Introducing GPT-Rosalind for Life Sciences Research

OpenAI has released GPT-Rosalind, a dedicated frontier reasoning model for biology, drug development, and translational medicine. The model focuses on accelerating complex research workflows that mix literature, specialized databases, experimental data, and hypothesis exploration.

Given that drug development takes 10-15 years from target discovery to regulatory approval, improvements at the early discovery stage can significantly shape downstream outcomes. GPT-Rosalind is designed to support multi-step tasks such as evidence synthesis, hypothesis generation, and experiment planning, enabling researchers to explore more possibilities and reach better conclusions earlier.

Key details include the following.

• Available as a Research preview to eligible customers in ChatGPT, Codex, and the API
• A Life Sciences research plugin for Codex provided free of charge, connecting to 50+ scientific tools and data sources
• Collaboration with organizations including Amgen, Moderna, Allen Institute, and Thermo Fisher Scientific to apply it to real-world workflows

As its name suggests, the model honors Rosalind Franklin and aims for deeper understanding and enhanced tool use across chemistry, protein engineering, and genomics as a whole. It has been tuned to deliver stronger performance than existing models particularly on multi-step tasks such as literature review, sequence-to-function interpretation, experimental design, and data analysis.

Evaluations also confirmed improvements. On BixBench, it showed leading performance based on published scores, and on LABBench2, it outperformed GPT-5.4 on 6 out of 11 tasks. In particular, there was a large improvement on CloningQA, and in a separate evaluation with Dyno Therapeutics, best-of-ten submissions in the Codex app exceeded the 95th percentile of human experts on prediction tasks and scored around the 84th percentile on generation tasks.

Access is still operated under a strict trusted access structure. It starts with eligible Enterprise customers in the US, and requires safeguards against biological misuse, organizational governance, and security controls. OpenAI stated that during the research preview period, it will not consume existing credits or tokens within the scope of the abuse guardrails.