OpenAI launches GPT-Rosalind to take on Google DeepMind in drug discovery
OpenAI unveiled GPT-Rosalind on Thursday, a specialized reasoning model built specifically for life sciences research. The launch marks OpenAI’s first serious entry into scientific computing and puts it in direct competition with Google DeepMind’s AlphaFold.
The model is named after Rosalind Franklin, the British chemist whose X-ray diffraction data was instrumental in discovering DNA’s double helix structure. That naming choice signals OpenAI’s ambitions: this isn’t just another chatbot wrapper, it’s a purpose-built tool for molecular biology, drug discovery, and translational medicine.
GPT-Rosalind targets a specific pain point in pharmaceutical research. Getting a new drug from initial target discovery to FDA approval typically takes 10 to 15 years. The early stages involve grinding through massive literature databases, experimental data, and evolving hypotheses. Scientists juggle specialized tools across fragmented workflows. OpenAI claims its model can accelerate this discovery phase by helping researchers synthesize evidence, generate hypotheses, plan experiments, and spot connections they might otherwise miss.
The model outperformed GPT-5.4 on six biology-focused benchmark tasks according to OpenAI’s internal evaluations. It’s designed to reason over molecules, proteins, genes, and disease pathways while integrating with scientific databases and tools in multi-step workflows.
OpenAI is offering GPT-Rosalind as a research preview through a trusted access program for qualified U.S. enterprise customers. It’s available via ChatGPT, Codex, and the API. The company also released a free Life Sciences research plugin for Codex that connects models to over 50 scientific tools and data sources.
The partner list reads like a who’s who of biotech and pharma: Amgen, Moderna, the Allen Institute, Thermo Fisher Scientific, Novo Nordisk, Oracle Health, NVIDIA, Benchling, and UCSF School of Pharmacy are all signed on. Amgen’s SVP of AI and Data, Sean Bruich, stated the collaboration could “accelerate how we deliver medicines to patients.”
This is a significant strategic pivot for OpenAI. While competitors like Anthropic focus on AI safety and Meta dumps open-weights models into the wild, OpenAI is betting that domain-specific models with enterprise guardrails will capture value in regulated industries where accuracy and auditability matter more than raw reasoning scores.
The drug discovery angle is also smart economics. Pharmaceutical companies spend billions on R&D. A model that genuinely accelerates early discovery could command premium pricing far beyond standard API rates. OpenAI is essentially selling picks and shovels during a gold rush, except the gold is cancer therapies and the miners are billion-dollar pharma companies.
The first release is just the beginning. OpenAI says it will continue expanding the model’s biochemical reasoning capabilities across “long-horizon, tool-heavy scientific workflows.” The compute infrastructure required to train and evaluate these domain-specific models explains why OpenAI has been raising billions for data center expansion.
Whether GPT-Rosalind actually delivers on its promises remains to be seen. Biology has a way of humbling AI researchers. But the launch demonstrates OpenAI’s ambition to move beyond general-purpose chatbots and become infrastructure for specific high-value industries.