Microsoft quietly rolled out a meaningful update to one of the internet’s most accessible image generators: Bing Image Creator now supports OpenAI’s GPT-4o image model alongside DALL·E 3, giving users a simple toggle between two different image-generation engines. The change is live on bing.com/create, the Bing mobile app, in Copilot Search, and even directly from the Bing search bar or the Edge address bar.

If you use Bing Image Creator, you now have two practical choices:

• GPT-4o — billed by Microsoft as a “major leap” in quality: more photorealism, better handling of fine details and text in images, and a deeper understanding of complex prompts. It’s now offered for free inside Bing Image Creator.
• DALL·E 3 — the older (and still-capable) OpenAI model that can generate multiple quick variations at once; Microsoft keeps this as the faster/more options route inside the same UI.

Practical quota: Bing Image Creator gives you 15 “fast” GPT-4o creations per day for free; after that, fast requests cost 10 Microsoft Rewards points each. Microsoft also says standard-speed generations are unlimited.

GPT-4o is part of OpenAI’s “omnimodal” push — a single model that can handle text, images, audio and (experimentally) video. Among the concrete image improvements OpenAI and reporters noted earlier this year are:

• Better “binding” — the model keeps relationships between multiple objects/attributes more reliably (e.g., “blue circle left of red triangle”).
• Improved text rendering on images — small labels and typography tend to be more readable and less garbled than with many earlier generators.
• The model uses an autoregressive approach (generating images sequentially) rather than the diffusion technique used by DALL·E, which designers and researchers say helps with certain detailed outputs.

Microsoft’s message leans into those strengths, pitching GPT-4o for photo-real product mockups, detailed visualizations, or any prompt that benefits from fidelity and nuance. That said, the company also points out you can switch back to DALL·E 3 when you want quicker batches and multiple variations.

A couple of practical tradeoffs matter for day-to-day use:

• Several writeups and hands-on observers report that GPT-4o in Bing currently generates one detailed image at a time, while DALL·E 3 can produce multiple variations faster. If you like to spray-and-pray with many quick options, DALL·E 3 may still be preferable.
• Microsoft’s free quota focuses on fast creations (15/day). If you regularly need many fast, high-fidelity outputs, you’ll either burn Rewards points or choose the standard-speed unlimited path, which Microsoft says remains free. The experience and latency for GPT-4o can be longer than DALL·E’s quick batch outputs.

In short: GPT-4o trades some throughput (and patience) for higher fidelity and better handling of tricky details like embedded text.

How to use it — where you’ll see it

Microsoft says the model switch is available:

• at bing.com/create,
• in the Bing mobile app,
• via Copilot Search, and
• directly from the Bing search bar or the Edge address bar (if Bing is your default).

It’s a UI toggle — pick GPT-4o when you want a higher-detail, photoreal finish; pick DALL·E 3 when you want multiple variants quickly. (Microsoft’s blog has screenshots of the mobile and desktop flows.)