DALL·E 3

DALL·E 3 is an image generation model from OpenAI, announced on 20 September 2023 ^[1][2]. A user describes a scene in ordinary language and the model returns a picture. It was the third model in OpenAI's DALL·E line, and it is remembered less for sitting in a standalone tool than for being placed inside ChatGPT ^[1].

Where it came from

DALL·E 3 is the successor of DALL·E 2 ^[1]. The change OpenAI emphasised was not only the pictures themselves but the layer between the user and the model: OpenAI said DALL·E 3 was built natively on ChatGPT, which generates a tailored, detailed prompt from a user's idea and can tweak an image on request ^[4]. In practice that means a short, vague request can be expanded into a fuller description before anything is drawn, and a follow-up request is handled by adjusting the existing image rather than by starting over with a new prompt ^[4].

How it was released

OpenAI announced the model on 20 September 2023 ^[1][2], then released it in ChatGPT for ChatGPT Plus and ChatGPT Enterprise customers in October 2023 ^[1]. Access through OpenAI's API and its Labs platform came in early November 2023 ^[1]. Microsoft also used it: DALL·E 3 was integrated into Bing Image Creator, and image generation in Microsoft Copilot ran on DALL·E 3 ^[1].

What it does

The core behaviour is text in, image out, with ChatGPT acting as the intermediary that writes the detailed prompt ^[4]. OpenAI also stated a rule about ownership: images created with DALL·E 3 belong to the user, who does not need permission to reprint, sell or merchandise them ^[4].

The model was also built with restrictions. It was designed to block requests for art in the style of living artists, and it used input and output filters, blocklists and refusals in ChatGPT to limit explicit content ^[1]. OpenAI said DALL·E 3 included mitigations to decline requests for a public figure by name ^[4]. These limits sit in the prompt-handling layer as well as around the image output, which is consistent with the idea of a model wrapped in a chat product rather than called directly.

Where it sits in the line

DALL·E 3 was the successor of DALL·E 2 and was itself succeeded by GPT Image in 2025 ^[1]. In March 2025, GPT Image's native image generation replaced DALL·E 3 inside ChatGPT ^[1]. For developers, the end came on a published schedule: OpenAI announced the API deprecation on 14 November 2025, and dall-e-3 was shut down on 12 May 2026 ^[3].

Read as a sequence, the model marks a stage in which image generation stopped being a separate destination and became a feature of a conversation: the same system that answers questions writes the prompt, draws the picture and revises it on request ^[4]. The later replacement of DALL·E 3 by GPT Image inside ChatGPT, and the eventual retirement of the dall-e-3 API endpoint, closed out that stage ^[1][3].