OpenAI API platform

What it is

The OpenAI API platform is OpenAI's commercial platform for developers. It gives programmers access to the AI models OpenAI develops, so that they can build those models into their own applications ^[1].

Background

OpenAI announced the API on 11 June 2020 as a way to access new AI models it had developed, starting with GPT-3 ^[1]. By 25 March 2021 OpenAI said more than 300 applications were delivering GPT-3-powered features through the API ^[3]. On 18 November 2021 the API became available with no waitlist, a change OpenAI linked to progress on safety ^[2].

Key features

The platform's growth was measured by how many developers used it. At its first DevDay, on 6 November 2023, OpenAI said that more than two million developers were building on the API ^[4]. The API has since added tools aimed at agents, which are programs that carry out multi-step tasks. The Responses API and the open-source Agents SDK were introduced on 11 March 2025 ^[6].

At DevDay on 6 October 2025 OpenAI announced AgentKit, built on the Responses API. Sam Altman said that 4 million developers had built with OpenAI and that the API processed more than 6 billion tokens per minute ^[5][7]. A token is the small unit of text that a model reads and writes.

Its place in the story

The platform has had a sequence of replacements. The older Assistants API was scheduled to shut down on 26 August 2026, and Agent Builder, Evals and reusable prompts were scheduled to shut down on 30 November 2026 ^[8]. Over the same period, newer interfaces arrived. An Agents API exposing the managed Codex harness entered public beta on 10 September 2026, and a Decisions API became generally available on 22 October 2026 ^[6].

The platform's history runs from a single model family in 2020 to a set of tools for building and running agents in 2026. It is the route through which many outside developers have used OpenAI's models, so changes to its interfaces affect those who build on it.