OpenAI closes DALL-E in ChatGPT, gives all users its newer image tool for free

Something changed inside ChatGPT that most users won’t notice until they look for the tool that isn’t there. The DALL-E GPT — a dedicated ChatGPT mode built specifically for image generation — has been retired, replaced by a newer built-in capability running on a more capable model, now available at no cost to every ChatGPT user.

The replacement is called ChatGPT Images, and it runs on gpt-image-1 and gpt-image-1-mini, OpenAI‘s current image models. The upgrade matters: the newer models handle complex scene descriptions with more accuracy than their predecessor, produce images with fewer visual artifacts, and understand compositional requests — things like specific lighting conditions or spatial relationships between objects — that earlier versions handled inconsistently. OpenAI made ChatGPT Images the default image tool across all account tiers, free and paid alike.

One practical step is worth taking before the transition is fully complete. Any image created with the DALL-E GPT lives inside the conversation where it was generated. When those conversations are eventually deleted, the images go with them. OpenAI’s guidance, published in its release notes when the shutdown was first announced, is to download anything worth keeping. To do this: open ChatGPT in a browser or the mobile app, navigate to the Images section, and select each image to save it locally. ChatGPT’s rebuilt search function can help locate images from older conversations.

The DALL-E GPT’s retirement fits into a broader consolidation OpenAI has been carrying out this year. The same DALL-E models were pulled from the developer API in May. Earlier this month, OpenAI retired the o3 reasoning model from ChatGPT following a ninety-day sunset period. The pattern is deliberate: fewer entry points that each do more, rather than a growing catalog of specialized tools that each require their own learning curve.

Not every capability transfers cleanly to the new arrangement. The premium ‘Images with thinking’ variant — which applies reasoning steps before generating, useful for technically complex or ambiguous prompts — remains exclusive to paid subscribers. Free accounts get the standard gpt-image-1 output, a meaningful step up from what came before, but the highest-capability image experience still sits behind a paywall. Third-party applications or workflows that connected directly to the DALL-E GPT will also need to redirect their users to the ChatGPT Images interface through the standard ChatGPT platform. Custom GPTs that include image generation are unaffected by today’s change.

ChatGPT Images is available globally across ChatGPT’s web and mobile platforms on all account tiers. OpenAI has indicated that ‘Images with thinking’ — currently limited to paid plans — is expected to reach more users in future updates.

Tags: OpenAI, chatgpt, AI image generation, ChatGPT Images, DALL-E, gpt-image-1