DALL-E 3
OpenAI's retired image generator, formerly built into ChatGPT
- Category
- Image Generation
- Pricing
- Retired May 12, 2026; was included with ChatGPT Plus ($20/mo) or ~$0.04-0.08/image via API — succeeded by GPT Image 2
- Best for
- DALL-E 3 is now relevant only as historical context — anyone evaluating OpenAI's image generation today should look at GPT Image 2 (ChatGPT Images), which replaced it.
- Official site
- openai.com/index/chatgpt
- Last updated
- August 2026
DALL-E 3 was OpenAI's third-generation text-to-image model, released in 2023 and made available inside ChatGPT (Plus, Team and Enterprise tiers) as well as through the standalone OpenAI API. It was, for much of 2023-2025, one of the most widely used AI image generators due to its tight integration with ChatGPT's conversational interface, which let users refine prompts in natural language rather than needing prompt-engineering skill.
As of this writing, DALL-E 3 has been discontinued: OpenAI retired DALL-E 2 and DALL-E 3 on May 12, 2026, and announced on July 31, 2026 that the legacy DALL-E interface inside ChatGPT would be fully shut down by August 30, 2026. It has been replaced by GPT Image 2 (branded as 'ChatGPT Images' inside the product), a natively integrated image model built directly into OpenAI's newer GPT models rather than shipped as a separate standalone model. This page covers DALL-E 3 as a legacy reference — anyone evaluating OpenAI's current image generation should look at GPT Image 2 instead.
DALL-E 3 is now relevant only as historical context — anyone evaluating OpenAI's image generation today should look at GPT Image 2 (ChatGPT Images), which replaced it. Historically, DALL-E 3 was best suited for ChatGPT users who wanted quick, conversational image generation without learning prompt engineering, and less suited for professionals who needed fine-grained creative control or the most artistically distinctive output — those users were generally better served by Midjourney or, for commercially-licensed work, Adobe Firefly.
Key features
Natural-language prompt refinement (legacy)
Because DALL-E 3 lived inside ChatGPT, users could describe what they wanted conversationally and ask for iterative changes without needing precise prompt syntax.
Automatic prompt rewriting (legacy)
ChatGPT would silently expand short user prompts into more detailed internal prompts to improve output quality and prompt adherence.
In-chat image generation (legacy)
Images generated directly in the same conversation as text, letting users combine research, copywriting and visuals in one session.
API access with size/quality tiers (legacy)
Developers could call DALL-E 3 directly via the OpenAI API, choosing standard or HD quality and several aspect ratios, billed per image.
Improved text rendering vs. DALL-E 2 (legacy)
DALL-E 3 was a meaningful step up from DALL-E 2 at rendering legible text and complex multi-subject scenes, though it still struggled with long or complex typography.
Content moderation and safety filters (legacy)
Built-in filters restricted generation of certain content categories, including depictions of real public figures.
Successor: GPT Image 2 / ChatGPT Images
The current replacement is natively built into OpenAI's GPT models, offering higher resolution (up to 4K), improved text rendering accuracy and resolved color-cast issues that DALL-E 3 was known for.
Pricing breakdown
Legacy: Included with ChatGPT Plus
- Generations included as part of a ChatGPT Plus subscription
- No separate per-image charge inside the ChatGPT interface
- Subject to daily/rate limits
Legacy: API (standard quality)
- Billed per generated image
- Lower cost than HD tier
- Used for developer integrations
Legacy: API (HD quality)
- Higher-fidelity output
- Higher per-image cost than standard quality
Current: GPT Image 2 (successor)
- Replaces DALL-E 3 entirely as of the 2026 retirement
- Higher resolution and improved text accuracy
Pros and cons
Pros
- Conversational refinement inside ChatGPT made it far more approachable for non-designers than prompt-heavy competitors
- Strong at interpreting complex, multi-element scene descriptions in a single generation
- Noticeably better text rendering than its predecessor, DALL-E 2
- Tight ChatGPT integration meant no separate account or tool-switching was needed for many casual users
- Per-image API pricing was inexpensive relative to comparable commercial models at the time
- Safety and moderation systems made it a comparatively low-risk choice for mainstream/consumer use
Cons
- Now retired — DALL-E 3 is no longer usable in ChatGPT as of the 2026 shutdown, making it a dead end for new evaluation
- Struggled with hands, fine anatomical detail and complex typography even at its peak
- Results sometimes only loosely matched detailed or unusual prompts, requiring multiple regenerations
- Offered less artistic stylization than Midjourney by default, which some users saw as flatter or more generic output
- Content moderation was, for some users, overly conservative, blocking benign prompts
- No dedicated, verified listing on G2 or Capterra specific to DALL-E 3 makes independent review data thin and mixed with DALL-E 2 feedback
What reviewers say
G2's combined DALL·E listing (which does not cleanly separate DALL-E 2 from DALL-E 3 feedback) shows solidly positive sentiment at 4.1/5, driven by ease of use and creative idea generation. Product Hunt and general reviews echo this: users consistently praised how quickly DALL-E 3 turned plain-language descriptions into usable images, while criticizing its handling of hands, fine detail and complex text. Reddit and community sentiment around the May 2026 retirement was largely neutral-to-positive about the move, with users describing GPT Image 2 as a clear technical improvement, though some expressed frustration at having to migrate saved prompts and workflows on short notice.
Frequently praised
- Fast, accessible image generation directly inside a chat interface
- Strong at interpreting complex, multi-part prompts
- Noticeable text-rendering improvement over DALL-E 2
Frequently criticized
- Inconsistent hands and fine anatomical detail
- Struggled with complex or lengthy on-image typography
- Prompt adherence could miss subtle requested details, requiring retries
- Now fully retired, so it's no longer usable or actively supported
Alternatives to DALL-E 3
Frequently asked questions
Is DALL-E 3 still available?
No. OpenAI retired DALL-E 2 and DALL-E 3 on May 12, 2026, and shut down the legacy DALL-E interface inside ChatGPT entirely by August 30, 2026.
What replaced DALL-E 3?
GPT Image 2, marketed inside ChatGPT as 'ChatGPT Images,' replaced DALL-E 3. It's natively built into OpenAI's newer models rather than being a separate standalone model, and offers higher resolution and better text rendering.
Can I still access old DALL-E 3 images I generated?
OpenAI gave users a migration window after the July 31, 2026 announcement to save existing work before the August 30, 2026 full shutdown; check OpenAI's help center for current export options.
How much did DALL-E 3 cost via the API?
Historically around $0.04 per standard-quality 1024x1024 image and roughly $0.08 for HD quality, billed per image through the OpenAI API.
Was DALL-E 3 better than Midjourney?
It depended on the use case: DALL-E 3 was generally easier to use conversationally and better at literal prompt-following, while Midjourney was consistently rated higher for artistic polish and stylization.
Do I need a new account to use GPT Image 2?
No — GPT Image 2 is accessed through the same ChatGPT or OpenAI API account previously used for DALL-E 3; no separate signup is required.
Ready to try DALL-E 3?
Head to the official site to explore pricing and start a free trial where available.
Visit DALL-E 3 →