OpenAI is rolling out significant upgrades to its GPT-Rosalind AI model, specifically targeting the complex demands of life sciences research and drug discovery. Built on the foundation of OpenAI GPT-5.5, the updated model boasts enhanced agent coding, tool usage capabilities, and deeper intelligence in critical areas such as medicinal chemistry and genomics.
Visual TL;DR. Drug Discovery Needs supports GPT-Rosalind upgrade. GPT-Rosalind upgrade enables enhanced inference. GPT-Rosalind upgrade enables workflow execution. Enhanced inference powers AI discovery. Executing workflows leads to enhanced AI detection. GPT-Rosalind upgrade evaluated by LifeSciBench benchmark. GPT-Rosalind upgrade facilitates data synthesis.
Drug discovery needs: The complex demands of life science research and drug discovery
GPT-Rosalind Upgrade: OpenAI’s GPT-Rosalind has been significantly upgraded for life sciences
Workflow execution: Bridging the gap between AI-driven inference and actual execution
LifeSciBench Benchmark: Comprehensive benchmark evaluated by external experts for real-world impact
Data synthesis: Synthesize data at various scales, from molecules to biological systems.
Powering AI Discovery: Powering drug discovery and genomics with improved capabilities
Visual TL;DR
This advancement aims to bridge the gap between AI-driven inference and the actual execution of scientific workflows. The GPT-Rosalind improvements are designed to synthesize data across a variety of scales, from molecules to biological systems, a key aspect of progress in this field.
Tackling scientifically valuable challenges
To assess real-world impact, OpenAI developed LifeSciBench, a comprehensive benchmark evaluated by external experts. This assessment covers six core life sciences workflow areas: evidence processing, analysis, design and optimization, scientific reasoning, validation, and communication.
In evaluation using LifeSciBench, the updated GPT-Rosalind demonstrated broad performance improvements across tasks identified by both academic and industry experts. This includes complex medicinal chemistry queries and complex biological analyses.
stronger scientific reasoning
GPT-Rosalind achieved industry-leading performance in medicinal chemistry as measured by MedChemBench. It reportedly performed 27.5% to 25.1% better than GPT-5.5 while using fewer tokens, and showed improvements in multimodal synthesis and mechanical inference.
For genomics and quantitative biology evaluated via GeneBench, GPT-Rosalind showed higher accuracy of 21.6% compared to 20.4% for GPT-5.5 and used 31% fewer tokens. This long-term, end-to-end analysis capability extends to functional genomics, spatial transcriptomics, and proteomics.
Supports real-world lab work
A new evaluation, LabWorkBench, specifically tests GPT-Rosalind’s ability to assist scientists with real-world experimental protocols, from troubleshooting to optimization. On this unique dataset, the model scores 63.2% compared to 55.8% for GPT-5.5, indicating a significant increase in real-world lab assistance and token efficiency.
From inference to executed workflow
OpenAI also introduced two plugins, Life Sciences Research and Life Sciences NGS Analysis, to provide a practical execution layer for repeatable scientific workflows. These plugins integrate evidence retrieval, biological interpretation, and bioinformatics execution, allowing researchers to connect external data to internal analysis.
Eligible enterprise users will now be able to utilize GPT-Rosalind to power these plugins, increasing the utility of Codex as a dynamic workbench. New interactive viewers for native biological file types such as sequences and alignments are also integrated, allowing scientists to work more closely with their data during AI-driven inference.
The company is providing expanded access to trusted organizations around the world through its Research Preview deployment structure, demonstrating a push toward deeper AI integration in enterprise-level life sciences research.
