AI BriefingOpenAI releases GPT-Rosalind-6 for trusted biological research access
·2026.11.02 23:44
Key point
The model shares GPT-6 Astra weights and achieves 49.6% on MedChemBench v7, exceeding the High capability threshold in three of four biological evaluations while remaining below Critical thresholds.
Details
- OpenAI launched GPT-Rosalind-6, a frontier reasoning model for biology and drug discovery, available via trusted access to qualified scientists, institutions, and government partners.
- The model uses the same weights as GPT-6 Astra without additional post-training, achieving 49.6% on MedChemBench v7 and 37.7% on GeneBench Pro v13, significantly outperforming GPT-Rosalind-5.5.
- GPT-Rosalind-6 exceeded the High capability threshold in three of four biological evaluations, including 85.0% on refusal-adjusted Biorisk knowledge, but failed to meet the Critical capability threshold in all four assessments.
- The model inherits GPT-6 Astra's cyber safety measures with a more conservative refusal boundary, recording a 14.3% average task success rate on historical jailbreak regression tests.
- Access requires business verification for non-government entities and is restricted to approved biosecurity and public health purposes for government partners, with mandatory human review before model outputs influence experiments or decisions.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
AWS Weekly Roundup: GPT-6.1 Terra on Bedrock, Step Functions as agent tools, and more (November 2, 2026)
OpenAI Merges ChatGPT and Codex Usage Limits for Plus, Pro, and Business PlansOpenAI Secures $11B at $1.25T Valuation, Missing $30B Target