Microsoft takes on Google and OpenAI with its own AI models
Digital Trends • • technology
Key Points:
- Microsoft has publicly released three proprietary AI models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—available via the Microsoft Foundry platform and MAI Playground, targeting speech-to-text, voice synthesis, and image generation use cases.
- MAI-Transcribe-1 supports transcription in 25 languages and operates 2.5 times faster than Microsoft's Azure Fast service, while MAI-Voice-1 can generate natural-sounding audio rapidly and create custom voices from short clips; MAI-Image-2 ranks in the top three on the Arena.ai image generation leaderboard.
- Until October 2025, Microsoft was contractually restricted from developing its own frontier AI models due to a 2019 deal with OpenAI, which licensed OpenAI’s models to Microsoft in exchange for cloud infrastructure support.
- Despite launching its own AI models, Microsoft maintains its partnership with OpenAI, with CEO Mustafa Suleyman emphasizing continued collaboration even as Microsoft pursues a parallel AI strategy.
- The MAI models are competitively priced below similar offerings from Amazon and Google and have the potential to become foundational to Microsoft’s AI product ecosystem if they perform well.
