Introducing new capabilities to GPT-Rosalind

Our GPT-Rosalind series, purpose-built for enterprise-scale life sciences research, has received a new model update. This update combines the agentic coding and tool-use capabilities of GPT-5.5 with stronger model intelligence in core drug-discovery domains like medicinal chemistry and genomics, enhancing performance across various life sciences analysis, design, and experimental workflows. GPT-Rosalind is currently available in research preview to eligible organizations globally through our trusted-access deployment structure.

Advancements in life sciences rely on synthesizing data and evidence across diverse scales and modalities, encompassing molecules, genes, pathways, and living systems. Our evaluations demonstrate that the updated GPT-Rosalind delivers broad performance gains in research tasks performed by biology experts, complex medicinal chemistry queries, quantitative biology, and wet lab troubleshooting.

To ensure continuous improvement and measure the real-world impact of GPT-Rosalind, we developed LifeSciBench. This externally expert-judged benchmark focuses on foundational aspects of life sciences research. Unlike existing benchmarks that assess isolated components of model performance or biological domains, LifeSciBench offers an end-to-end perspective on scientifically valuable work. It draws tasks from six critical workflow areas in life sciences research: evidence handling, analysis, design and optimization, scientific reasoning, validation and operations, and translation and communication. This benchmark helps align our progress with the practical needs and realities of life sciences research.

GPT-Rosalind consistently leads performance across scientifically-valuable tasks, as identified by both industry and academic experts. One example involves extracting, reconciling, and auditing scientific evidence from various sources like papers, figures, tables, and experimental records.