原文 · 未翻译
OpenAI launches GPT-Rosalind, a reasoning model built for life sciences research
Key Points
OpenAI has introduced GPT-Rosalind, a reasoning model built for life sciences research that assists with tasks like evidence synthesis, hypothesis generation, experiment design, and data analysis.
Named after chemist Rosalind Franklin, the model is said to handle reasoning about molecules, proteins, genes, and disease-relevant biology more accurately than earlier GPT versions.
In internal benchmarks, GPT-Rosalind outperforms GPT-5, GPT-5.2, and GPT-5.4 in chemistry, biochemistry, and experiment design. For now, it is only accessible as a research preview to qualified US enterprise customers through a Trusted Access Program.
OpenAI has introduced GPT-Rosalind, a reasoning model tailored for the life sciences. It's designed to help researchers move faster from hypothesis to experiment. Access is tightly controlled for now.
The model is named after chemist Rosalind Franklin, whose work helped uncover the structure of DNA. The name fits the mission: GPT-Rosalind is built to tackle problems in the biosciences, drug discovery, and translational medicine. It's meant to help researchers synthesize evidence, generate hypotheses, plan experiments, and work through multi-step research tasks.
OpenAI says the model sets itself apart from earlier GPT versions by being tuned specifically for scientific workflows. It's meant to reason more accurately about molecules, proteins, genes, signaling pathways, and disease biology, while making better use of scientific tools and databases across multi-step workflows. Supported tasks include literature research, interpreting sequence-function relationships, experiment planning, and data analysis.
In OpenAI's own evaluations, GPT-Rosalind outperforms its predecessors GPT-5, GPT-5.2, and GPT-5.4 across chemistry, biochemistry and protein understanding, phylogenetics, experiment design, and tool usage.
On the public BixBench benchmark for bioinformatics and data analysis, GPT-Rosalind scored 0.751 on Pass@1. According to OpenAI, that puts it ahead of GPT-5.4 (0.732), Grok 4.2 (0.698), GPT-5 (0.728), and Gemini 3.1 Pro (0.550). On LABBench2, which covers tasks like literature research, database access, sequence manipulation, and protocol design, GPT-Rosalind beats GPT-5.4 on 6 out of 11 tasks, OpenAI says. The biggest jump came in CloningQA, which requires fully designing DNA and enzyme reagents for molecular cloning protocols.
A free plugin that connects more than 50 scientific data sources
Alongside the model, OpenAI is releasing a freely available life sciences research plugin for Codex on GitHub. The plugin provides modular skills for common research workflows and hooks models up to more than 50 public multi-omics databases, literature sources, and biology tools. It spans human genetics, functional genomics, protein structure, biochemistry, clinical evidence, and public study discovery.