What DALL-E is, and the short answer
DALL-E is gone. OpenAI retired the model on 12 May 2026 and replaced it, via GPT Image 1.5 and GPT Image 2, with GPT Image 2.5 (released 8 September 2026), built into ChatGPT.
So if you came here looking for DALL-E, there is nothing to sign up for. There is no DALL-E website, no DALL-E subscription and no DALL-E app. There is image generation inside ChatGPT, and that is a different product with a different name.
This review covers what replaced it, because that is the actual question.
What’s new in GPT Image 2.5: sharper detail, more natural lighting and texture, better preservation of people and products from reference photos, more reliable editing across a long back-and-forth, and generation up to 50% faster than 2.0. It adds a Sketch feature that turns a drawing into a generated image. The API ships two variants: Flare (faster) and Sunburst (more detailed but slower).
How to access OpenAI image generation in 2026
| Plan | Price | What you get for images |
|---|---|---|
| Free | $0/mo | Very limited generation with GPT Image, sometimes watermarked |
| Go | $8/mo | More generations per month, fewer restrictions |
| Plus | $20/mo | GPT Image 2.5 at higher volume, native 2K resolution |
| Pro | from $100/mo | 5-20x higher usage, generation priority |
There is no image-only tier. You are buying a ChatGPT subscription that happens to include images.
What actually changed
GPT Image 2.5 is better than DALL-E 3 was at prompt adherence and text rendering, and it outputs at higher resolution. That is a real improvement.
What did not change is the ceiling. Put a GPT Image 2.5 result next to Midjourney or Flux and the difference is obvious: OpenAI’s images are competent and generic, the others are striking.
The one thing it does better than anyone
Conversational editing. You generate an image, then say “make the background darker and remove the text on the left”, and it does that to the image you already have.
Every other tool makes you rewrite the prompt and generate again, which produces a different image rather than an edited one. For iterating toward a specific result without learning a tool, nothing else comes close.
GPT Image (formerly DALL-E) vs Midjourney vs Leonardo AI
| GPT Image (ex-DALL-E) | Midjourney | Leonardo AI | |
|---|---|---|---|
| Entry price | Included with ChatGPT Plus ($20/mo) | $10/mo | Limited free · from $12/mo |
| Aesthetic quality | ★★★☆☆ | ★★★★★ | ★★★★☆ |
| Ease of use | ★★★★★ | ★★★☆☆ | ★★★★☆ |
| Conversational editing | ★★★★★ | ★★☆☆☆ | ★★★☆☆ |
| Fine control (ControlNet, styles) | ★★★☆☆ | ★★★★☆ | ★★★★★ |
| Text rendering | ★★★★☆ | ★★★☆☆ | ★★★☆☆ |
Choose GPT Image (ChatGPT) if: you already pay for ChatGPT Plus and want to generate images without leaving the conversation, without worrying about complex prompts.
Choose Midjourney if: you prioritize artistic aesthetics above everything else and don’t mind a more technical workflow via Discord or the web.
Choose Leonardo AI if: you need fine control over styles, trainable models and tools built for professional visual production.
Who is this for in 2026?
People who already use ChatGPT: the most convenient option, since image generation comes included at no extra cost on top of the subscription.
Developers who used the DALL-E API: forced to migrate to gpt-image-1 or gpt-image-2, with new pricing and limits worth checking in OpenAI’s documentation before assuming old code still works.
Not recommended if: you’re specifically looking for “DALL-E” as a product - it no longer exists on its own, and for superior artistic quality there are better dedicated options like Midjourney.
What users say
Reaction to the sunset was mixed. On OpenAI’s developer forum, several users called the DALL-E 3 deprecation “a mistake”, pointing out they lost a stable, predictable model for one with different behavior and, in some cases, a higher per-image cost via the API.
On the other hand, regular ChatGPT users barely noticed the transition: for most, “asking for an image” simply started returning results with better context understanding and more legible text, with no explicit announcement of the engine change. Brand confusion is the most-cited problem: plenty of people still search for “DALL-E” without realizing they’re already using something else.
Verdict
The 2.5 rating is not a judgement on the quality of GPT Image 2.5. It reflects that DALL-E, the thing you searched for, does not exist, and what replaced it is a feature of another subscription rather than a product you can choose.
If you already pay for ChatGPT, you have decent image generation and you should use it before buying anything else. If you do not, do not subscribe to ChatGPT for the images: $10/mo of Midjourney or the free tier of Ideogram will serve you better.
👉 The best AI image generators in 2026
👉 Midjourney vs DALL-E vs Firefly: which to choose
Community rating
Affiliate link - no extra cost to you