Microsoft Rolls Out MAI Models, Cuts Contact-Center GPU Costs 89%

Microsoft has relied heavily on its OpenAI partnership to supply frontier models for Copilot and Azure, but it is steadily expanding its internally developed MAI lineup. Building more models in-house gives the company greater control over inference costs, performance and integration across products including Bing, PowerPoint and Dynamics 365. Offering the technology through Microsoft Foundry also turns its model-training expertise into enterprise cloud services while reducing dependence on a single outside supplier.

Microsoft on July 23, 2026, placed MAI-Image-2.5-Pro and MAI-Voice-2-Flash in public preview. The voice model, deployed in Dynamics 365 Contact Center and available through Azure Voice Live, cuts GPU costs by as much as 89%. Microsoft said it is twice as fast as MAI-Voice-2 and 32% cheaper, with preview pricing of $15 per 1 million characters. MAI-Image-2.5-Pro costs $5 per 1 million text-input tokens, $8 per 1 million image-input tokens and $106 per 1 million image-output tokens.