OpenAI builds life sciences model series for drug discovery

On this episode of the OpenAI Podcast, host Andrew Mayne speaks with Joy Jiao (research lead) and Yunyun Wang (product lead) about OpenAI's work in life sciences. The centerpiece of the discussion is the newly announced life sciences model series — a biochemistry-focused model series built around complex, long-trajectory research workflows. The series prioritizes mechanistic understanding starting with genomics and protein understanding, targeting early discovery use cases where greater compute and more capable AI models can help overcome core research bottlenecks. The team's stated ambition is captured in their internal tagline: "scale test time compute to cure all disease."

These are designed for enterprise reproducibility and are aimed at translational biology users, with aspirations to expand into clinical use cases while remaining broadly useful for foundational biology.

A key product component is the life sciences research plugin, which ships with over 50 skills — templatized, repeatable workflows covering tasks like cross-evidence literature search and pathway analysis. These are designed for enterprise reproducibility and are aimed at translational biology users, with aspirations to expand into clinical use cases while remaining broadly useful for foundational biology. The models are being deployed across multiple surfaces: ChatGPT for literature synthesis workflows and Codex for long-trajectory agentic workflows. Joy Jiao describes current model capabilities as analogous to a computational biologist — running open-source protein structure prediction tools, reviewing outputs, and iterating on inputs — with the next step being deeper integration into autonomous lab environments, including a collaboration with Ginkgo Bioworks. The episode also addresses biorisk and biosecurity as explicit design constraints, framing responsible deployment as inseparable from advancing capability in a field with real-world safety stakes.

Key facts

1. 01OpenAI announced a new biochemistry-focused life sciences model series targeting early drug discovery workflows.
2. 02The series focuses on mechanistic understanding starting with genomics and protein understanding.
3. 03A life sciences research plugin ships with over 50 templatized, repeatable workflow skills.
4. 04Skills cover tasks such as cross-evidence literature search and pathway analysis with a one-click deploy option.
5. 05Models are deployed across ChatGPT (literature synthesis) and Codex (long-trajectory agentic workflows).
6. 06