OpenAI's first industry-specific model, GPT-Rosalind, targets the pharmaceutical sector, achieving over 95% human expert-level accuracy in RNA folding prediction.

According to VentureBeat, OpenAI has released GPT-Rosalind, the company's first industry-specific advanced reasoning model tailored for biology, drug discovery, and translational medicine. Named after scientist Rosalind Franklin, who made crucial contributions to revealing DNA structure, the model is currently available as a research preview on ChatGPT, Codex, and the API, open only to qualified US enterprise customers during the preview period without consuming existing credits.

Choosing the life sciences as the first vertical industry is itself a strategic signal. OpenAI's rationale is that a new drug typically takes 10 to 15 years from target discovery to approval, and AI acceleration in the early discovery phase will compound downstream, leading to better target selection, stronger biological hypotheses, higher-quality experiment designs, and ultimately increasing the success rate of the entire pipeline.

In terms of performance, GPT-Rosalind has achieved the highest score among publicly released models on the BixBench bioinformatics benchmark. On the LABBench2 benchmark, which covers 11 research tasks such as literature retrieval, sequence manipulation, and experimental protocol design, 6 tasks surpass GPT-5.4, with the most significant improvement seen in molecular cloning experiment design (CloningQA). Evaluation conducted in collaboration with AI gene therapy company Dyno Therapeutics used privately held, uncontaminated RNA sequences; the model's best result in ten submissions ranks in the top 95th percentile of historical human expert scores for sequence function prediction and around the 84th percentile for sequence generation.

OpenAI has also open-sourced the Codex Life Sciences Research Plugin on GitHub, integrating with over 50 public multi-omics databases, literature sources, and bioinformatics tools, covering areas such as human genetics, functional genomics, protein structures, and biochemistry. The plugin is free for all users, not limited to GPT-Rosalind, and can be used with general models as well. Early collaborators include Amgen, Moderna, the Allen Institute, Thermo Fisher Scientific, and OpenAI is exploring AI-guided protein and catalyst design with the Los Alamos National Laboratory. GPT-Rosalind is the first version in its life science model series, with plans to continuously enhance bioinformatics reasoning capabilities.