Microsoft Launches Three In-House AI Models

Microsoft has released three new foundational AI models built entirely in-house—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—marking the most concrete evidence yet that the company intends to compete directly with OpenAI and Google on model development, not just distribution. The models are available immediately through MicrosoftFoundry and a new MAI Playground, representing the first output from Microsoft's superintelligence team formed just six months ago under Mustafa Suleyman.

MAI-Transcribe-1 is the headline release, achieving the lowest average Word Error Rate on the industry-standard FLEURS benchmark across 25 languages. According to Microsoft's benchmarks, it beats OpenAI's Whisper-large-v3 on all 25 languages, Google's Gemini 3.1 Flash on 22 of 25, and delivers 2.5x faster batch transcription thanexisting Azure offerings. The company is already testing it inside Copilot Voice mode and Microsoft Teams for conversation transcription.

The launch comes at a pivotal moment for Microsoft. Until October 2025, the company was contractually prohibited from pursuing artificial general intelligence independently under its original OpenAI partnership terms. Therenegotiated agreement freed Microsoft to build its own frontier models while retaining license rights to OpenAI models through 2032. Suleyman emphasized the OpenAI partnership remains intact, but the models signal Microsoft's strategic push toward what he calls "AI self-sufficiency."

Read the full article at VentureBeat →