OpenAI · Tools

DALL-E

DALL-E — demo listing for marketupdate.

At a glance

Pricing model
freemium
Price
Pricing on request
Hosting region
Global
Vendor
OpenAI
Report data / suggest a correction

Profile

DALL-E is OpenAI's original text-to-image generation model line — DALL·E 2 (2022) and DALL·E 3 (2023) — which turned written prompts into images inside ChatGPT and through a pay-per-image API. As of July 2026 it is a legacy product: OpenAI deprecated both models on November 14, 2025, and shut down all API access to DALL-E on May 12, 2026, moving image generation entirely to its newer GPT Image model family. There is no ongoing way to use DALL-E today — it survives only in OpenAI's documentation as a reference point for anyone migrating an old integration or comparing historical pricing.

Who builds it

DALL-E is built by OpenAI, the company behind ChatGPT, the GPT API, and Sora. It was OpenAI's first dedicated image-generation product, predating the GPT Image models that power image creation in ChatGPT and the API today. OpenAI's replacement lineage moved from DALL-E's older diffusion-based architecture to the newer GPT Image family, starting with gpt-image-1 and continuing through gpt-image-1.5 and the current gpt-image-2.

Core features

  • Text-to-image generation — DALL·E 2 and DALL·E 3 generated images purely from natural-language prompts, with no image input required.
  • Multiple output sizes — DALL·E 3 supported 1024×1024, 1024×1536, and 1536×1024 pixel outputs; DALL·E 2 offered comparable square and rectangular formats at lower resolutions.
  • ChatGPT integration — DALL·E 3 was built directly into ChatGPT Plus and Enterprise as the default image generator, before being superseded by GPT Image models.
  • Simple, prompt-only API — the dall-e-2 and dall-e-3 model endpoints accepted a text prompt and a size parameter and returned finished images, with far less editing and instruction-following depth than current GPT Image models.
  • Fully deprecated — as of the shutdown date, neither the dall-e-2 nor the dall-e-3 endpoint accepts new API requests.

Pricing

DALL-E can no longer be purchased or used — it does not appear on OpenAI's current API pricing page. While the models were live, OpenAI's own model documentation listed per-image API pricing of $0.04–$0.08 per image for DALL·E 3, depending on resolution and quality setting, and $0.016–$0.02 per image for DALL·E 2. There was never a metered free tier; usage in ChatGPT required a paid Plus or Enterprise subscription. OpenAI announced the deprecation on November 14, 2025, and shut the API down on May 12, 2026, pointing developers to gpt-image-2, gpt-image-1, or gpt-image-1-mini as replacements.

Who it's for

DALL-E is now relevant mainly to developers auditing or migrating legacy integrations built on the old dall-e-2/dall-e-3 endpoints, and to anyone who needs accurate historical context on OpenAI's image-generation lineage — for example when comparing what changed between DALL-E's diffusion-based approach and the newer, more instruction-following GPT Image models. It's also a useful reference point for content teams and researchers writing about the history of AI image generation, since DALL·E 2's 2022 launch and DALL·E 3's ChatGPT integration were milestone moments for the category. Anyone starting a new image-generation project today, however, should use OpenAI's current GPT Image models instead, since DALL-E itself is no longer accessible in any form — neither through the API nor through ChatGPT.

Frequently asked questions

What is DALL-E?

DALL-E is OpenAI's original text-to-image generation model line, consisting of DALL·E 2 (released 2022) and DALL·E 3 (released 2023), which generated images directly from natural-language prompts. It was OpenAI's dedicated image model before the company moved image generation to its newer GPT Image family.

Is DALL-E still available in 2026?

No. OpenAI deprecated DALL·E 2 and DALL·E 3 on November 14, 2025, and fully shut down API access to both models on May 12, 2026. Since then, the dall-e-2 and dall-e-3 endpoints no longer accept requests.

How much did DALL-E cost via the API?

According to OpenAI's own model documentation, DALL·E 3 cost $0.04–$0.08 per image depending on resolution and quality (1024×1024, 1024×1536, 1536×1024), while DALL·E 2 cost $0.016–$0.02 per image. These prices applied while the models were live and no longer appear on OpenAI's current pricing page.

What replaced DALL-E, and how does it relate to GPT Image?

OpenAI's official deprecation notice directs former DALL-E users to gpt-image-2, gpt-image-1, or gpt-image-1-mini — the GPT Image family that is now OpenAI's current image-generation lineup. GPT Image 2 is described by OpenAI as its state-of-the-art image generation model for fast, high-quality image generation and editing, built on a different (non-diffusion) approach than DALL-E.

How was DALL-E accessed — ChatGPT or the API?

DALL-E was available both ways: as a text-to-image endpoint in the OpenAI API (model IDs dall-e-2 and dall-e-3) and, for a period, as ChatGPT's built-in image generator on Plus and Enterprise plans. Both access paths have since been retired in favor of GPT Image models.

Was DALL-E free to use?

No. DALL-E was never free. API access was billed per image ($0.016–$0.08 depending on model, size, and quality), and ChatGPT access required a paid Plus or Enterprise plan. OpenAI's documentation lists no free tier for either DALL·E 2 or DALL·E 3.