OpenAI gives 100,000 researchers free frontier AI access

OpenAI launched ChatGPT for Academic Researchers on July 29, a program that will give 100,000 scientists, mathematicians and engineers free access to its frontier AI tools. The first 10,000 participants will receive access this summer, with the rollout expanding through 2027. Early institutions include the Institute for Advanced Study in the United States and France's École normale supérieure.

Participants will be able to use OpenAI's most capable models across ChatGPT, ChatGPT Work and Codex, including GPT-5.6 Sol Pro at launch. They will also receive larger context windows, expanded deep-research access, higher usage limits and more than 75 life-science skills. Connectors can bring in scientific literature, genomic and clinical databases, satellite imagery, notebooks and reference managers. Approved researchers may invite up to four verified collaborators from the same institution.

OpenAI says the workspaces include business-grade privacy and security protections, and that submitted data will not be used to train its models by default. Training and hands-on support are part of the program as well. Applications opened immediately, but access is limited to qualifying researchers at selected, recognised degree-granting institutions with substantial research activity.

The scale makes this more than a conventional academic pilot. OpenAI says roughly 1.3 million people already use ChatGPT each week for advanced science and mathematics, generating about 8.4 million messages. The company is now lowering the cost barrier for researchers who may lack the budgets of major corporate labs. The initiative forms part of a commitment of more than $250 million through 2027 to support external scientific research and discovery.

For researchers, the immediate opportunity is practical: faster literature reviews, code generation, data analysis, hypothesis testing and grant preparation. For universities, it creates a new reason to set clear rules around verification, reproducibility, confidential data and authorship before AI becomes embedded in research workflows.

There is an important limit. This is free product access, not an open release of model weights, training data or internal methods. Independent researchers can test what the systems do through OpenAI's interfaces, but they still cannot fully inspect how the models were built. Even so, putting frontier tools in the hands of 100,000 academics could broaden who gets to experiment with advanced AI—and give OpenAI a large stream of expert feedback from real scientific work.