GPT-Rosalind-5.5: benchmark results
OpenAI GPT-Rosalind model incrementally trained from GPT-5.5 for biology, drug discovery, and translational medicine research. Provider: OpenAI. Released 2026-06-03. Access: Private.
Aggregate Rating 1833 ± 44, rank #46 of 10552 rated models, from 11 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenAI GPT-Rosalind-5.5 System Card - Hard Negative Protein Binding Prediction | 3.13 | pass@4 (%) | 100 |
| OpenAI GPT-Rosalind-5.5 System Card - TroubleshootingBench | 53.31 | pass@1 (%) | 100 |
| OpenAI LifeSciBench | 36.1 | Exact Pass Rate (%) | 100 |
| SecureBio BioTIER Permit | 100 | Benign prompts answered (%) | 100 |
| SecureBio HPCT | 64.8 | Accuracy (%) | 86.6 |
| SecureBio VCT v2 | 50.29 | Accuracy (%) | 84.6 |
| SecureBio BioTIER Refuse | 61.4 | Hazardous prompts refused (%) | 51.3 |
| OpenAI GPT-Rosalind-5.5 System Card - Biorisk Knowledge | 81.1 | cons@32 (%) | 50 |
| OpenAI GPT-Rosalind-5.5 System Card - DNA Sequence Design for TF Binding | 13.64 | pass@1 (%) | 0 |
| OpenAI GPT-Rosalind-5.5 System Card - Multi-select Virology Troubleshooting | 51.2 | pass@1 (%) | 0 |
| OpenAI GPT-Rosalind-5.5 System Card - ProtocolQA Open-Ended | 36.4 | pass@1 (%) | 0 |
Interactive version: theaggregate.ai/model?slug=gpt-rosalind-5-5 · How It Works · Data refreshed daily, snapshot 2026-11-16.