Ten days after it was announced, DALL-E 3 appears to have already been made available to the public through Microsoft's Image Creator tool as part of its browser-based AI suite.

Early Saturday morning, digital art creator communities on Reddit and Twitter discovered significantly improved images and text interpretation capabilities were available through Bing. The upgrade, widely believed to be DALL-E 3, is still not available via the OpenAI website.


DALL-E 3 is a much-anticipated upgrade to the immensely popular DALL-E 2 AI image generator from OpenAI. While ChatGPT, OpenAI's generative AI chatbot and text-based tool, enjoys a dominant position in its space, DALL-E faces stiff competition from tools like Midjourney and Stable Diffusion. A head-to-head comparison using DALL-E 2 revealed some ways it fell short.

In July, AI art enthusiasts were able to experiment with OpenAI's in-progress upgrades to DALL-E 2. YouTube influencer MattVidPro demonstrated it outperforming its predecessor as well as other image generators.

"This blows anything we've seen before out of the water, it's insane,” he said at the time. “Midjourney cannot compete at this level—I don't even think Midjourney version six would be able to compete at this level.”

That technology preview lacked the protections put in place by major AI developers to prevent malicious or deceptive uses, allowing MatVidPro to demonstrate a fully uncensored implementation of DALL-E.

When DALL-E 3 was officially announced on Sept. 20, its improved capabilities were clear—with guardrails meanwhile put in place for the public. At the time, it appeared that paid customers of OpenAI's ChatGPT Plus and Enterprise products would get access first.


Late Friday, MattVidPro revealed that a public release of DALL-E 3 was imminent, noting that some users could get early access via Bing Image Creator by logging in and out of their Microsoft accounts, using more obscure web browsers, or clearing their browser cache.  As of this writing, however, such tricks are no longer required.

"We know that DALL-E 3 is capable of more in the ChatGPT [interface], but no one really seems to be getting access to that yet, but people are definitely getting access via Bing Image Creator," he said.

Although the Bing Image Creator interface doesn't disclose the version of DALL-E being used, MattVidPro shared test prompts that would reveal whether you were using the new version. Among them are the ability to render readable text rather than random gibberish, and the ability to depict historical figures and famous celebrities—although some noted that other tools may be purposefully limiting their capabilities on this front.

"This model is so powerful, it's hard to quantify," he said. "Dall-E 3 is absolutely insane—this is easily the best AI image generator ever."

"Hands down, every single generation that comes out of this thing is better than anything I've ever seen out of any other image generator before," he added.

Jose Antonio Lanz contributed to this report.

Generally Intelligent Newsletter

A weekly AI journey narrated by Gen, a generative AI model.