References discovered in ChatGPT's web build confirm OpenAI is testing a new subscription tier called "ChatGPT for Science," designed for research workflows rather than general conversation. The leak, first flagged on X by TestingCatalog and picked up by multiple outlets, shows the company expanding its specialized AI portfolio beyond the standard personal, Teams, and Enterprise tiers.
OpenAI has not announced the plan or commented on the references. But the web build code is active, suggesting internal testing is underway.
Access will almost certainly be restricted. OpenAI's existing tier structure offers a roadmap: Teams requires a company domain and three users; Enterprise requires legal entity verification. Early indications point to ChatGPT for Science limiting access to verified universities, research institutes, and labs, not individual researchers.
The subscription appears to build on OpenAI's work with GPT-Rosalind, a model announced earlier this year that runs on GPT-5.5 architecture. Rosalind is no science-skinned chatbot -- it's a purpose-built system for enterprise-scale life sciences research, locked behind a "trusted-access deployment structure" that restricts use to pharmaceutical companies (Novo Nordisk is a known partner), biotech firms, and verified research institutions.
ChatGPT for Science could extend those capabilities to a wider academic audience. Rather than keeping advanced research tools limited to a handful of enterprise partners, the new tier would offer stronger grounding in scientific literature and specialized research topics compared to standard ChatGPT subscriptions.
No launch date has been set. OpenAI is still actively testing the feature on the web build, and an announcement is likely weeks away.
