Microsoft’s new in-house image generator has already cracked the LMArena AI benchmark’s top 10 — challenging claims that it’s just an “OpenAI reseller”

by

Microsoft’s AI chief, Mustafa Suleyman, stated the company aims to develop its own advanced AI models internally, ranging in size, but will also realistically use models from other sources when appropriate. This reflects Microsoft’s goal of becoming independent in generative AI, including building its own chip infrastructure, and reducing its dependence on OpenAI’s technology.

Recently, the software company launched MAI-Image-1, its own text-to-image generator, created by its internal team (according to The Verge). The company described the model as a key advancement in their work.

Microsoft is creating its own AI models that will be a few months behind the cutting-edge technology from OpenAI, according to the company’s AI CEO, Mustafa Suleyman. They recently released two of these models – MAI-Voice-1 and MAI-1-preview – back in August.

Microsoft’s new AI image generator, like many others today, can create incredibly realistic pictures – everything from landscapes to dramatic lighting effects. Importantly, Microsoft worked with creative professionals during development to ensure the images aren’t cookie-cutter or overly generic.

What’s really impressive is that this AI model can create images from requests quicker than many larger, more complex models. It’s already proving its worth, currently ranking as a top 10 image generator on the LMArena benchmark.

The news arrives as the relationship between Microsoft and OpenAI appears to be weakening, stemming from disagreements about OpenAI becoming a for-profit company and issues with cloud computing. Earlier this year, OpenAI announced its $500 billion ‘Stargate’ project, which aims to build data centers throughout the U.S. to support its cutting-edge AI technology.

Microsoft is no longer the only cloud provider for OpenAI, although they still have the first chance to partner. Salesforce CEO Marc Benioff believes Microsoft won’t integrate OpenAI’s technology into its products going forward. The recent release of Microsoft’s own AI models and image generator suggests Benioff might be right.

I’ve been following the reactions to Microsoft’s AI work, and Benioff has been quite vocal with his criticism. He recently said, back in October 2024, that he thinks Microsoft’s approach to AI actually hurts the industry. He even compared their Copilot to the old Microsoft Clippy, implying it’s not very useful or effective.

The executive also stated that Microsoft essentially rebranded OpenAI’s technology and integrated it into Excel. Benioff added that customers are unhappy with Copilot, viewing it simply as a resale of OpenAI’s offerings.

OpenAI has expressed concerns that Microsoft isn’t providing enough cloud computing power to support its work. They’ve suggested that if another AI company reaches a major milestone – achieving artificial general intelligence (AGI) – it might be because Microsoft’s resources were insufficient. This is the reason OpenAI is looking for cloud computing options from other providers in addition to Microsoft.

Reports suggest Microsoft cancelled plans for two large data centers because they didn’t want to offer extra training resources for ChatGPT. However, OpenAI’s CEO, Sam Altman, appears to disagree, stating that OpenAI isn’t currently limited by computing power.

Up until now, Copilot’s image creation and overall functionality have largely relied on technology from OpenAI, including DALL-E 3 and GPT-4o.

Can Microsoft produce enough GPUs to handle all the demand for creating images with this new technology? Let me know what you think in the comments!

Stay up-to-date with the latest news, insights, and features from Windows Central by following us on Google News!

Read More

2025-10-14 15:10