We make ChatGPT and our API available in connection with extensive safety measures, as further detailed below. We also make certain models⁠, such as our speech-to-text model Whisper and our image understanding model called CLIP, available on an open source basis after evaluating the potential risks of such releases.

We intend to continue to make ChatGPT available for free and will earn revenue from users and businesses that choose to pay for premium services. Given the high costs of developing and offering large scale foundation models, our organization is not profitable and does not expect to be profitable for the near future—our goal continues to be to make AI benefits broadly and safely available to the world.

OpenAI makes access to our leading foundation models available primarily through ChatGPT and our API.

Developing an advanced language model like GPT‑4 requires (1) teaching it intelligence, such as the ability to predict, reason, and solve problems, as well as (2) aligning it to human values and preferences. The former is done in a process called “pre-training”, which involves showing the model a vast amount of human knowledge over months. To then incorporate human choice into the model, we use a second step, called “post-training”, where we make the model safer and more usable.