OpenAI quietly ships GPT-6 Rosalind for approved life-sciences labs at $15/$75 per million tokens

OpenAI has released a GPT-6 version of its life-sciences model without an announcement. The only public trace is a one-line entry dated October 30 in the API changelog: "Released GPT-6 Rosalind, bringing specialized biological reasoning with scientific tool use to complex life sciences research." The model ID is gpt-6-rosalind. OpenAI's main accounts on X had not posted about it as of early October 31 UTC.

The details are in an updated help center article.

Who gets it

GPT-Rosalind is gated behind OpenAI's trusted-access program and comes in two tiers. Rosalind Discovery covers everyday work such as literature review, data analysis and protocol interpretation. It runs on the existing Rosalind 5.5 model with stricter monitoring. Rosalind Research is for "frontier biological work," and the new model sits only in that tier, as GPT-6 Rosalind Research. Discovery approval does not carry over; moving up to Research needs a separate review.

GPT-6 Rosalind Research is available to "a limited set of approved Research organizations" in ChatGPT's Work tab and in Codex, but not in the Chat tab. API access needs its own approval. In the API it is restricted to internal research tools, and customer-facing products and external commercial applications are excluded. In Enterprise workspaces it starts in a two-week Admin Preview, switched off by default, and existing Research approval does not skip that step. Academic labs on the Rosalind Academic Plan get no API access.

What it costs

The pricing page lists gpt-6-rosalind at standard rates of:

Prices are per million tokens. That is three times the $5/$25 still charged for gpt-rosalind-research and gpt-rosalind-discovery, and 1.5 times the $10/$50 standard rate for GPT-6 Astra, OpenAI's general flagship. Regional processing adds the usual 10% uplift.

OpenAI has not published benchmarks, a model card or a system card for the GPT-6 version, and the help article describes the family only in general terms: target research, genomics, protein and sequence analysis, medicinal chemistry, wet-lab troubleshooting and experiment planning. It says the models are "currently most valuable for early discovery and hypothesis generation."

The release lands three days after Anthropic's progress report on drug-development work with Claude. Both labs are pushing into life sciences, and both keep their strongest biology capabilities behind application-based access.