Microsoft on Thursday announced its first in-house AI models called MAI-Voice-1 and MAI-1 Preview. Notably, Microsoft has largely relied on OpenAI's GPT models to power the AI experience on Copilot, but as the tech giant forays into its own AI models, it increasingly stands in competition with the ChatGPT maker, along with a host of other AI companies.
What can Microsoft's new AI models do?
MAI-Voice-1 is a natural speech generation model, which the company claims can generate a full minute of audio in under a second and using only a single GPU, which the company says makes it one of the most efficient speech models out there.
The new model is now powering the Copilot Daily and Podcast feature, and Microsoft is also launching it in Copilot Labs.
Meanwhile, MAI-1 Preview is a foundational model from Microsoft and is now available for public tests on LMArena, a crowdsourced benchmarking platform for large language models (LLMs). MAI-1 Preview currently ranks in the 13th position in the LMArena leaderboard, way below the likes of GPT-5, Gemini 2.5 Pro, and even DeepSeek R-1 and Grok-3.
MAI-1 Preview has been pre-trained and post-trained by Microsoft on 15,000 NVIDIA H100 GPUs, which is a bargain compared to the over 100,000 employed by rivals like Grok for training its models.
Microsoft's AI roadmap:
In an interaction with Semafor, Microsoft AI Chief Mustafa Suleyman said that the company took some learnings from the open-source community in order to stretch the capabilities of its model with minimal resources.
“Increasingly, the art and craft of training models is selecting the perfect data and not wasting any of your flops on unnecessary tokens that didn’t actually teach your model very much,” Suleyman said.
Suleyman also said that Microsoft AI is already working on the next version of its AI models and is working on the largest data centers in the world equipped with Nvidia's latest GB200s chips.
“We have big ambitions for where we go next. Not only will we pursue further advances here, but we believe that orchestrating a range of specialized models serving different user intents and use cases will unlock immense value. There will be a lot more to come from this team on both fronts in the near future,” Microsoft wrote in a blog post.
