The Frontier
Sign inJoin free
Agents & software

Microsoft ships three speech models on October 1

Microsoft ships three speech models on October 1 - featured image
Original source, captured October 2026
1 min readfrontier

Share

Microsoft ships three speech models on October 1

On October 1, 2026 Microsoft shipped three speech models: MAI-Transcribe-2-Streaming, MAI-Voice-2.1, and MAI-Voice-2.1-Flash

Microsoft shipped three new speech models

Microsoft launched three speech models on October 1, 2026. MAI-Transcribe-2-Streaming handles real-time transcription, while MAI-Voice-2.1 and MAI-Voice-2.1-Flash handle text-to-speech

The models are available through Microsoft Azure AI services. They represent Microsoft's first streaming transcription model

Who gets Microsoft's speech models

The speech models are available to Azure AI customers. Developers can access them through the Azure AI Speech service

Microsoft did not announce specific pricing for the new models. They are available to existing Azure AI subscribers

What is actually new about Microsoft's speech models

MAI-Transcribe-2-Streaming is Microsoft's first streaming transcription model. It processes audio in real time rather than waiting for the full recording

The three-model suite covers both transcription and speech synthesis, giving developers a complete speech pipeline

What the launch does not settle

Microsoft has not published detailed benchmarks comparing the new speech models to competitors like OpenAI's Whisper or Google's speech models

No independent evaluation of the speech models' accuracy or latency has been published as of October 9, 2026

What to do now

Developers building speech applications should evaluate Microsoft's new models for real-time transcription and text-to-speech needs

Organizations already using Azure AI Speech should test the new models for performance improvements

Microsoft AI -

Share

Cite this piece

Canonical URL

https://www.thefrontier.dev/articles/microsoft-mai-speech-models-october-2026

Attribution

The Frontier, “Microsoft ships three speech models on October 1”, 9 Oct 2026

TLDR

On October 1, 2026 Microsoft shipped three speech models: MAI-Transcribe-2-Streaming, MAI-Voice-2.

Plain text

The Frontier. “Microsoft ships three speech models on October 1.” The Frontier. 9 Oct 2026. https://www.thefrontier.dev/articles/microsoft-mai-speech-models-october-2026

BibTeX

@misc{frontier_microsoft_mai_speech_models_october_2026_2026,
  title = {Microsoft ships three speech models on October 1},
  author = {{The Frontier}},
  howpublished = {The Frontier},
  year = {2026},
  month = oct,
  url = {https://www.thefrontier.dev/articles/microsoft-mai-speech-models-october-2026}
}

Full text may be reprinted with canonical link and byline.

Related reading