Microsoft Launches Three In-House AI Models to Challenge OpenAI and Google

Quick Facts

• Microsoft launched three new AI models built entirely in-house: speech transcription, voice generation, and image creation
• MAI-Transcribe-1 achieves 3.9% average Word Error Rate across 25 languages, claiming world’s most accurate transcription
• Models run on half the GPU footprint of competitors and are priced below major cloud rivals

Microsoft on Thursday launched three foundational AI models built entirely in-house, marking the company’s most direct challenge to OpenAI and Google in model development.

The three models target core AI capabilities. MAI-Transcribe-1 delivers speech-to-text transcription with a 3.9% average Word Error Rate across 25 languages. Microsoft claims it outperforms OpenAI’s Whisper-large-V3 and Google’s Gemini 3.1 Flash-Lite on benchmark tests.

MAI-Voice-1 generates 60 seconds of audio in under one second on a single GPU. The text-to-speech model powers custom voice creation and handles expressive audio generation.

MAI-Image-2 ranks third on the Arena.ai image generation leaderboard. The text-to-image model is already integrated into Bing and PowerPoint.

Microsoft priced the models aggressively to undercut competitors. MAI-Transcribe-1 costs $0.36 per audio hour and runs 2.5 times faster than Microsoft’s previous Azure Fast offering. MAI-Voice-1 is priced at $22 per million characters, while MAI-Image-2 costs $5 per million text input tokens.

“We’re pricing them to be the very best of any hyperscaler,” Microsoft AI CEO Mustafa Suleyman told VentureBeat. “So there will be the cheapest of any of the hyperscalers out there.”

Each model was built by teams of fewer than 10 engineers. The models run on half the GPU footprint of comparable competitors, reducing Microsoft’s infrastructure costs across Teams, Copilot, Bing, and PowerPoint.

The launch follows a renegotiated contract with OpenAI that previously prohibited Microsoft from independently pursuing artificial general intelligence until October 2025. Suleyman said Microsoft’s goal is to be “completely independent” across all model modalities.

Microsoft’s stock has fallen 17% year-to-date as investors worry about the company’s dependence on OpenAI. During recent earnings, analysts noted OpenAI represents 45% of Microsoft’s future sales backlog. Microsoft shares lost $357 billion in market value following those concerns.

The models are available immediately through Microsoft Foundry and a new MAI Playground. Microsoft has committed over $13 billion toward AI development and partnerships.

“We’re now a top three lab just under OpenAI and Gemini,” Suleyman said, positioning Microsoft as a direct competitor rather than just a distributor of AI capabilities.

Read more: Microsoft launches 3 new AI models in direct shot at OpenAI and Google