenAI's most advanced general models, including GPT-5.4, showing marked improvements in chemistry, protein engineering, genomics, and experimental design. It achieved a leading Pass@1 score of 0.751 on BixBench, a benchmark simulating real-world bioinformatics and data analysis tasks.
To support the model, OpenAI is providing a free Life Sciences research plugin for Codex, which connects to over 50 public multi-omics databases and biological tools. Due to the potential for biological misuse, access to GPT-Rosalind itself will be tightly controlled via a trusted access model, initially rolling out only to eligible enterprise customers in the U.S. who conduct legitimate, public-interest research and have robust governance systems. OpenAI positions this as the first step in a dedicated life science modeling series, with future plans to enhance biochemical reasoning and support longer, tool-dependent research cycles.
- Outperforms GPT-5.4 in chemistry & experimental design, scoring 0.751 on BixBench bioinformatics benchmark.
- Designed for multi-stage research workflows: literature review, hypothesis generation, and experiment planning.
- Rolling out with strict trusted access in the U.S. to prevent misuse, paired with a free 50+ tool plugin.
Why It Matters
It targets the most time-consuming phase of drug discovery, potentially saving years and billions in R&D costs for new therapies.