Microsoft AI announced on August 29, 2025, the start of public testing for its first two major in-house AI models, confirming its strategy to diversify its AI resources. The first model, MAI-1-preview, is a foundational language model created for a wide range of consumer scenarios. It will operate in parallel with models from OpenAI, complementing them and allowing Microsoft to optimize performance and costs in its products like Copilot and Windows. The second model, MAI-Voice-1, is a specialized speech generation model. Its key advantage is high speed and low latency, which will provide a more natural and seamless real-time voice interaction experience. MAI-Voice-1 will be integrated into Copilot, Daily, and other Microsoft services. This launch is a landmark event as it demonstrates that Microsoft, while remaining a key partner of OpenAI, is also actively developing its own powerful R&D to create foundational models tailored to its ecosystem.
Microsoft Launches Public Testing of In-House AI Models MAI-1 and MAI-Voice-1
