GPT-Rosalind is here: a frontier model built for life sciences research!

Hey there, it's Shiichan! Today I've got a shiny new AI model that steps right into life sciences research, and I can't wait to tell you about it!

OpenAI News openai.com

What was announced?

From OpenAI's News comes GPT-Rosalind, a frontier reasoning model built to support research across biology, drug discovery, and translational medicine! It's optimized for scientific workflows, pairing deeper understanding across chemistry, protein engineering, and genomics with improved tool use.

The name honors Rosalind Franklin, whose rigorous research helped reveal the structure of DNA. For now it's available as a research preview in ChatGPT, Codex, and the API for qualified customers through the trusted access program. OpenAI is already working with names like Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific to apply it across discovery workflows.

Why it matters

Getting a new drug from target discovery to regulatory approval takes roughly 10 to 15 years on average in the United States. And research isn't slowed only by the hard science itself, but by how complex the workflows are.

Scientists have to work across huge volumes of literature, specialized databases, experimental data, and ever-evolving hypotheses, and that's time-intensive, fragmented, and hard to scale. GPT-Rosalind aims to accelerate the earliest stages of discovery by helping with multi-step research tasks like evidence synthesis, hypothesis generation, and experimental planning. Gains at those early stages compound downstream, so speeding them up is a big deal.

What changes

GPT-Rosalind is strong at reasoning over molecules, proteins, genes, pathways, and disease-relevant biology, and it's good at using scientific tools and databases in multi-step workflows like literature review, sequence-to-function interpretation, experimental planning, and data analysis.

On top of that, there's a Life Sciences research plugin for Codex arriving at the same time. It lets you connect models to more than 50 scientific tools and data sources. Eligible Enterprise users can pair it with GPT-Rosalind for deeper biological reasoning, while everyone else can use the plugin with the mainline models.

Dive Deep

The evaluations are the fun part, so let's get specific.

• On BixBench (a benchmark around real-world bioinformatics and data analysis), GPT-Rosalind delivered leading performance among models with published scores.
• On LABBench2 (tasks like literature retrieval, database access, sequence manipulation, and protocol design), it outperformed GPT-5.4 on 6 of 11 tasks. The biggest jump was CloningQA, which needs end-to-end design of DNA and enzyme reagents for molecular cloning protocols.
• Partnering with Dyno Therapeutics on an RNA sequence-to-function prediction and generation task, using unpublished, uncontaminated sequences, the model was compared against 57 historical scores from human experts in the AI-bio field. Evaluated in the Codex app, best-of-ten submissions ranked above the 95th percentile of human experts on prediction and around the 84th percentile on sequence generation.

About that plugin: the Life Sciences research plugin is available on GitHub today, packed with modular skills for common research workflows spanning human genetics, functional genomics, protein structure, biochemistry, clinical evidence, and public study discovery. It reaches more than 50 public multi-omics databases, literature sources, and biology tools.

Access works a bit differently: to guard against biological misuse, it ships through a trusted-access structure. It starts with qualified Enterprise customers in the U.S., judged on three principles: beneficial use, strong governance and safety oversight, and controlled access with enterprise-grade security. Organizations that want in can request access through a qualification and safety review process.

A quick note on pricing: during the research preview, using this model won't consume your existing credits or tokens (subject to abuse guardrails). OpenAI says it will share more on pricing and availability as the program expands.

Wrap-up

• GPT-Rosalind is a frontier reasoning model for life sciences research, targeting biology, drug discovery, and translational medicine, and named after Rosalind Franklin.
• It's a research preview in ChatGPT, Codex, and the API via the trusted access program, with partners like Amgen and Moderna.
• It leads on BixBench and beats GPT-5.4 on 6 of 11 LABBench2 tasks, and ranked in high human-expert percentiles in the Dyno Therapeutics evaluation.
• The Life Sciences research plugin for Codex is on GitHub, connects to 50+ tools and data sources, and works with the mainline models too.
• During the preview it won't burn your existing credits or tokens.

If you're a researcher running the search-hypothesize-design-experiments loop in drug discovery or bioinformatics, this is the announcement for you!