Microsoft AI has unveiled three new foundational models designed to generate text, audio, and images, marking a strategic move to build a proprietary stack of multimodal technologies. Announced on Thursday, the release demonstrates the company's intent to compete directly with rival AI labs while maintaining its existing partnership with OpenAI. The newly launched suite includes MAI-Transcribe-1, a speech-to-text model capable of transcribing speech across 25 languages. According to Microsoft, this model operates 2.5 times faster than the company's previous Azure Fast offering. The second model, MAI-Voice-1, is an audio-generation tool that can produce 60 seconds of high-quality audio in just one second. It also features the capability to create custom voices. The third model, MAI-Image-2, is a video-generation tool. Previously tested in March on a platform called MAI Playground, all three models are now officially available on Microsoft Foundry, with the transcription and voice models also accessible via MAI Playground. These tools were developed by the MAI Superintelligence team, an AI research unit led by Mustafa Suleyman, the CEO of Microsoft AI. Suleyman stated that the team was formed in November 2025 with a focus on Humanist AI, prioritizing human-centric design and practical communication needs. He emphasized that the models are optimized for real-world application and predicted more releases in Foundry and Microsoft products in the near future. In a market crowded with large language models, Microsoft is highlighting cost efficiency as a key differentiator. The pricing structure for the new models is significantly lower than comparable offerings from Google and OpenAI. MAI-Transcribe-1 starts at $0.36 per hour of audio. MAI-Voice-1 is priced at $22 per one million characters, while MAI-Image-2 charges $5 per one million tokens for text input and $33 for image output. Despite introducing these in-house solutions, Suleyman reaffirmed Microsoft's commitment to its partnership with OpenAI, noting that a recent renegotiation of the agreement allowed the company the flexibility to pursue its own superintelligence research without conflicting with the existing collaboration. Microsoft has invested more than $13 billion in its AI research initiatives and continues to host these models across its product ecosystem through a long-term multi-year agreement. The company maintains a diversified strategy, producing its own hardware chips alongside purchasing from external providers, mirroring its approach to AI model development.