OpenAI offers US scientists up to $17 million in Codex and API access

OpenAI has announced support for the US Department of Energy’s Genesis Mission, connecting 17 national laboratories, universities and industry. It is offering roughly 2,000 researchers Codex and API access through programs worth up to $17 million.

Codex and API credits are entering specific research campaigns

The commitment includes $4 million in Codex access and $3 million in API support for two campaigns. OpenAI also offers up to $10 million in additional API use against $2.5 million spent by participants.

Selected researchers will receive access to GPT-Rosalind for biology. Trusted national laboratory leaders are promised early access to some models and features, while defense research will receive broader access to advanced cyber capabilities.

The default tool will shape future research workflows

Credits can accelerate coding, simulation work and experiment design. The longer strategic value for OpenAI is that researchers begin adapting procedures, evals and data interfaces to its models.

A laboratory that embeds Codex in daily work will later evaluate alternatives against an established workflow. Adoption before results therefore matters more to the vendor than a one-off sponsorship announcement.

Credit consumption is not a scientific result

The program can easily count users, prompts and consumed tokens. Those metrics cannot show whether a model helped produce a valid hypothesis or a reproducible discovery.

Sensitive biology and cyber projects also require data audits, access controls and human validation. Comparisons across multiple vendors are essential if public research is to avoid dependence on a single proprietary layer.

Publications and replication will measure the contribution

The two named campaigns concern high-temperature superconductors and the Atlas of the Machine-Accessible Frontier. Useful evidence will be published methods, independently replicated results and open evals. Workshops and exhausted credits can demonstrate usage, but they cannot establish scientific progress.

Lilith's verdict

OpenAI is placing Codex beside the laboratory notebook before the results exist. If scientists begin recording experiments in its syntax, the next vendor contest will take place on a board that is already covered in writing.