Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash

After unveiling MAI-Image-2.5 back in May, Microsoft has now announced MAI-Image-2.5-Pro and MAI-Voice-2-Flash, two new models designed to power everything from image generation to enterprise voice experiences. The company says these models already power millions of experiences across Bing, PowerPoint, OneDrive, Dynamics 365, and Azure.

One model focuses on image quality, the other on speed

The first newcomer is MAI-Image-2.5-Pro, Microsoft’s highest-quality image generation model so far. It focuses on producing detailed visuals, accurate in-image text, and more reliable image editing using natural language prompts.

Alongside it comes MAI-Voice-2-Flash, a lightweight voice model built for low-latency conversations. Microsoft says it delivers the same natural speech quality as MAI-Voice-2 while running twice as fast and costing 32% less.

Both models are now available in public preview, giving developers more flexibility to choose between quality, speed, and cost depending on their workloads.

Microsoft is already using them across its products

Microsoft says Bing Image Creator now runs entirely on MAI-Image-2.5, replacing external image models with its own technology. PowerPoint is also using MAI-Image-2.5 for image editing, reducing GPU costs by up to 84% compared to GPT-Image-2. Over in OneDrive, Microsoft says the model has improved save rates by 26%, reduced latency by roughly 25%, and increased efficiency by 2.5x.

On the voice side, MAI-Voice-2-Flash now powers Dynamics 365 Contact Center and Azure Voice Live, helping businesses build more responsive AI-powered voice agents while cutting GPU costs by as much as 89%.

Microsoft also confirmed that Dragon Copilot now uses MAI-Transcribe-1.5 across 58 languages, significantly reducing transcription and language-identification errors for healthcare professionals.