OpenAI Updates GPT-Rosalind with Enhanced Life Sciences Capabilities

OpenAI released an updated version of its GPT-Rosalind model series on June 3, 2026, designed specifically for life sciences research at enterprise scale. The update integrates GPT-5.5’s agentic coding and tool-use capabilities with improved performance in medicinal chemistry, genomics, and broader workflows including biology analysis, experimental design, and validation.

The model now leads on LifeSciBench, an expert-judged benchmark covering six workflow areas: evidence handling, analysis, design and optimization, scientific reasoning, validation and operations, and translation and communication. It outperforms GPT-5.5, Grok 4.3, and Gemini 3.1 Pro across these tasks, with particular gains in extracting and auditing evidence from papers, figures, and records.

Additional benchmarks show targeted improvements. On MedChemBench, GPT-Rosalind scores 27.5 percent compared to GPT-5.5’s 25.1 percent while using 7.2 percent fewer tokens. On GeneBench for long-horizon genomics and quantitative biology tasks, it reaches 21.6 percent accuracy versus 20.4 percent with 31 percent fewer tokens. LabWorkBench, which evaluates assistance with real wet lab protocols, records 63.2 percent for the updated model against 55.8 percent previously, alongside a 5.3 percent token reduction.

GPT-Rosalind remains available only through research preview to eligible organizations via OpenAI’s trusted-access program. The announcement includes a detailed example of the model critiquing a gene therapy data package for an FDA Type B meeting on AAV9-microDys-X for Duchenne muscular dystrophy, identifying gaps in assay specificity, surrogate endpoint validity, biopsy design, statistical controls, durability, and safety monitoring.

The update follows the initial GPT-Rosalind launch in April 2026. It aims to align model progress more closely with practical needs in drug discovery and life sciences by synthesizing data across molecules, genes, pathways, and living systems.