GPT-Rosalind

// Overview

GPT-Rosalind is OpenAI's first domain-specific frontier reasoning model, purpose-built for life sciences research, drug discovery, and genomics. Named after Rosalind Franklin, it was introduced on April 16, 2026 and is available in research preview through OpenAI's trusted-access deployment structure to qualified enterprise customers in ChatGPT, Codex, and the API. Launch partners include Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific.

The model is tuned for multi-step scientific workflows over molecules, proteins, genes, pathways, and disease-relevant biology — covering evidence synthesis, hypothesis generation, experimental planning, genomics analysis, and interpretation of research data. In OpenAI's Codex evaluations, best-of-ten GPT-Rosalind submissions ranked above the 95th percentile of human experts on a prediction task and reached the 84th percentile on a sequence-generation task.

On June 3, 2026, OpenAI shipped expanded capabilities for GPT-Rosalind, combining GPT-5.5's agentic coding and tool-use with stronger drug-discovery domains such as medicinal chemistry and genomics. The updated model uses 31% fewer tokens than GPT-5.5 while improving accuracy on quantitative-biology workloads (21.6% vs 20.4% on GeneBench). Access remains gated to eligible research organizations through the trusted-access program; during the preview, use of GPT-Rosalind does not consume existing credits or tokens.

// Benchmarks

This model's scores

1. GeneBench (genomics & quantitative biology, accuracy)21.6%
2. Codex prediction task (best-of-ten percentile vs human experts)95%ile
3. Codex sequence-generation task (percentile vs human experts)84%ile

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Strengths

• Frontier reasoning tuned for biology, drug discovery, medicinal chemistry, and genomics
• Higher accuracy than GPT-5.5 on genomics/quantitative-biology tasks while using 31% fewer tokens (per OpenAI's June 2026 update)
• Top-tier results on OpenAI's Codex prediction and sequence-generation evaluations vs human experts
• Inherits GPT-5.5's agentic coding and tool-use abilities for scientific tool integration and Codex workflows
• Trusted-access deployment structure designed for legitimate research organizations with strong governance

// Best for

• Reach for it for evidence synthesis and literature review across genomics, transcriptomics, proteomics, and clinical evidence.
• Reach for it for hypothesis generation, experimental planning, and wet-lab troubleshooting in drug-discovery workflows.
• Reach for it for medicinal-chemistry queries — turning molecules into useful drugs — and genomics or quantitative-biology analysis.
• Reach for it as an agentic lab assistant in Codex, connecting models to scientific tools and data sources via the trusted-access API.

// How to access