- OpenAI launched GPT-Rosalind for life sciences and drug discovery.
- It supports research tasks with strong benchmark results.
OpenAI has introduced a new artificial intelligence model, GPT-Rosalind, aimed at supporting research in biology, drug discovery, and translational medicine. The model is named after Rosalind Franklin and is designed to support scientific workflows in areas such as biochemistry and genomics.
The model is aimed at life sciences research. Drug development can take 10 to 15 years, and early-stage work often involves coordination across multiple systems and datasets. The average cost of developing a single drug is estimated at around $2.6 billion. The market for artificial intelligence in drug discovery is projected to grow from about $3.25 billion in 2026 to more than $10 billion by 2031.
GPT-Rosalind is built to handle multi-step research tasks, including evidence synthesis, hypothesis generation, and experimental planning. “By supporting evidence synthesis, hypothesis generation, experimental planning, and other multi-step research tasks, this model is designed to help researchers accelerate the early stages of discovery,” OpenAI said in a blog post.
Model capabilities and performance
The model is based on OpenAI’s latest internal systems and is optimised for scientific reasoning across molecular biology and chemistry. OpenAI said it can reason across biological domains, including proteins and genes, as well as disease pathways. On BixBench, a benchmark for bioinformatics and data analysis, it recorded leading results among models with published scores.
In LABBench2 testing, GPT-Rosalind outperformed GPT-5.4 in six of eleven tasks. The benchmark covers work such as literature retrieval, database access, sequence manipulation, and protocol design, which are common parts of early-stage research. The largest improvement was observed in CloningQA, a task involving reagent design for molecular cloning workflows.
OpenAI also conducted evaluations with Dyno Therapeutics using unpublished RNA sequence data. In these tests, the model’s outputs ranked above the 95th percentile of human experts for prediction tasks and around the 84th percentile for sequence generation when assessed within the Codex environment. The evaluation focused on tasks such as sequence-to-function prediction and generation using RNA data.
Access, tools, and deployment
The model is integrated with a new Life Sciences research plugin for Codex, available via GitHub. OpenAI said the plugin acts as an orchestration layer across scientific tools and datasets. It connects to more than 50 public databases and literature sources for tasks such as protein structure analysis and sequence searches. Researchers can also use it for clinical research workflows without switching between separate tools and databases. OpenAI said the plugin and related connectors are available more broadly, allowing use of its mainline models for life sciences tasks without access to GPT-Rosalind.
GPT-Rosalind is being released as a research preview through OpenAI’s trusted access framework, with availability limited to qualified enterprise users in controlled environments. Access requires a qualification and safety review process, and organisations must demonstrate legitimate research use while maintaining governance controls to prevent misuse, including risks related to biological applications. The model can generate and analyse biological data within research workflows. OpenAI also said use of the model during the preview will not consume standard credits or tokens, subject to safeguards.
Usage is restricted to approved users within secure environments, and participating organisations must comply with defined research and access policies.
Industry adoption
OpenAI said it is working with companies including Amgen, Moderna, and Thermo Fisher Scientific to apply the model across research workflows.
“The life sciences field demands precision at every step. The questions are highly complex, the data are highly unique, and the stakes are high,” said Sean Bruich, senior vice president of artificial intelligence and data at Amgen. He said the collaboration enables the company to apply OpenAI’s tools to research and drug development processes.
Collaborations also include institutions such as the Allen Institute. OpenAI said it is continuing to develop the model’s capabilities for tasks such as protein and catalyst design in collaboration with research institutions.
OpenAI also recently introduced GPT-5.4-Cyber, a variant of its flagship model designed for cybersecurity applications. The release follows a separate announcement from Anthropic, which introduced its frontier model, Mythos.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
