OpenAI hands 100,000 researchers free frontier-model access

OpenAI is opening its frontier models to 100,000 academic researchers free through 2027, under a program called ChatGPT for Academic Researchers. It starts small — roughly 10,000 researchers this summer, seeded at places like the Institute for Advanced Study and École normale supérieure — and scales from there. Each participant can invite up to four collaborators from their own institution.

What you actually get: GPT-5.6 Sol Pro and the rest of the frontier lineup, in a workspace with business-grade privacy where data isn't used for training by default. That's roughly a year of the $200/month ChatGPT Pro tier per seat — call it ~$2,400 of list value each, and the program sits inside OpenAI's stated $250M+ commitment to external science through 2027.

The catch is who's excluded. Eligibility skews to biology, chemistry, physics, math, CS and engineering — people doing science with AI, not people studying AI. Interpretability and safety researchers have been asking for model weights and training data, and a Pro subscription doesn't touch that request. Free API-adjacent credits are a distribution play as much as a philanthropic one: get the next generation of published work running on OpenAI's stack, and citations become lock-in.

Why it matters: the cost floor for frontier-model research just hit zero for 100k people — and OpenAI gets to pick which 100k.