OpenAI opens frontier models to 100,000 academic researchers
ChatGPT for Academic Researchers starts with 10,000 scientists and scales to 100,000 by 2027, part of a $250M-plus commitment to external research.
OpenAI launched ChatGPT for Academic Researchers on Wednesday, a programme that gives scientists at selected institutions free access to its frontier models, starting with 10,000 researchers this summer and expanding to 100,000 through 2027.
Participants get access to OpenAI's top-end models, including GPT-5.6 Sol Pro at launch, in a workspace roughly equivalent to a year of the $200-per-month ChatGPT Pro tier, and each researcher can invite up to four collaborators from their own institution. The workspaces carry business-grade privacy and security terms, and OpenAI says submitted data is not used for model training by default. Eligibility spans biology, chemistry, computer science, engineering, mathematics and physics; access is already live at institutions including the Institute for Advanced Study and École normale supérieure. OpenAI framed the programme as part of a commitment of more than $250 million through 2027 to support external scientific research, and argued that the benefits of frontier AI should not concentrate in a handful of well-resourced labs.
What the programme does not include is model weights. Researchers get inference access under commercial terms — useful for doing science with the models, largely unusable for doing science on them. Interpretability, robustness and evaluation work that requires internals, fine-tuning or activation access remains outside the offer, which is why a substantial part of academic AI research still runs on open-weight models from Chinese and European labs.
The timing is not incidental. Anthropic, Google and OpenAI have all been expanding academic and national-science programmes, and Washington is simultaneously weighing restrictions on Chinese open-weight models that many university labs currently depend on. Free credits at scale are also a recruiting and standard-setting instrument: a generation of graduate students trained inside one vendor's workspace becomes that vendor's default in industry.
Why it matters: Compute access has become the main stratifier in academic AI and adjacent sciences, and 100,000 seats at frontier-tier capability is a meaningful redistribution of it. It also further entrenches the split the open-weights fight is about — labs can now be generous with access while remaining closed about the artefacts, and the science that needs the artefacts still has to look elsewhere.