What the pricing page shows
OpenAI's API pricing page now carries a second life-sciences model the company has never announced. The entry, gpt-rosalind-discovery, sits under a section titled "Life sciences models" directly beside gpt-rosalind-research. testingcatalog.com documented the change after the AI Tracker Bot account surfaced it on X, and aiweekly.co carried it as a top story in its October 12 edition. The listing appeared without a launch post or press note.
The rate card is identical for both SKUs: $5 per million input tokens, $0.50 per million cached input tokens, and $25 per million output tokens. The page states that cache-write pricing does not apply. For gpt-rosalind-research, billing is dated: it began on October 5, 2026. Two separately named products sharing one price point in the same category is a deliberate catalog decision, not a rounding artifact.
Access is the other half of the listing. The page ties both models to OpenAI's trusted-access program, restricted to approved internal research, which in practice means no public endpoint is reachable for either SKU today. testingcatalog.com's capture of the page also shows the start of a data residency note. A product can be priced, catalogued, and gated simultaneously; the pricing page is doing all three jobs for a model with no documented behavior.
How Rosalind fits OpenAI's life-sciences push
Rosalind is OpenAI's life-sciences brand, and biggo.com maps it across three delivery paths: models, workspaces, and coding plugins. Named initial partners include Amgen and Moderna. The discovery SKU extends the model track rather than opening a new one, which places it inside an existing commercial program with a billing structure already in place — the October 5, 2026 start date for research — instead of a speculative brand extension.
Positioning matters more than branding here. biggo.com describes OpenAI as operating purely as a tool supplier in life sciences, contrasted with Google DeepMind's research-tool approach and Anthropic's hands-on drug research. The same outlet speculates that gpt-rosalind-discovery will primarily target drug discovery, a reading consistent with the name but unverified by OpenAI. The vertical push is happening while price competition among general-purpose large language models intensifies, giving specialized SKUs a clearer commercial rationale.
The pricing comparison inside OpenAI's own catalog is telling. biggo.com reports that the company's cybersecurity model costs three times the life-sciences model, and reads that gap as a signal that OpenAI currently prioritizes accumulating pharma clients over maximizing price per token. Two life-sciences SKUs at one rate card fit that reading: identical pricing lowers the barrier for a second workload without forcing a procurement decision between them.
What remains unconfirmed
What is missing is documentation. testingcatalog.com reports there is no blog post, no help article, and no system card for gpt-rosalind-discovery. Without a system card, nothing is on record about what the model does, what it was evaluated against, or how it differs from gpt-rosalind-research beyond the name. The pricing entry is currently the only primary source, which is an unusual evidentiary state for a catalogued commercial product.
The identical pricing admits two readings. It could indicate the same underlying capability tier sold under a task-oriented name, or a portfolio pricing decision meant to keep discovery and research interchangeable for trusted-access users. The evidence does not distinguish between them. What the shared rate card does establish is that both SKUs are priced for one commercial frame rather than tiered by use case.
Three signals would move this from listing to launch: a system card, a billing start date for discovery comparable to research's October 5 date, and any change to the trusted-access gate. Until then the record is silent on release timing and availability. orcarouter.ai logged the SKU on October 10, aiweekly.co on October 12; the documentation gap has persisted across that window.
Sources
- testingcatalog.com — OpenAI Quietly Lists a Second GPT-Rosalind Model for Life Sciences
- aiweekly.co — AI News Today, October 11: Top Stories (2026-10-11)
- biggo.com — OpenAI Quietly Lists a New Life Sciences Model, Targeting the Drug-Tool Supplier Track… (2026-10-11)
- orcarouter.ai — GPT-Rosalind Discovery: Unannounced Second Rosalind SKU (2026-10-10)
- New Horizon daily digest, 2026-10-12 (topic provenance)