DALL-E 3

DALL-E 3 demo video

OpenAI's 2023 launch demo, kept as a record of what DALL-E 3 did. The model has since been retired.

►

Official video by OpenAI via YouTube, embedded for reference. ToolChase does not host or claim this video.

What is DALL-E 3?

DALL-E 3 was OpenAI's image generation model, built into ChatGPT and reachable through the API. It is now retired on both counts: OpenAI shut down the dall-e-3 and dall-e-2 API models on May 12, 2026, and removed the official DALL-E GPT from ChatGPT on August 30, 2026. OpenAI's own DALL-E 3 page is now an archived research post that points visitors to ChatGPT Images 2.0, and its help center answers the question plainly: you cannot generate images with DALL-E in ChatGPT any more. This page is kept as a record of what the model was and where its job went.

What DALL-E 3 was good at, from 2023 until its retirement, was doing what you asked. It followed complex, multi-part prompts closely instead of interpreting them, which made it useful for illustrations, diagrams and marketing assets where specific elements had to land in specific places. It also rendered legible text inside images, a real weakness of every competitor of its era. The workflow was conversational: you described an image in plain language, ChatGPT rewrote your prompt, and you refined the result by asking for changes rather than by learning syntax.

Its replacement in ChatGPT is ChatGPT Images 2.0, introduced April 21, 2026 and available on every tier including the free one, with an editor for selecting and changing part of an image and an aspect ratio picker. Images with thinking is limited to Plus, Pro and Business. For developers the replacements are the GPT Image models (gpt-image-2, gpt-image-1.5, gpt-image-1 and gpt-image-1-mini), which are billed per million tokens rather than per image. The old criticisms of DALL-E 3 still belong on the record: its aesthetic quality trailed Midjourney on artistic and photorealistic work, and its safety filters were the most restrictive of the major generators, refusing many prompts other tools handled without complaint.

DALL-E 3 Pricing

DALL-E 3 has no price, because you cannot buy it. What follows is the cost of the thing that took over its job, image generation inside ChatGPT, at the plan prices listed on OpenAI's own pricing page in September 2026. These are true monthly rates, not annual rates divided by twelve: OpenAI sells monthly billing on Go and Plus and reserves annual billing for Business and Enterprise.

ChatGPT Free, $0/mo, limited and slower image generation. ChatGPT Images 2.0 is available on all tiers, so this is a real free plan and not a trial.

ChatGPT Go, $6/mo, more image creation than Free, plus longer memory and more uploads. OpenAI notes this plan may include ads.

ChatGPT Plus, $20/mo, more complex and accurate image creation, and the tier where Images with thinking starts (Plus, Pro and Business only).

ChatGPT Pro, from $120/mo, unlimited and faster image creation, subject to OpenAI's abuse guardrails. Read the "from" literally: Pro is sold at 5x or 20x more usage, so $120 is the entry rate rather than the only one.

API access, retired, OpenAI shut down dall-e-3 and dall-e-2 on May 12, 2026, so the old per-image rates are gone and dall-e-3 no longer appears in the model list or the pricing table. Programmatic image generation now runs on the GPT Image models, priced per million tokens instead of per image: gpt-image-2 at $8 input and $30 output, gpt-image-1.5 at $8 and $32, gpt-image-1 at $10 and $40, gpt-image-1-mini at $2.50 and $8. Batch rates are half of each.

Key Features

• Retired model, DALL-E 3 is no longer reachable: the API shut down May 12, 2026 and the official DALL-E GPT left ChatGPT on August 30, 2026. Everything below describes what it did
• ChatGPT integration, generated images through natural conversation, with automatic prompt refinement and iterative editing in the same chat
• High prompt adherence, followed detailed, multi-element prompts more accurately than most contemporaries, placing objects where they were described
• Text rendering in images, the strongest text-in-image capability of its era, producing readable words, labels, signs, and typography
• Inpainting & editing, select an area of a generated image and change it while keeping the rest intact
• API access (retired), dall-e-3 and dall-e-2 shut down May 12, 2026; programmatic image generation now runs on gpt-image-2, gpt-image-1.5, gpt-image-1 and gpt-image-1-mini
• Fixed aspect ratios, square (1024x1024), portrait (1024x1792) and landscape (1792x1024). ChatGPT Images 2.0 lifted this: it offers an aspect ratio picker and gpt-image-2 accepts edges up to 3840px within a 3:1 ratio
• Safety filters, built-in moderation blocked harmful, violent, or explicit content, and declined prompts naming public figures or living artists
• Conversational refinement, describe changes in plain language to iterate without re-prompting from scratch
• Style versatility, illustrations, photorealistic images, 3D renders, watercolors, pixel art and more from style instructions
• Commercial use, OpenAI's DALL-E 3 page still states the images you create are yours to reprint, sell or merchandise without asking permission

Pros & Cons

Best For

Nobody, as a live purchase. DALL-E 3 cannot be reached from ChatGPT or the OpenAI API any more, so there is no current use case for it and this entry stays up as a record. If you came for the workflow, describing an image in plain language and refining it by conversation, that is ChatGPT Images 2.0, on every ChatGPT plan including the free tier, with an editor and an aspect ratio picker. If you came for the API, use OpenAI's GPT Image models: gpt-image-2 for quality, gpt-image-1-mini when cost matters. If you came for readable text inside images, the trait DALL-E 3 was known for, test ChatGPT Images 2.0 and Ideogram against each other before committing.