Microsoft AI: MAI-Image-2.5

MAI-Image-2.5 is an image generation model from Microsoft, published under the Microsoft AI banner and identified on api14 as microsoft/mai-image-2.5. It sits in the image category, and Microsoft describes it as a high-quality image generation model available via Azure AI Foundry. If you are already working within Microsoft's ecosystem or evaluating it, this is the model that backs that offering.

The model takes text and image inputs and returns images. That means you can drive it with a prompt alone, or supply a single input reference image alongside your prompt to guide the result. Its context window is 4096 tokens, and the maximum output is 1024. Generation is controlled through a small set of image parameters: aspect_ratio, which accepts 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3 or auto; n, which is fixed at exactly 1, so each request returns a single image; and input_references, which accepts between 0 and 1 reference images. Microsoft positions the output as both photorealistic and artistic, so it is intended to cover a fairly wide stylistic range rather than one narrow look.

The practical uses follow directly from those capabilities. Text to image generation covers the usual ground: concept art, marketing visuals, illustrations, and asset drafts where you need a specific framing, which is where the aspect ratio options matter, since you can request square, landscape or portrait output without post-processing crops. Because the model accepts one input reference image, it also suits workflows where you want a new image guided by an existing one, such as producing a variation on a composition or carrying a visual style from a reference into a new scene. The single image per request constraint means batch work is a matter of issuing multiple calls rather than asking for several images at once.

Calling it through api14 works the same way as any other model on the gateway, using the OpenAI-compatible API with the model id microsoft/mai-image-2.5. The supported request parameters are max_completion_tokens, max_tokens and temperature, alongside the image parameters described above for aspect ratio and input references. If you have existing OpenAI client code, you change the model id and the base URL, and the request shape stays familiar.

A few limits are worth knowing before you commit. The context window of 4096 tokens is modest, so very long or heavily detailed prompts have less room than on larger-context models. Output is capped at 1024, n cannot exceed 1, and only one input reference is accepted, so tasks needing multiple reference images or several images per call will not fit here. If your work is text rather than image generation, or you need a much larger context, a different model on api14 will serve you better.