2 min read AI-generated

GPT-Rosalind Hits Research Preview — and Moves Into Codex

Copy article as Markdown

OpenAI's life-sciences model is now available as a research preview in ChatGPT, Codex, and the API. It comes with a Life Sciences plugin for Codex that connects researchers to 50+ scientific tools and data sources.

Featured image for "GPT-Rosalind Hits Research Preview — and Moves Into Codex"

OpenAI just moved GPT-Rosalind a step closer to real lab work. The model, built for biology and drug discovery, is now available as a research preview in ChatGPT, Codex, and the API — though not for everyone, but through a trusted-access program for qualified organizations.

A model that thinks in molecules

GPT-Rosalind isn’t a general-purpose chatbot. It’s a reasoning model optimized specifically for scientific workflows: chemistry, protein engineering, genomics. In OpenAI’s evaluations it delivers the best results on tasks that require reasoning over molecules, proteins, genes, pathways, and disease-relevant biology. But the key isn’t raw knowledge — it’s tool handling. The model is meant to use scientific tools and databases sensibly across multi-step flows: literature review, sequence-to-function interpretation, experimental planning, data analysis.

The idea behind it: accelerate research, especially the early stages of discovery. Generate hypotheses, synthesize evidence, plan experiments — the tedious groundwork that precedes every exciting result.

Codex gets a scientific plugin

Interesting for anyone who codes: OpenAI also released a freely accessible Life Sciences plugin for Codex. It connects models to over 50 scientific tools and data sources. That shifts GPT-Rosalind from a pure chat experience toward something you can wire into real research pipelines — where code, data, and models already come together.

Access stays deliberately controlled. OpenAI describes a trusted-access structure for organizations doing legitimate research with clear public benefit, strong governance, and safety oversight. For a model that bundles knowledge about proteins and disease biology, that caution makes sense — the dual-use debate is never far away here.

My take

What interests me about this isn’t the model itself so much as the pattern: the big providers are increasingly building domain-specialized models instead of just inflating the one giant general-purpose model further. Biology, law, finance — specialized tools with their own access regimes are popping up everywhere. For science that could be a real boost. At the same time, a piece of research infrastructure is moving into the hands of a few companies. Both true at once — which is exactly why it’s worth watching.

Sources: OpenAI: Introducing GPT-Rosalind, OpenAI: New capabilities to GPT-Rosalind