OpenAI, the company behind ChatGPT and the GPT models

From nonprofit lab to public benefit corporation

OpenAI was founded in San Francisco in 2015 with a stated mission: to ensure that artificial general intelligence benefits all of humanity. It started as a nonprofit. In 2019 it created a for-profit arm to pay for the enormous computing power it needed, still governed by the nonprofit.

On October 28, 2025, OpenAI announced an updated structure. The nonprofit is now called the OpenAI Foundation, and the for-profit has become OpenAI Group PBC, a public benefit corporation. That legal form requires the company to pursue its mission alongside its commercial goals, and the Foundation appoints the board of the operating company.

The model families

The core of the business is the GPT family: large language models that have become multimodal, reading text and images, with variants for real-time voice, transcription and image generation. The developer documentation separates general-purpose flagship models from specialized ones for cybersecurity, life sciences, audio and images.

Alongside them, gpt-oss is a family of open-weight models released under the Apache 2.0 license. You can download, modify and use them commercially, subject to the accompanying usage policy. It is the exception in a strategy that otherwise remains largely closed.

How to access the models

• ChatGPT: the consumer assistant on the web, mobile and desktop. Plans and features are covered on the ChatGPT page.
• The OpenAI API: an API billed by the price per million tokens, documented on OpenAI’s developer portal.
• gpt-oss: download the weights and run them without going through OpenAI’s servers.

What sets OpenAI apart

In practice, OpenAI brought conversational AI to the mainstream when it launched ChatGPT in November 2022, and its API format has become a de facto standard: several rivals, such as DeepSeek and xAI, offer APIs that work with OpenAI’s client libraries. Its ecosystem is huge, but its flagship models remain black boxes: you cannot host them yourself or inspect their weights. To weigh them against other labs on price and capabilities, use the model comparison tool.