In August 2026, Microsoft is set to put its inaugural native real-time speech model, MAI Realtime, through its paces. This innovative model boasts support for 16 languages and 2 distinct speech styles. It facilitates seamless simultaneous listening and speaking during conversations, along with automatic language identification and the ability to switch languages mid-dialogue. Microsoft has extended invitations to select partners to trial the model on the MAI Playground platform. However, the timeline for its official launch on Microsoft Foundry or integration into Copilot's speech functionalities remains shrouded in mystery. MAI Realtime adopts a full-duplex interaction strategy, obviating the necessity for rigid turn-taking and enabling concurrent listening and response. The debugging panel is designed to showcase real-time latency, the model's cognitive process, and the sequential steps of processing.