Microsoft launches MAI-Image-2-Efficient, a new AI image model

Microsoft has unveiled MAI-Image-2-Efficient, a new version of its image generation model that boasts lower costs and higher speed. This launch occurred in Microsoft Foundry and MAI Playground without the need for a waitlist, highlighting the company's ambition to build a self-sufficient AI ecosystem that does not rely on OpenAI.

The new model is priced at $5 per million text tokens and $19.50 per million images, representing approximately a 41% reduction from its predecessor, MAI-Image-2. Microsoft claims that MAI-Image-2-Efficient operates 22% faster and achieves four times the throughput efficiency on GPUs, as demonstrated on NVIDIA H100 hardware.

MAI-Image-2-Efficient is aimed at high-volume production tasks such as product photography and marketing material creation. This model allows for effective handling of text within images and is designed to work within tight time and budget constraints.

Microsoft's strategy is to position MAI-Image-2-Efficient and MAI-Image-2 as complementary tools, reflecting an approach seen in other areas of AI. This enables businesses to choose between an efficient model for mass production and a more precise model for complex tasks.

The launch of MAI-Image-2-Efficient also underscores the shifting dynamics between Microsoft and OpenAI. A recent internal memo from OpenAI indicated that their partnership with Microsoft limits their market capabilities, signaling increasing competition and the necessity for Microsoft to develop its own AI solutions.