Microsoft AI, the technology giant’s research facility, unveiled three core AI models on Thursday capable of generating text, voice, and images.

This launch underscores Microsoft’s ongoing effort to expand its arsenal of multimodal AI models to compete with other AI research entities, despite its ongoing association with OpenAI.

MAI-Transcribe-1 converts speech into text in 25 different languages and operates 2.5 times quicker than Microsoft’s Azure Fast service, based on a company press announcement. MAI-Voice-1 is designed for audio production. This voice model enables users to produce 60 seconds of audio in just one second and customize their own voice. MAI-Image-2 is a model for video generation.

MAI-Image-2 was first introduced on MAI Playground, a new platform for testing large language models, on March 19. Presently, all three models are being rolled out on Microsoft Foundry, with the transcription and voice models also accessible in MAI Playground.

The models were crafted by Microsoft’s MAI Superintelligence team, an AI research group under the leadership of Mustafa Suleyman, the CEO of Microsoft AI, established and announced in November 2025.

“At Microsoft AI, we’re developing Humanist AI. Our approach to creating AI models is distinct — prioritizing human needs, optimizing for real human communication, and training for practical applicability,” Suleyman expressed in the blog entry. “You can expect more models from us shortly in Foundry and integrated into Microsoft products and services.”

In an ever-competitive LLM market, MAI aims to differentiate these models by noting their lower costs compared to those offered by Google and OpenAI, as mentioned in the blog post.

MAI-Transcribe-1 is priced starting at $0.36 per hour. MAI-Voice-1 begins at $22 for each 1 million characters, while MAI-Image-2 starts at $5 for every 1 million tokens for text input and $33 for every 1 million tokens for image output.

Even while unveiling its own models, Suleyman reiterated Microsoft’s dedication to its collaboration with OpenAI in a discussion with VentureBeat — although a recent renegotiation allowed Microsoft to genuinely explore this superintelligence research, he informed The Verge.

Microsoft has poured over $13 billion into the AI research lab and integrates its models into various products through a long-term partnership. Microsoft adopts a similar strategy with chips, both manufacturing its own and procuring from external sources.