Microsoft Launches 3 New MAI AI Models on Foundry to Take on OpenAI, Google
Microsoft is positioning these models around three key metrics: speed, efficiency, and cost.
Microsoft has announced a new set of in-house AI models under its MAI (Microsoft AI) family, signalling a deeper push into building its own AI ecosystem rather than relying heavily on partners like OpenAI. The company has introduced three models, MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, all of which are now available through its Azure-based Microsoft Foundry platform.
Three models, three core capabilities
Each of the new models focuses on a specific use case.
— MAI-Transcribe-1 is designed for speech-to-text tasks and supports transcription across 25 major languages. Microsoft claims it delivers state-of-the-art accuracy while also being significantly faster than its previous offerings, with batch processing speeds up to 2.5 times higher.
— MAI-Voice-1 focuses on text-to-speech generation. It is built to produce natural, expressive audio with emotional nuance, while also preserving speaker identity across longer outputs. The model can even generate up to 60 seconds of audio in about a second, making it suitable for real-time applications and voice agents.
— MAI-Image-2 targets image generation and creative workflows. Microsoft says the model improves both speed and output quality, offering more accurate lighting, textures, and in-image text rendering. It is already being integrated into tools like Copilot, Bing, and PowerPoint.
Built for speed, cost, and scale
Microsoft is positioning these models around three key metrics: speed, efficiency, and cost. The company claims that all three models are designed to deliver competitive price-to-performance ratios compared to alternatives available in the market. Pricing starts as low as $0.36 per hour for transcription, with similar cost-focused positioning for voice and image generation.
The models are also optimised for real-world usage, meaning they are designed to perform reliably in less controlled environments rather than ideal lab conditions.
Available through Microsoft Foundry
All three models are accessible via Microsoft Foundry, the company’s AI development platform that allows developers to build, test, and deploy AI applications using a wide range of models. Foundry acts as a central hub where enterprises can access models from Microsoft as well as partners like OpenAI, Meta, and others, while also providing tools for deployment, governance, and scaling. The new MAI models are also available through the MAI Playground, where developers can experiment with them before integrating them into applications.
Published By : Shubham Verma
Published On: 3 April 2026 at 19:47 IST