If you’re researching DALL·E in mid-2026, there’s something important you need to know before anything else: DALL·E 3 was officially retired on May 12, 2026. OpenAI shut it down and replaced it with a newer model called GPT Image 1.5, built directly into ChatGPT. This guide explains what actually happened, what replaced it, and what that means if you want AI image generation today.
What Happened to DALL·E
DALL·E 3 was OpenAI’s dedicated text-to-image model, called externally by ChatGPT as a separate system. In March 2025, OpenAI began transitioning to GPT Image 1, and DALL·E 3 was fully deprecated from the API on May 12, 2026. The current model, GPT Image 1.5, works differently: it’s autoregressive and built natively into ChatGPT’s conversation flow, rather than being a separate diffusion model ChatGPT calls out to.
In practice, this means image generation now understands your instructions more precisely, renders text inside images accurately, and lets you edit specific parts of an image conversationally instead of regenerating the whole thing from scratch.
What It Costs Now
Since image generation lives inside ChatGPT rather than being a standalone product, pricing follows ChatGPT’s plan structure:
| Plan | Price | Image generation |
| Free | $0 | Roughly 2-3 images per day |
| Plus | $20/mo | Roughly 50images per 3-hour window |
| Pro | $200/mo | Effectively unlimited, subject to fair-use limits |
| Business/Enterprise | Custom | “Virtually unlimited,” subject to fair-use policy |
For developers building image generation into their own apps, the API bills per image rather than per subscription: standard-quality images start around $0.011 with the newer Mini model, while higher-quality outputs run up to roughly $0.167 each.
What’s Actually Good About the Current Version
- Included with a subscription you may already have: if you already pay for ChatGPT Plus, image generation costs nothing extra — a real advantage over standalone tools that require a separate subscription
- Best-in-class text rendering: GPT Image 1.5 handles text inside images (logos, signage, labels) noticeably better than most competitors, including Midjourney
- Conversational editing: you can ask for a specific change to part of an image without starting over, which dedicated image generators typically can’t do
Where It Falls Short
- Generation speed lags behind dedicated tools — a single image can take one to two minutes, compared to seconds on purpose-built generators
- Artistic and cinematic quality still trails Midjourney for creative, atmospheric work
- Free tier limits (2-3 images/day) are tight enough to feel more like a demo than a usable tool
- OpenAI’s rate limits (roughly 50 images per 3-hour window on Plus) aren’t officially published, so exact numbers shift based on server load
ChatGPT Images vs. Midjourney: The Honest Comparison
These two tools solve different problems, and plenty of professionals use both. ChatGPT’s image generation is the stronger choice for text rendering, logos, product mockups, and iterative editing within a conversation. Midjourney remains the stronger choice for pure artistic quality, cinematic lighting, and atmospheric or conceptual art. If your work leans commercial and text-heavy, ChatGPT is the more practical tool; if it leans artistic and exploratory, Midjourney still has the edge.
Who Should Use It
Free is fine only for testing the waters — 2-3 images a day won’t support any real workflow.
Plus ($20/mo) is the practical entry point if you need images regularly alongside writing and research — especially since you likely already need ChatGPT Plus for other reasons.
Pro ($200/mo) only makes sense if you’re generating high volumes daily and hitting Plus’s limits consistently.
Bottom Line
The tool most people still call “DALL·E” no longer technically exists — it’s been fully absorbed into ChatGPT as GPT Image 1.5, with noticeably better text accuracy and conversational editing than its predecessor. If you already pay for ChatGPT Plus, you already have access to it at no extra cost. If pure artistic quality is your priority, it’s still worth comparing directly against Midjourney before choosing.
Transparency note: this analysis is based on OpenAI’s official deprecation announcements and independently verified reporting as of July 2026, not the author’s direct hands-on use of every plan tier.
_________________________________________________________________________________________________________
Last updated: July 2026. Pricing, limits, and product names are subject to change; always verify on OpenAI’s official pricing page before subscribing.