Microsoft has launched its first internally developed AI models: MAI-Voice-1 – a voice generation model – and MAI-1-preview, a language model in the testing phase.

Microsoft has launched MAI-Voice-1, a speech model capable of generating one minute of audio in less than a second, using a single GPU. It is already integrated into features like Copilot Daily, providing AI-narrated news briefs. Users can experiment with MAI-Voice-1 in Copilot Labs. Additionally, MAI-1-preview, a text model based on 15,000 Nvidia H100 GPUs, is intended for managing daily instructions. These initiatives are consumer-focused, in contrast to enterprise-oriented AI partnerships, and represent just the beginning of Microsoft's ambitions in the AI field.