DALL-E
by OpenAI
OpenAI's text-to-image model, integrated into ChatGPT and the Images API
Last reviewed 2026-06-20
DALL-E is OpenAI's family of text-to-image models, first announced on January 5, 2021. It generates images from natural-language prompts and went through three main versions: the original DALL-E (a 12-billion-parameter GPT-3 variant using a discrete VAE and CLIP), DALL-E 2 (April 2022, a CLIP-conditioned diffusion model with higher resolution and realism), and DALL-E 3 (released September 2023, integrated into ChatGPT and Microsoft's Bing Image Creator and Copilot), which focused heavily on prompt fidelity. The name is a portmanteau of the Pixar robot WALL-E and the surrealist painter Salvador Dali. DALL-E is a generation tool, not an autonomous agent: a person writes a prompt and the model returns images that the user then re-prompts, edits, or regenerates. As of late 2025 and into 2026, OpenAI superseded DALL-E with its newer GPT Image models (GPT Image 1 / 1.5 / 2), which replaced DALL-E 3 as the default image generator in ChatGPT. OpenAI announced on November 14, 2025 that the dall-e-2 and dall-e-3 API snapshots would be deprecated and removed from the API on May 12, 2026, directing developers to the GPT Image endpoints instead.
What it can do
Generate images from text prompts
AssistantProduces images from natural-language descriptions. DALL-E 3 (released September 2023) emphasized prompt fidelity, following longer, more detailed prompts more closely than DALL-E 2.
sourceIn-chat image generation via ChatGPT
AssistantDALL-E 3 was integrated directly into ChatGPT (October 2023), letting users describe an image in conversation and have ChatGPT expand the prompt and generate it; OpenAI later replaced this with its newer GPT Image models as the default.
sourceImage generation and editing via the OpenAI Images API
AssistantThe dall-e-2 and dall-e-3 model snapshots were exposed through OpenAI's Images API for programmatic generation (and, for DALL-E 2, edits and variations). OpenAI deprecated these snapshots and scheduled their removal from the API for May 12, 2026, directing developers to gpt-image-2, gpt-image-1, or gpt-image-1-mini.
sourceContent safety filtering
AssistantOpenAI applies content moderation to DALL-E, reportedly blocking prompts involving named public figures and filtering objectionable content, though earlier reports noted some filters could be bypassed with substitute wording.
source
Strengths
- +Strong prompt fidelity in DALL-E 3 (followed long, detailed prompts more literally than many peers)
- +Tightly integrated into ChatGPT and Microsoft Copilot/Bing, so generation happened inside tools people already used
- +Simple, well-documented Images API with usage-based per-image pricing
Limitations
- −Deprecated: dall-e-2 and dall-e-3 API snapshots were scheduled for removal on May 12, 2026, in favor of GPT Image models
- −An assistant, not an autonomous agent: the human prompts, curates, and re-rolls every output
- −Content filters were sometimes criticized as overly restrictive (Microsoft Copilot) and, earlier, as bypassable
Overview
DALL-E is OpenAI's family of text-to-image models, first announced on January 5, 2021. The name combines the Pixar robot WALL-E and the surrealist painter Salvador Dali. It generates images from natural-language prompts and is a creative generation tool, not an autonomous agent: a person describes an image, the model returns options, and the person re-prompts, edits, or regenerates until satisfied.
What it does
The lineage runs through three main versions. The original DALL-E was a 12-billion-parameter GPT-3 variant that used a discrete VAE and CLIP. DALL-E 2 (April 2022) switched to a CLIP-conditioned diffusion approach with higher resolution and realism. DALL-E 3 (released September 2023) focused on prompt fidelity, following longer and more detailed prompts more closely, and was integrated into ChatGPT in October 2023 as well as Microsoft's Bing Image Creator and Copilot. OpenAI applies content moderation, reportedly blocking prompts involving named public figures and filtering objectionable content. As of late 2025 and into 2026, OpenAI replaced DALL-E with its newer GPT Image models as the default generator in ChatGPT.
Integrations & setup
DALL-E was used in three main ways: inside ChatGPT (describe an image in chat and ChatGPT expands the prompt and generates it), inside Microsoft's Bing Image Creator and Copilot, and programmatically through OpenAI's Images API (model snapshots dall-e-2 and dall-e-3). It runs on OpenAI's own proprietary models.
Pricing
Through the API, DALL-E was billed per image on a usage basis. DALL-E 3 standard quality cost about $0.040 per 1024x1024 image and $0.080 for larger sizes, with HD quality at roughly $0.080 to $0.120 per image. In-chat generation was bundled into ChatGPT subscriptions. With deprecation, generation moves to the GPT Image models and their pricing.
Best for / not for
Historically a good fit for anyone wanting strong prompt-following image generation inside tools they already used (ChatGPT, Copilot) or via a simple API. As of 2026 it is deprecated: new work should target OpenAI's GPT Image models (gpt-image-2, gpt-image-1, gpt-image-1-mini), since the dall-e-2 and dall-e-3 API snapshots were scheduled for removal on May 12, 2026. Not suited to anyone needing autonomous, multi-step agentic behavior; it is a single-shot generation tool.
Alternatives
Midjourney offers a more stylized, opinionated art look; Ideogram is strong on legible in-image text; Stable Diffusion is the open-weights option for self-hosting and fine-tuning; Recraft targets brand and vector design with an API. Within OpenAI's own stack, the direct successor is the GPT Image model family.
What people are saying
We aggregate real LinkedIn discussion into sentiment for the agents people search most. DALL-E isn't tracked yet, want it added? Request tracking.
FAQ
Is DALL-E an AI agent?+
No. DALL-E is a text-to-image generation model. A person writes a prompt and the model returns images, which the user then re-prompts or edits. It operates at the assistant level with no independent multi-step action.
Is DALL-E still available in 2026?+
OpenAI superseded DALL-E with its GPT Image models, which replaced DALL-E 3 as the default image generator in ChatGPT. OpenAI announced on November 14, 2025 that the dall-e-2 and dall-e-3 API snapshots would be removed from the API on May 12, 2026, pointing developers to the GPT Image endpoints (gpt-image-2, gpt-image-1, gpt-image-1-mini).
What did DALL-E 3 cost via the API?+
Before deprecation, DALL-E 3 standard quality was priced around $0.040 per 1024x1024 image and $0.080 for larger sizes, with HD quality at roughly $0.080 to $0.120 per image, billed per image on a usage basis.
Sources
- DALL-E (Wikipedia) · accessed 2026-06-20
- DALL-E deprecations (OpenAI API docs) · accessed 2026-06-20
- DALL-E 3 (OpenAI) · accessed 2026-06-20
- Deprecation Reminder: DALL-E will be shut down on May 12, 2026 (OpenAI Developer Community) · accessed 2026-06-20
Last reviewed 2026-06-20