• Source:JND

Microsoft, which has so far leaned heavily on OpenAI’s models for its AI products, is now beginning to showcase its own work. Under the leadership of Mustafa Suleyman, the company’s AI division has introduced two new models: MAI-1-preview and MAI-Voice-1.

According to a company blog post, MAI-1-preview “offers a glimpse of future offerings inside Copilot,” while MAI-Voice-1 is capable of generating a 60-second audio clip in just one second using a single GPU. Microsoft claims this makes it one of the most efficient speech-generation systems available today.

ALSO READ: iPhone 17 Pro Max Price In India Revealed Through Leaks! This Is How Much It Might Cost You

The company describes MAI-Voice-1 as a “highly expressive and natural speech generation model”, which is already being used in Copilot Daily and Podcasts. Plans are also in place to bring it to Copilot Labs, where users will be able to test it with storytelling demos.

Meanwhile, MAI-1-preview is a large language model currently accessible on LMArena. It was trained on about 15,000 NVIDIA H100 GPUs and is designed to excel “in following instructions and providing helpful responses to everyday queries.” Microsoft says the model will soon appear in select text-based use cases. For context, the training setup was far smaller than that of xAI’s Grok, which reportedly required over 100,000 of these chips.

Suleyman had earlier made Microsoft’s consumer-first approach clear, telling The Verge that the in-house models won’t be focused on enterprise use cases but will instead “create something that works extremely well for the consumer and really optimise for our use case.”

ALSO READ: Samsung Galaxy A17 Series Launched: Check Price In India, Features And Specs

This comes against a backdrop of tension between Microsoft and OpenAI, despite the billions Microsoft has invested in the ChatGPT maker. In June, reports surfaced that OpenAI executives had even considered accusing Microsoft of anticompetitive practices within their partnership.