Models, applications and agents
OpenAI develops the GPT model family and products built around it. ChatGPT began as a conversational research preview on November 30, 2022. Codex is its coding-agent product, with interfaces for engineering work across local and cloud environments. These are related parts of the organization: a model supplies capabilities, while an application or agent adds tools, interaction and execution. [5][3]
OpenAI's December 2015 announcement established it as a nonprofit research organization. Its current structure has two distinct entities: the nonprofit OpenAI Foundation and the commercial OpenAI Group public benefit corporation. Describing the entire organization simply as a public benefit corporation loses that distinction. [2][1]
Nonprofit control of a commercial group
Following the October 2025 recapitalization, the Foundation retained control through special voting and governance rights. It appoints the Group's board and can replace its directors. The structure page therefore establishes control independently of any particular investor's economic stake; an investment is not the same as control of OpenAI. [1]
The October 7, 2026 GPT-6 announcement concerned ChatGPT's interactive answers and reasoning experience, beginning a staged rollout across its plans. The announcement explicitly separated that change from Work and Codex. Product availability should therefore be read at the level of the named product and release, rather than assuming every OpenAI surface changed at once. [4]