OpenAI quietly lists gpt-rosalind-discovery on its API pricing page — the Rosalind life-sciences line is becoming a family
OpenAI has listed an unannounced gpt-rosalind-discovery model on its API pricing page under ‘Life sciences models,’ at the same $5/$25 rate as gpt-rosalind-research — the clearest sign yet that the Rosalind biology line is becoming a family of models. The listing was spotted by the AI Tracker Bot on X and reported by TestingCatalog on October 10, 2026.
OpenAI has listed an unannounced gpt-rosalind-discovery model on its API pricing page under ‘Life sciences models,’ at the same $5/$25 rate as gpt-rosalind-research — the clearest sign yet that the Rosalind biology line is becoming a family of models. The listing was spotted by the AI Tracker Bot on X and reported by TestingCatalog on October 10, 2026.
OpenAI has listed an unannounced gpt-rosalind-discovery model on its API pricing page under ‘Life sciences models,’ at the same $5/$25 rate as gpt-rosalind-research — the clearest sign yet that the Rosalind biology line is becoming a family of models. The listing was spotted by the AI Tracker Bot on X and reported by TestingCatalog on October 10, 2026.
OpenAI's API pricing page has a new row. Under the section headed "Life sciences models," a model called gpt-rosalind-discovery now sits alongside gpt-rosalind-research — and OpenAI has said nothing about it. No blog post, no help article, no system card. The listing was first flagged by the AI Tracker Bot on X and reported by TestingCatalog on October 10, 2026.
The price column is the first clue. Both Rosalind rows carry identical rates: $5 per million input tokens, $0.50 per million cached input tokens, and $25 per million output tokens, with no cache-write pricing. The page's fine print states that billing for gpt-rosalind-research began on October 5, 2026, that access is limited to approved internal research through OpenAI's trusted-access program, and that data-residency endpoints add a 10% uplift for eligible models released on or after March 5, 2026. TokenCost's pricing breakdown of the same page confirms the row-and-footnote structure: a "Life Sciences" row for gpt-rosalind-research at $5.00/$0.50/$25.00, standard tier, short context, no batch or flex discount.
A family, not a one-off
The telling detail is in OpenAI's own footnote: "all eligible organizations will continue to get access to the latest GPT-Rosalind models as they're released." That sentence describes a product line, not a single model — and a "discovery" variant is the most natural second member. The name is suggestive too: OpenAI introduced GPT-Rosalind in April 2026 as a frontier reasoning model for biology, drug discovery and translational medicine, named after Rosalind Franklin, with Amgen, Moderna, the Allen Institute and Thermo Fisher Scientific among its first partners. Its help center describes the current model as most useful for early discovery and hypothesis generation — target biology, literature synthesis, omics interpretation — so a variant tuned specifically for the discovery end of that spectrum fits the shape of the program.
The Rosalind line has been moving fast. OpenAI took it out of research preview on September 11, 2026 and opened it to eligible organizations worldwide through the trusted-access program; published pricing took effect October 5; and the latest version, GPT-Rosalind-5.5, is trained on top of GPT-5.5. Background coverage from Orcarouter's Rosalind briefing traces that arc from the April trusted-access research preview to today's globally available, gated commercial offering — complete with a Life Sciences Research Plugin for Codex that connects the model to more than 50 scientific tools and databases.
Why it matters
The lab bench is quietly becoming an AI product category. Anthropic has been moving in the same direction: Claude Operon, a biology-focused mode in the Claude desktop app, and Claude Science, currently in beta — alongside OpenAI's own ChatGPT for Academic Researchers. Together with Rosalind, the labs are assembling a full research stack, from literature review to experimental planning to data analysis. A dedicated discovery model would push that stack deeper into the earliest, highest-value stage of pharma R&D: deciding what to test in the first place.
There is also a pricing signal worth reading carefully. At $5 per million input tokens and $25 per million output, the Rosalind rate card lands at exactly the same level as Anthropic's Claude Opus 5 — a clean specialist-versus-generalist matchup at identical token prices. Orcarouter's comparison makes the bet explicit: OpenAI is wagering that life-sciences teams will pay flagship rates for a model tuned on molecules, proteins, genes and pathways rather than breadth.
That bet is inseparable from how OpenAI distributes the model. Rosalind is not in the ChatGPT picker and cannot be self-served: it reaches organizations through the trusted-access program via the API, Codex and ChatGPT Enterprise, with governance and misuse controls as the gate. OpenAI has used the same playbook for other dual-use specialist models, including GPT-5.4-Cyber — restricted by default, expanded only to vetted organizations. If discovery follows the research model's path, expect the same gating, the same three surfaces, and the same fine print about eligible organizations receiving new Rosalind releases as they ship.
The caveats, stated plainly: nothing is confirmed. This could be a separately trained discovery model or a differently tuned variant of the research model; its exact purpose is unclear; and the trusted-access gate will almost certainly stay, given the stakes. Rosalind-5.5's system card says the model is trained not to refuse sophisticated biology questions, with gated access as the main safeguard instead — the same posture OpenAI took when it expanded trusted access for GPT-5.4-Cyber. The pattern to watch: OpenAI routinely lists models on its pricing page shortly before announcing them, so a formal announcement — with a system card, help-center docs, and possibly its own trusted-access terms — may not be far off.