DALL-E
OpenAI's pioneering text-to-image model series, retired in 2026.
Retired
DALL-E appears to be discontinued or shut down — DALL-E 2/3 removed from API May 2026; DALL-E GPT retired Aug 2026 This page is kept as an archive record and is excluded from search engines.
About
DALL-E was OpenAI's flagship text-to-image model line, starting with the original DALL-E in 2021 and reaching mainstream awareness with DALL-E 2 (2022) and DALL-E 3 (2023). DALL-E 3 introduced a GPT-4 powered prompt relabeler that rewrote user prompts before generation, which made it unusually good at following detailed instructions and rendering text inside images. It shaped how millions of people first experienced AI image generation.
Historical entry: DALL-E 3 was removed from ChatGPT in December 2025, and both DALL-E 2 and DALL-E 3 were fully deprecated and removed from the OpenAI API on May 12, 2026. The official DALL-E GPT was retired from ChatGPT on August 30, 2026. OpenAI's image surface is now the gpt-image family (gpt-image-1, 1.5, and gpt-image-2, an autoregressive architecture with native reasoning and stronger text rendering). This entry is kept for reference; new work should target ChatGPT Images / the gpt-image models instead.
What it can do
- Text-to-image generation (discontinued)
- Prompt relabeling for detailed instruction following
- Retired from ChatGPT December 2025; API deprecated May 12, 2026; DALL-E GPT retired August 30, 2026
- Succeeded by the gpt-image model family
Setup
No longer available. Use OpenAI's ChatGPT Images (gpt-image models) via ChatGPT or the images API instead.
Prerequisites
None — product retired. Existing DALL-E generations may need to be downloaded from ChatGPT before surfaces are removed.
Use cases
Works with
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 71; Trending score 65/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.