Microsoft unveils MAI-Image-2.5-Pro and MAI-Voice-2-Flash AI models for image generation and speech recognition

Microsoft has launched public previews of two new internally developed AI models. MAI-Image-2.5-Pro is the flagship image generator, offering high-quality generation, detailed editing, and accurate text rendering. MAI-Voice-2-Flash is optimized for high-throughput voice applications like call centers, running twice as fast and costing 32% less than the base model. The company reports that Bing Image Creator now fully runs on MAI-Image-2.5, and PowerPoint has reduced GPU resource costs by 84% compared to using OpenAI's GPT-Image-2. MAI-Voice-2-Flash deployed in Dynamics 365 Contact Center cut GPU costs by up to 89% and is also integrated into Azure Voice Live. Microsoft's Dragon Copilot, used by 170,000 healthcare organizations, was migrated to MAI-Transcribe-1.5 supporting 58 languages. The lightweight MAI-Code-1-Flash model on GitHub Copilot shows 10% better code acceptance than GPT-5.4 Mini and Claude Haiku 4.5, with lower token consumption, and was further trained for Excel tasks where it performs at GPT-5.6 level while running on older Nvidia H100 and A100 accelerators.