Microsoft announced three new foundational AI models on Thursday that generate text, voice, and images, signaling the tech giant's push to build its own multimodal AI stack while competing directly with OpenAI and Google on pricing. The models, developed by Microsoft AI's Superintelligence team, are positioned as cheaper alternatives to competitors' offerings, marking a significant shift in how Microsoft approaches AI research and commercialization .

What Are the Three New Models Microsoft Released?

Microsoft unveiled three distinct models designed to handle different types of AI tasks. MAI-Transcribe-1 converts speech into text across 25 languages and operates 2.5 times faster than Microsoft's previous Azure Fast offering. MAI-Voice-1 generates audio, allowing users to create 60 seconds of speech in just one second and customize voice characteristics. MAI-Image-2, a video-generating model originally released on Microsoft's MAI Playground testing platform in March, now becomes available across Microsoft's broader product ecosystem .

All three models are now accessible through Microsoft Foundry, the company's model deployment platform. The transcription and voice models are also available in MAI Playground, Microsoft's new large language model testing software, giving developers immediate access to experiment with the technology .

How Do Microsoft's Prices Compare to Competitors?

Pricing represents the clearest competitive advantage Microsoft is leveraging. MAI-Transcribe-1 starts at $0.36 per hour of transcription, while MAI-Voice-1 begins at $22 per million characters generated. MAI-Image-2 pricing starts at $5 per million tokens for text input and $33 per million tokens for image output . Microsoft explicitly highlighted in its announcement that these models are cheaper than comparable offerings from Google and OpenAI, a direct challenge to the market leaders.

For context, this pricing strategy matters because organizations deploying AI at scale face substantial costs. A company transcribing thousands of hours of audio annually, for example, could see meaningful savings with MAI-Transcribe-1's aggressive pricing. The move suggests Microsoft is willing to compete on cost rather than solely on performance metrics.

Steps to Access and Deploy Microsoft's New AI Models

• For Developers: Access MAI Playground to test all three models before committing to production deployment, allowing teams to evaluate performance on their specific use cases.
• For Production Use: Deploy models through Microsoft Foundry, which integrates with existing Microsoft product infrastructure and enterprise systems.
• For Voice and Transcription: Both MAI-Voice-1 and MAI-Transcribe-1 are available in MAI Playground, enabling immediate experimentation without enterprise licensing.

What Does This Mean for Microsoft's OpenAI Partnership?

Microsoft's move might seem contradictory given its $13 billion investment in OpenAI and multi-year partnership to integrate OpenAI models into Microsoft products. However, the company has clarified its strategy. Mustafa Suleyman, CEO of Microsoft AI, reaffirmed the partnership's importance while noting that a recent renegotiation allowed Microsoft to pursue its own superintelligence research independently .

"At Microsoft AI, we're building Humanist AI. We have a distinct view when creating our AI models, putting humans at the center, optimizing for how people actually communicate, training for practical use," stated Mustafa Suleyman, CEO of Microsoft AI.

Mustafa Suleyman, CEO of Microsoft AI

This approach mirrors Microsoft's strategy with semiconductor chips, where the company both manufactures its own processors and purchases from external suppliers. The dual strategy allows Microsoft to reduce dependency on any single partner while maintaining strategic relationships. The renegotiated OpenAI partnership apparently gave Microsoft more freedom to develop competing models, suggesting both companies benefit from this arrangement .

Why Is Microsoft Building Its Own Models Now?

The three models come from Microsoft AI's Superintelligence team, formed and announced in November 2025 under Suleyman's leadership. This dedicated research group signals Microsoft's commitment to competing in foundational AI model development, not just integrating third-party models into products. The timing reflects broader industry trends where major tech companies recognize that controlling foundational models provides strategic advantages in pricing, customization, and product differentiation .

In an increasingly crowded large language model market, where dozens of companies offer similar capabilities, Microsoft's emphasis on cost efficiency and practical usability suggests the company believes price and real-world applicability matter more than raw performance benchmarks. This positioning could appeal to enterprise customers balancing AI adoption with budget constraints.

The announcement indicates Microsoft plans to release additional models soon through Foundry and integrate them directly into Microsoft products and services, suggesting this is the beginning of a broader product rollout rather than a one-time release .