For the first time, Microsoft has launched its own AI models: MAI-Voice-1 and MAI-1-preview. This marks a new direction in the company’s AI strategy. MAI-Voice-1 can generate one minute of speech in under a second using just a single graphics processor (GPU). It’s already being used in features like Copilot Daily, where it reads the news or creates podcast-style conversations.
MAI-1-preview, trained on 15,000 Nvidia H100 GPUs, is designed to understand instructions and provide helpful answers. Unlike OpenAI’s more business-focused models, Microsoft is targeting everyday users. Both models are being integrated into Copilot and tested on platforms like LMArena. Further development is planned.