Microsoft debuts MAI-Image-2-Efficient
Microsoft announced MAI‑Image‑2‑Efficient, a production‑ready image model that it says is faster, more efficient and cheaper to run, and is now available in Microsoft Foundry and MAI Playground. The company provided specific percentage improvements for speed, efficiency, cost and latency in its announcement. (x.com)
Microsoft has released MAI-Image-2-Efficient, a new image generator it says is built to make large batches of pictures faster and cheaper. (microsoft.ai) The model went live on April 14, 2026 in Microsoft Foundry and MAI Playground. Microsoft said it runs 22% faster than MAI-Image-2, uses 4 times less compute when normalized for latency and graphics processor use, and cuts output pricing to $19.50 per 1 million image tokens from $33 for MAI-Image-2. (microsoft.ai; techcommunity.microsoft.com; techcommunity.microsoft.com) Text-to-image systems turn written prompts into pictures by starting with visual noise and refining it step by step. Microsoft’s documentation says MAI-Image-2 uses that diffusion process to align an image with a text prompt. (learn.microsoft.com) Microsoft is pitching the efficient version for jobs where speed matters more than squeezing out every last detail. Its examples include product shots, marketing assets, user interface mockups, batch pipelines, and images generated mid-conversation inside chatbots or design tools. (microsoft.ai; techcommunity.microsoft.com) The release lands less than a month after Microsoft introduced MAI-Image-2 on March 19 and less than two weeks after it brought that model into Foundry on April 2. Microsoft said the original model was aimed at higher-fidelity work such as portraits, photorealistic scenes, stylized illustration, and longer text inside images. (microsoft.ai; techcommunity.microsoft.com; microsoft.ai) That gives Microsoft a two-model lineup inside its own image stack: one model for premium output and one for throughput. The company said both models are part of a broader push to make Foundry a full platform for developers building with Microsoft’s first-party voice, transcription, and image systems. (techcommunity.microsoft.com; techcommunity.microsoft.com) Microsoft also tied the launch to its own products. The company said MAI-Image-2-Efficient is rolling out across Copilot and Bing, with PowerPoint listed as a future surface, while the earlier MAI-Image-2 announcement said that model was already beginning to roll out in Copilot and Bing Image Creator. (microsoft.ai; microsoft.ai) Microsoft’s headline comparison with rivals is speed. In its launch post, the company said MAI-Image-2-Efficient was 40% faster on average than other leading text-to-image models in tests run on April 13, 2026, though the benchmark details and model settings were set by Microsoft. (microsoft.ai) For customers, the immediate change is practical: the new model is already available in Foundry and Playground, and Microsoft is selling it at the same $5 per 1 million text input tokens as MAI-Image-2 while lowering the image output price. (microsoft.ai; techcommunity.microsoft.com)
