Microsoft’s MAI-Image-2-Efficient model accelerates the company’s move away from OpenAI

Microsoft Corp. pressure. of artificial intelligence is gaining momentum with today’s release of MAI-Image-2-Efficient, a leaner and gentler version of its advanced image generation model that debuted earlier this month.

The new version is intended to deliver high-quality visuals faster, and at a fraction of the cost of its predecessor, the company said. The original MAI-Image-2 model was released by Microsoft’s new superintelligence MAI team earlier this month, and a team led by Mustafa Suleyman is credited with launching a version of that model a few weeks later.

As the name suggests, MAI-Image-2-Efficient is all about efficiency, with Microsoft promising up to four times faster per image processing unit when running on Nvidia Corp.’s H100 processors. In terms of raw performance, it is about 22% faster than the main MAI-Image-2 model, surpassing Google LLC’s Gemini 3.1 Flash with 40% higher latency, the company said, citing its benchmarks.

Microsoft said MAI-Image-2-Efficient is more cost-effective as well, with prices starting at $5 per million input tokens, the same as the flagship model, and $19.50 per million output tokens, 41% cheaper. This two-tier pricing strategy, with users able to choose between a high-end model for reliable creative work and an efficient version for volume production, mirrors the playbooks of Google and startups like OpenAI Group PBC and Anthropic PBC. The company said MAI-Image-2-Efficient is particularly suited to “assembly line” tasks such as user mockups, product images and marketing materials, where cost-per-image is a key consideration.

Early response to the original MAI-Image-2 has been positive, with the model claiming the number 3 spot on Arena.ai’s leaderboard and earning praise for its realistic image and ability to render complex typography – an area where many other image production models struggle. But it is not yet clear whether MAI-Image-2-Efficient suffers from the same issues as its sibling. For example, the original received heavy criticism for its 30-second “cooling” time between prompts, its inability to produce images with anything other than a 1:1 aspect ratio, and its overly aggressive content filtering.

From OpenAI

Microsoft’s intention to improve its models is driven by its desire to reduce its reliance on OpenAI, in which it has invested heavily over the years. There was a time when the Windows software maker was almost entirely dependent on OpenAI models for its AI products and capabilities. But it seems that it is wary of being “locked” in a company that is growing strongly and has shown a desire to seek independence.

Just yesterday, CNBC reported that it uncovered an internal memo from OpenAI’s new Chief Revenue Officer, Denise Dresser, who told employees that the company’s partnership with Microsoft “reduced our ability to meet businesses where they are.” The memo pointed to OpenAI’s new partnership with Amazon Web Services Inc.’s Bedrock platform as a key driver of growth, noting that internal demand was “amazing, amazing” since the partnership was announced in February.

OpenAI also looked to diversify its cloud infrastructure base from providers such as CoreWeave Inc., Google LLC and Oracle Corp. at Microsoft’s expense. On the other hand, Microsoft has added OpenAI to its official list of competitors in mid-2024.

Meanwhile, Valoir analyst Rebecca Wettemann told SiliconANGLE in January that Microsoft’s reliance on OpenAI is becoming a major concern for investors, as OpenAI faces increased competition from rivals such as Google and Anthropic. “Investors are getting impatient, especially because most of Microsoft’s returns are related to the money from OpenAI that is now widely considered,” he explained.

Building its own models would be a huge benefit for Microsoft. MAI models are now the default choice whenever someone asks the company’s Copy tool to render an image, instead of OpenAI’s DALL-E model. That means Microsoft can keep all the costs it incurred, rather than paying a license fee to OpenAI.

MAI-Image-2-Efficient also helps accelerate Microsoft’s AI strategy. With the launch of services like Copilot Cowork and Agent 365, Microsoft is looking towards a world where AI performs complex, multi-step tasks and workflows on behalf of users. If a freelancer is asked to put together the materials for a new marketing campaign, low latency and reduced costs are essential, enabling it to scale.

Microsoft said MAI-Image-2-Efficient is now available through Microsoft Foundry and MAI Playground.

Image: Microsoft