DALL-E

DALL-E is OpenAI’s older image generation model family. DALL-E 2 and DALL-E 3 are still useful as historical labels, but they are not the path for new OpenAI image builds. OpenAI’s current image-generation docs center GPT Image models, and dall-e-2 and dall-e-3 were removed from the API on May 12, 2026 per OpenAI’s deprecation schedule.

What it was

DALL-E launched in 2021 as OpenAI’s first text-to-image model family. DALL-E 2 followed in 2022, and DALL-E 3 later became the better-known ChatGPT-era image model.

Through 2023 and 2024 it served as the default OpenAI image surface, pitched against Midjourney and Stable Diffusion. Its strengths were prompt adherence and text rendering for a consumer chat product. Its weakness was aesthetic ceiling: Midjourney routinely beat it on stylized and cinematic output.

What replaced it

OpenAI’s current developer path is GPT Image. The image-generation guide describes GPT Image as the latest and most advanced OpenAI image-generation model family, with better instruction following, text rendering, editing, and world-knowledge use than the DALL-E series.

For API work, that means choosing current GPT Image model IDs and endpoint behavior rather than preserving a DALL-E abstraction. The Image API now centers GPT Image models; the legacy dall-e-2 and dall-e-3 IDs were retired on May 12, 2026. The Responses API image-generation tool is designed around GPT Image.

For broader site context on OpenAI’s image-generation transition, see the local coverage of ChatGPT Images 2.0 and gpt-image-2, but use the official OpenAI docs linked below as the source of truth for current API model IDs.

API status

As verified on 2026-06-12:

• OpenAI’s deprecation schedule retired dall-e-2 and dall-e-3 from the API on May 12, 2026. New calls should target GPT Image model IDs.
• OpenAI’s image guide now centers GPT Image as the current path for image-generation work; DALL-E appears only as historical context.
• GPT Image models support editing and output controls that the older DALL-E 3 Image API endpoint did not.
• Any code still referencing the dall-e-3 model string will fail against the live API and needs migration.
• Current pricing should be read from OpenAI’s pricing page because GPT Image costs depend on image tokens, quality, size, and input images.

Migration notes

Treat DALL-E as a historical label when auditing old prompts, docs, and comparisons. If a vendor, workflow, or internal playbook still says “DALL-E,” check whether it means the retired ChatGPT image surface, the DALL-E 3 API, or OpenAI image generation in general. Those are no longer the same operational choice.

For buyers, the decision path is straightforward. If you want OpenAI’s current image workflow inside a consumer or team assistant, evaluate ChatGPT and the current GPT Image path. If you want a pure image model with a different aesthetic ceiling, compare Midjourney, Flux, Adobe Firefly, and other active image tools instead of ranking DALL-E as if it were a current buying option.

For developers, the key risk is migration timing. New integrations should not be planned around the DALL-E brand. Update product copy, prompt libraries, tests, and API abstractions so they refer to current model IDs and a current source of truth. Keep this page for provenance, not as a buying recommendation.

Migration checklist

• Inventory any dall-e-2 or dall-e-3 model strings in code, prompt tests, docs, examples, and customer-facing copy.
• Replace product copy that says “DALL-E” when the real feature now means OpenAI image generation generally.
• Re-test prompt libraries because GPT Image behavior, output controls, editing support, and moderation settings are not identical to DALL-E behavior.
• Update pricing assumptions from fixed DALL-E-era per-image thinking to current GPT Image token, size, quality, and editing costs.
• Keep compatibility notes for historical comparisons, but route new product decisions to the current OpenAI image docs.