OpenAI updates GPT‑Rosalind for enterprise life sciences and debuts LifeSciBench

OpenAI released a research preview update to its GPT‑Rosalind model series, purpose‑built for enterprise‑scale life sciences research. The update combines GPT‑5.5’s agentic coding and tool‑use capabilities with stronger intelligence in core drug‑discovery domains such as medicinal chemistry and genomics, and is available to eligible organizations globally through a trusted‑access deployment.

To measure real‑world impact, OpenAI developed LifeSciBench, an externally expert‑judged benchmark that takes an end‑to‑end view of scientifically valuable work across six workflow areas: evidence handling, analysis, design and optimization, scientific reasoning, validation and operations, and translation and communication. As an example, OpenAI published an evaluation in which GPT‑Rosalind pressure‑tested a gene‑therapy package for Duchenne muscular dystrophy, identifying gaps in assay specificity, quantification standards, and the use of external natural‑history controls.

OpenAI also reported that GPT‑Rosalind achieves industry‑leading performance in medicinal chemistry on its MedChemBench evaluation, scoring 27.5 percent while using 7.2 percent fewer tokens than GPT‑5.5’s 25.1 percent. The benchmark tests multimodal chemical structure understanding, structure‑activity relationships, ADME prediction, and multiparameter lead‑optimization.

Source: OpenAI Blog (https://openai.com/blog)