KKnowFlowHub
NewsAuthorKnowFlowHub

Microsoft Unveils First In-House AI Models MAI-Voice-1 and MAI-1-preview (November 1, 2026)

On November 1 Microsoft unveiled its first in-house models: MAI-Voice-1 makes a minute of audio in under a second on one GPU; MAI-1-preview trained on ~15,000 H100s.

微软MAI自研模型语音生成

One-sentence summary

On November 1 Microsoft unveiled its first in-house AI models: MAI-Voice-1 generates a minute of audio in under a second on a single GPU, and MAI-1-preview, trained on about 15,000 Nvidia H100s for everyday conversation, is in Copilot and open for testing on Copilot Labs and LMArena.

Event Overview ​

On November 1, 2026, Microsoft unveiled its first in-house AI models, MAI-Voice-1 and MAI-1-preview, signaling a push toward consumer-facing AI independence from OpenAI.

Key Information ​

  • MAI-Voice-1 is an ultra-fast speech generator that delivers a minute of audio in under a second on a single GPU.
  • MAI-1-preview is a language model for everyday conversational tasks, trained on about 15,000 Nvidia H100 GPUs.
  • Both are integrated into Copilot features and available for public testing on Copilot Labs and LMArena.

Official Sources ​

AuthorKnowFlowHub
Published
KEEP READING

Keep reading

Article

Microsoft's MAI Strategy: From Relying on OpenAI to In-House Voice and Conversational Models

Compare

Microsoft MAI vs OpenAI GPT vs Google Gemini: In-House Model Strategies Compared

Dataset

Microsoft MAI-Voice-1 and MAI-1-preview Model Data

COPYRIGHT

Copyright & Infringement Notice

Content on this site is compiled from public sources for reference and learning. If you believe any content on this site infringes your rights (including but not limited to text, data, images and tables), please contact us and we will handle it promptly after verification:

  • Email us at: liu@chenra.com
  • In the email, please specify: the exact page URL of the infringing content, your proof of rights (e.g. copyright certificate, authorization document), and your contact information
  • We will verify and process within a reasonable period after receiving a valid notice, including removal or source attribution

All content on this site is attributed to its data sources. If any source is inaccurate or you believe a citation exceeds a reasonable scope, please let us know.