Is GPT-6 Rosalind state of the art?Pending benchmarks.Announced Oct 30, 2026; no independent results yet.
OpenAI's life sciences model for biology, drug discovery and translational medicine. Its system card says it uses the same weights as GPT-6 Astra, with no added training data or post-training; what differs is the deployment: trusted access only, a different refusal boundary on dual-use biology, and extra safeguards. Approved Rosalind Research organizations can use it in ChatGPT Work and Codex, and in the API as gpt-6-rosalind where separately approved; it is not in the ChatGPT Chat tab. API price: $15 input and $75 output per 1M tokens. No open weights.
The details so far
- Lab
- OpenAI
- Announced
- Oct 30, 2026
- API
- Waitlist
- Open weights
- Closed · Not released
- API model ID
- gpt-6-rosalind (listed in the OpenAI API changelog on Oct 30, 2026)
- Names
- GPT-6 Rosalind (API changelog); GPT-6 Rosalind Research (ChatGPT, Codex, Help Center); GPT-Rosalind-6 (system card)
- Weights relation
- Same weights as GPT-6 Astra; OpenAI says it added no training data and did no additional post-training for this release
- Access
- Trusted access only, for organizations and users approved for Rosalind Research; Rosalind Discovery approval does not include it. In Enterprise workspaces it starts in a two-week Admin Preview, off by default
- Available
- ChatGPT Work and Codex (with the Life Sciences Research plugin), and the OpenAI API where separately approved; not in the ChatGPT Chat tab. API use is limited to internal research, not customer-facing products
- Price (OpenAI API)
- $15 input, $1.50 cached input, $75 output per 1M tokens; cache writes $18.75 (30 min) and $37.50 (12 h); long-context rates $30 input and $112.50 output
- Preparedness classification (OpenAI)
- High in Biological and Chemical (below Critical), Critical in Cybersecurity, below High in AI Self-Improvement
- Related models
- Separate from GPT-Rosalind-5.5 (gpt-rosalind-research, generally available to approved orgs since Sep 8, 2026 at $5 / $25 per 1M tokens) and gpt-rosalind-discovery
- Size
- Not stated
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: fewer refusals than general-access models on dual-use biological queries, while still refusing requests for biological weaponization.
Lab claim: at the highest reasoning budget tested, 49.6% on MedChemBench v7 (GPT-Rosalind-5.5: 28.9%), 37.7% single-attempt success on GeneBench Pro v13 (16.7%) and 61.3% on LifeSciBench Gold (44.0%).
Lab claim: after adjusting for refusals, it exceeds the indicative High capability thresholds on three of four biological Preparedness evaluations.
Sources
LABOpenAI API changelog: GPT-6 Rosalinddevelopers.openai.com/api/docs/changelog · Oct 30, 2026LABGPT-Rosalind-6 System Card Addendum (OpenAI Deployment Safety Hub)deploymentsafety.openai.com/gpt-rosalind-6 · Oct 30, 2026LABOpenAI API pricing: life sciences modelsdevelopers.openai.com/api/docs/pricingLABGPT-Rosalind for life sciences research (OpenAI Help Center)help.openai.com/en/articles/20001193-gpt-rosalind-for-life-sciences-researchPart of the GPT-6 family
See every GPT-6 model side by side: GPT-6 family.
Tracking ends Jan 29, 2027 unless the API or weights become available.