OpenAI updates GPT-Rosalind life-sciences model, adds Novo Nordisk and cuts genomics token use 31% vs GPT-5.5

On June 3, 2026, OpenAI announced a major update to GPT-Rosalind, its frontier reasoning model purpose-built for life-sciences research, drug discovery and translational medicine (named for Rosalind Franklin). The refreshed model folds in GPT-5.5's agentic coding and tool-use while sharpening domain performance — and does so more efficiently, using 31% fewer tokens than GPT-5.5 on genomics work. Reported benchmark gains: MedChemBench (medicinal chemistry) 27.5% vs GPT-5.5's 25.1% using 7.2% fewer tokens; GeneBench (genomics) 21.6% vs 20.4%; and LabWorkBench (wet-lab protocols) 63.2% vs 55.8% using 5.3% fewer tokens.

The model is built for scientific workflows — evidence synthesis, hypothesis generation, experimental planning, medicinal chemistry, protein engineering and bioinformatics — with new plugins for evidence retrieval. OpenAI expanded availability: GPT-Rosalind is now in research preview to eligible organizations worldwide through its trusted-access deployment program. Named partners include newest addition Novo Nordisk alongside Amgen, Moderna, the Allen Institute and Thermo Fisher Scientific.

The update lands amid persistent skepticism about AI's translational payoff: despite roughly $17 billion invested in AI-driven drug discovery since 2019, no AI-developed drug has yet completed large-scale clinical trials, underscoring that benchmark wins don't equal therapeutic breakthroughs. Still, the release signals OpenAI's push into vertical, domain-specialized frontier models — pairing general agentic intelligence with measurable efficiency and accuracy gains in a high-value scientific niche and broadening enterprise access globally.