On Thursday, Microsoft AI, the technology giant’s research division, unveiled three new foundational AI models capable of generating text, voice, and images.
This launch indicates Microsoft’s ongoing effort to develop its own collection of multimodal AI models and compete with other AI research institutions, despite its continued alliance with OpenAI.
As detailed in a company press release, MAI-Transcribe-1 converts speech from 25 languages into text, operating 2.5 times quicker than Microsoft’s Azure Fast service. MAI-Voice-1, an audio generation model, enables users to produce 60 seconds of audio in a single second and offers the capability to create personalized voices. MAI-Image-2 is designed for video generation.
Initially launched on March 19 via MAI Playground, a new testing platform for large language models, MAI-Image-2 is now available alongside all three models on Microsoft Foundry. The transcription and voice models are also accessible through MAI Playground.
These models were created by Microsoft’s MAI Superintelligence team, an AI research group established and announced in November 2025, under the leadership of Mustafa Suleyman, CEO of Microsoft AI.
“At Microsoft AI, our focus is on developing Humanist AI. Our approach to creating AI models is unique: we prioritize humans, optimize for natural communication, and train for practical applications,” Suleyman stated in the blog post. He added, “Expect to see more models from us shortly in Foundry and integrated directly into Microsoft products and experiences.”
Amidst a growingly competitive large language model (LLM) market, MAI aims to position these models with a key advantage: their cost-effectiveness compared to offerings from Google and OpenAI, as noted in the company’s blog post.
The pricing begins at $0.36 per hour for MAI-Transcribe-1. MAI-Voice-1 is priced from $22 per 1 million characters, while MAI-Image-2 costs $5 for 1 million tokens of text input and $33 for 1 million tokens of image output.
Despite launching its proprietary models, Suleyman reiterated Microsoft’s dedication to its collaboration with OpenAI during an interview with VentureBeat. However, Suleyman also informed The Verge that a recent renegotiation of this partnership was crucial in enabling Microsoft to actively pursue this superintelligence research.
Through a multi-year agreement, Microsoft has allocated over $13 billion to the AI research lab and integrates its models across various products. Microsoft adopts a similar dual approach concerning chips, both manufacturing its own and procuring them from external suppliers.