OpenAI commits Codex and API access worth up to $17 million to the DOE's Genesis Mission

OpenAI has detailed a new package of commitments to the U.S. Department of Energy's Genesis Mission, offering national-laboratory and university researchers several million dollars in free Codex and API access along with early looks at unreleased models and specialized bioscience and cybersecurity tools.

What's new

The company says it will provide $4 million in Codex access to roughly 2,000 Genesis researchers across the National Laboratories and universities, and commit $3 million in API support to two large-scale scientific campaigns designed to test frontier models against major open scientific problems. Separately, OpenAI is offering participating researchers up to $10 million in API usage for every $2.5 million they spend — a 4x match structured to stretch existing federal science budgets further.

Beyond raw compute access, OpenAI is giving selected national-laboratory researchers working on eligible biology projects access to GPT-Rosalind, its specialized bioscience model, and expanding trusted access to its advanced cyber capabilities for national-laboratory cybersecurity researchers working on defensive research. Trusted national-laboratory leaders will also get early access to selected unreleased models and features, so labs can build out the workflows and evaluations needed before a model ships broadly.

OpenAI frames two "initial large-scale campaigns" it wants to pursue under this support: one chasing breakthroughs in high-temperature superconductors that operate at higher temperatures and more practical pressures, and a second — which it calls "An Atlas of the Machine-Accessible Frontier" — aimed at mapping where AI can already meaningfully accelerate science using existing knowledge and data, versus where problems still require new physical evidence.

Context

This builds on a formal OpenAI-DOE collaboration announced roughly a year ago, which already included deploying advanced reasoning models on Venado, the supercomputer at Los Alamos National Laboratory shared across National Nuclear Security Administration labs, and a joint bioscience-safety evaluation effort with Los Alamos. OpenAI also points to an "AI Jam Session" earlier this year in which more than 1,000 scientists across nine National Laboratories tested frontier models on domain-specific problems.

The announcement lands two days after Google detailed its own $40 million AI-credit commitment to the same Genesis Mission program, and alongside Meta's open-source vision models already being used on a separate Genesis-linked beamline-analysis project at Berkeley Lab. Genesis Mission itself is the DOE's push to integrate AI with federal scientific data, computing, and lab infrastructure, with a stated goal of doubling the productivity of American research within a decade.

Why it matters

Genesis Mission has quickly become a proving ground where every major AI lab is staking out a specific, differentiated role rather than making an identical pledge: Google is leading with raw credits, Meta with open model weights already deployed on real workloads, and now OpenAI with a mix of compute access, matched funding, and staged access to frontier and specialized models like GPT-Rosalind. For OpenAI specifically, tying its highest-value, not-yet-public capabilities to national-lab work gives it a controlled setting to find real scientific use cases — and a favorable venue to demonstrate frontier-model value to policymakers at a moment when AI infrastructure spending and government AI policy are both under heavy scrutiny.