Microsoft Privately Launches Two AI Models Amidst Vigorous AI Expansion, Employment Drive
Microsoft has taken a significant step forward in artificial intelligence (AI) development, announcing the release of two new AI models under its Microsoft AI (MAI) umbrella. The first model, MAI-Voice-1, and the second, MAI-1-Preview, are set to make a splash in the AI landscape.
MAI-Voice-1: Powering Microsoft Copilot Daily and Podcasts
MAI-Voice-1, a natural speech generation model, is already in use, powering the Microsoft Copilot Daily and Podcasts features. This model generates high-fidelity audio for single and multi-speaker conversations, demonstrating Microsoft's commitment to delivering top-notch AI experiences.
MAI-1-Preview: A Consumer-Focused Model
MAI-1-Preview, the first Microsoft foundation model trained end-to-end, is designed with consumers in mind. This model specializes in following instructions and providing helpful responses to everyday queries, making it an ideal choice for those seeking AI models that cater to their daily needs.
Public Testing and Rollout
The second model, MAI-1-Preview, is currently being publicly tested on LMArena. Over the coming weeks, it will be rolled out for text-based use cases within Copilot.
AI Agent & Copilot Summit
Microsoft Copilot and agents will be the focus of the upcoming AI Agent & Copilot Summit, taking place in San Diego from March 17-19, 2026. This AI-first event promises to provide more details about these groundbreaking models and the future of AI at Microsoft.
Copilot Labs: A Hub for Experimental AI Initiatives
In Copilot Labs, customers can try expressive speech and storytelling demos, offering a unique opportunity to experience the capabilities of these AI models firsthand.
A Push for Developers and Improving AI
Microsoft is actively hiring for software engineering roles as part of its push for developers. The goal is to improve from user feedback and evaluate millions of unique interactions to deliver optimal outcomes.
Computing Power and GPU Efficiency
The company's compute roadmap is supported by its operational NVIDIA GB200 cluster, and MAI-1-Preview, in particular, is an in-house mixture-of-experts model pre-trained and post-trained on approximately 15,000 NVIDIA GPUs. Impressively, it can create a full minute of audio in under one second on a single GPU.
The Search for Names and Future Developments
While the search results do not provide the names of the Microsoft employees involved in developing the AI models MAI-Voice-1 and MAI-1-Preview, the future of AI at Microsoft continues to unfold. As these models are further refined and rolled out, we can expect to see even more innovative applications of AI technology.
For more details about the AI Agent & Copilot Summit, visit the event's official website. Stay tuned for further updates on these exciting new AI models from Microsoft.
