Microsoft Launches MAI-Voice-1 and MAI-1-Preview, Its First In-House AI Models

Microsoft Launches MAI-Voice-1 and MAI-1-Preview, Its First In-House AI Models

Microsoft’s AI division has introduced its first homegrown AI models: MAI-Voice-1 and MAI-1-preview. The launch marks a major milestone as the company begins building its own consumer-focused AI systems to power Copilot and other tools.

What is Microsoft MAI-Voice-1?

MAI-Voice-1 is Microsoft’s new speech AI model that can generate a full minute of audio in less than a second using just one GPU. This makes it extremely efficient compared to traditional speech models.

The model already powers features like Copilot Daily, which delivers the day’s top news stories in an AI voice, and generates podcast-style discussions that explain trending topics.

Users can also try it themselves on Copilot Labs, where they can input custom text, adjust the voice style, and even change speaking tones.

Introducing Microsoft MAI-1-Preview

Alongside MAI-Voice-1, Microsoft also launched MAI-1-preview, a powerful text-based model trained on around 15,000 Nvidia H100 GPUs.

The company describes it as an instruction-following AI, designed to provide helpful responses to everyday queries. Unlike Microsoft’s enterprise-focused AI collaborations with OpenAI, MAI-1-preview is built with consumers in mind.

Why Microsoft Is Building Its Own AI Models

According to Microsoft AI chief Mustafa Suleyman, the focus is on consumer-first AI experiences. Instead of competing directly with enterprise-grade models, Microsoft wants to optimize AI for daily use cases such as search, productivity, and personal assistance.

Suleyman emphasized that Microsoft has access to massive amounts of consumer data and telemetry, which can help fine-tune models for more predictive and useful outputs.

What’s Next for Microsoft AI?

Microsoft plans to integrate MAI-1-preview into Copilot, which currently relies on OpenAI’s models. The company has also started testing it publicly on the AI benchmarking platform LMArena.

In its announcement, Microsoft said it envisions a multi-model approach, with different specialized AI systems working together to handle a wide range of user needs.

“We have big ambitions for where we go next,” Microsoft wrote. “Orchestrating specialized models across user intents will unlock immense value.”

With MAI-Voice-1 and MAI-1-preview, Microsoft is taking its first step into building a more independent AI ecosystem—tailored for everyday users.

TechCity