Overview
- Tool name: DALL·E 3 (Retired)
- Developer: OpenAI
- Official website: https://openai.com/index/dall-e-3
- Category: AI Models
- Status: API shut down on May 12, 2026 - no longer available
DALL·E 3 was OpenAI's third-generation text-to-image model, released in 2023, and the first widely used image generator that reliably followed long, detailed prompts. Its integration into ChatGPT made conversational image editing normal, and it genuinely influenced the field.
Its retirement happened in stages. In December 2025 OpenAI quietly switched the ChatGPT default to GPT Image 1.5. On March 4, 2026 DALL·E 3 stopped accepting new API deployments. On May 12, 2026 the DALL·E 2 and DALL·E 3 API model IDs were shut down, and on August 30, 2026 the official DALL·E GPT inside ChatGPT was retired. OpenAI's image generation now runs entirely on the GPT Image family.
The successor is the GPT Image family: the current line is GPT Image 2.5, shipped September 8, 2026 as gpt-image-2.5-flare (fast, up to 50% lower latency) and gpt-image-2.5-sunburst (editing precision). Unlike DALL·E's diffusion architecture, GPT Image models generate autoregressively, which is what enables instruction accuracy to move from roughly 90-95% to a claimed 99%, plus image-to-image editing. It is available inside ChatGPT across all tiers (Free through Enterprise) and billed per image token on the API: $8 per million image input tokens, $30 per million output, working out to roughly $0.006-$0.211 per 1024x1024 image depending on the quality setting.
Can you still use DALL·E 3?
No. The dall-e-3 and dall-e-2 API model IDs are gone; any code still referencing them will fail. Existing integrations need to migrate to the GPT Image endpoints. Image generation in ChatGPT switched automatically - users there need to do nothing.
Migration notes
- ChatGPT users: Nothing to do. You are already on GPT Image 2.5, which outperforms DALL·E 3 across the board.
- API developers: Replace dall-e-3 model strings with gpt-image IDs. Note the new models bill per image token rather than per image, support editing and output controls DALL·E 3 never had, and need fresh cost modeling.
- Choosing fresh in 2026: The realistic shortlist is GPT Image 2.5 (instruction following and editing), Midjourney (aesthetic quality), Ideogram (text rendering), Adobe Firefly (commercially safe assets), and FLUX.2 or Stable Diffusion (self-hosted).
Legacy and limitations
DALL·E 3 proved that long-prompt image generation could work at scale; its weaknesses - poor text rendering, shaky complex compositions - are largely fixed in its successors.
Frequently asked questions
Is DALL·E 3 shut down?
Yes. The API closed on May 12, 2026, and the DALL·E GPT inside ChatGPT was retired on August 30, 2026. OpenAI's current image model is GPT Image 2.5.
What replaced DALL·E 3?
Within the OpenAI ecosystem: GPT Image 2.5, built into ChatGPT and available via the API. For aesthetics, compare Midjourney; for text-in-image, Ideogram.
Are my old DALL·E images safe?
They live in your conversations. Deleting a chat deletes its images, so request a data export from Settings > Data controls if you want a backup.
*This page is not affiliated with OpenAI. Retirement timeline and successor details summarized as of October 2026; confirm with OpenAI's official documentation.*
