Microsoft AI launches its first streaming transcription model, MAI-Transcribe-2-Streaming, plus MAI-Voice-2.1

Microsoft AI, the company's in-house model group, announced on 1 October its first streaming transcription model, MAI-Transcribe-2-Streaming, together with two speech-generation models, MAI-Voice-2.1 and MAI-Voice-2.1-Flash.

Microsoft says the transcription model “delivers low-latency, real-time transcripts in 60 languages”, detects the language automatically, and produces its first partial transcript “in just over 100ms of receiving audio.” The company claims it ranks first on Artificial Analysis's leaderboard for transcription. “MAI-Transcribe-2-Streaming is available at an introductory price of $0.54 per hour of audio through the end of the year,” Microsoft says.

On the voice side, MAI-Voice-2.1 covers 23 languages and 26 locales and can clone a voice from a few seconds of reference audio; it costs $22 per million characters. The Flash version has 150ms end-to-end latency and costs $15 per million characters.

All three models are available in Microsoft Foundry and the MAI Playground, and the voice models are also on OpenRouter. Microsoft lists Vercel and Azure Voice Live too.

The release puts Microsoft's own models into the market for real-time voice agents, where it competes with specialist speech providers and with OpenAI, its long-time partner. The leaderboard ranking is Microsoft's description of a third-party benchmark and was not checked here. A separate Microsoft AI story, the end of pre-training for MAI-Base-2, ran on 17 October.

• Confirmed MAI-Transcribe-2-Streaming transcribes live audio in 60 languages at an introductory $0.54 per hour through the end of the year. Microsoft AI
• Claimed Microsoft says the model debuts at no. 1 on Artificial Analysis and returns first partials in just over 100ms. Microsoft AI
• Confirmed MAI-Voice-2.1 costs $22 per million characters; MAI-Voice-2.1-Flash costs $15, with 150ms end-to-end latency. Microsoft AI