What happened: OpenAI introduced GPT‑Rosalind, a life sciences model series aimed at biology, drug discovery, and translational medicine, and said it is available as a research preview in ChatGPT, Codex, and the API for qualified customers via its trusted access program.

Why it matters: Because this is “AI for science” turning into a product shape: tool-heavy workflows, domain evaluations, and access controls. Also, nothing says “high stakes” like putting powerful bio reasoning behind a gate instead of a tweet thread.

Wider context: OpenAI is positioning domain models as workflow accelerators, not just chat. It also highlights an open Life Sciences research plugin for Codex with modular skills that connect to 50+ scientific tools and data sources, which is basically an ecosystem land-grab in lab coats.

Background: OpenAI cites drug development timelines (about 10 to 15 years from target discovery to approval in the US) and argues earlier-stage gains compound. It says the model was evaluated on benchmarks like BixBench and LABBench2, and in a partner evaluation with Dyno Therapeutics.

Introducing GPT‑Rosalind for life sciences research — OpenAI

Singularity Soup Take: This is the part where “general intelligence” quietly becomes “enterprise-grade scientific plumbing.” The punchline is the same as always: the model is impressive, but the real control plane is trusted access, tool connectors, and who gets to run the button-mashing in biology.

Key Takeaways:

• Trusted Access Gate: OpenAI says GPT‑Rosalind is offered as a research preview for qualified customers through its trusted access program, explicitly framing availability as something you earn via controls rather than something you just download.
• Benchmarks And Workflow Tests: The post describes evaluations across public benchmarks (including BixBench and LABBench2) plus a partner test with Dyno Therapeutics using unpublished sequences, aiming to show it can handle tool-heavy, multi-step scientific tasks.
• Codex Plugin Ecosystem: OpenAI also released a Life Sciences research plugin for Codex on GitHub, with modular “skills” that provide access to 50+ public databases, literature sources, and biology tools, positioning Codex as a hub for scientific orchestration.