Skip to main content

Microsoft AIAudio modelTestable in MAI Labs, restricted Foundry access

MAI Realtime 1

Summary

MAI Realtime 1 is Microsoft's first speech-to-speech model. It processes spoken input and responds with speech. Microsoft offers a demo in MAI Labs and a limited-access Foundry preview.

All specifications

Architecture

Model class
Real-time speech model for direct speech input and speech output.
Source: Model class

Input and output

Input
Real-time audio.
Source: Input
Output
Real-time audio.
Source: Output
Capabilities
Handles interruptions, brief acknowledgements, and background noise; can connect to tools and enterprise systems.
Source: Capabilities

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking
Unknown
no source
Watermarking technique
Not known
no source

Licence and access

Access
Testable through MAI Labs. Microsoft Foundry offers a limited-access preview requiring registration. Microsoft says preview customers will not be charged before public preview begins.
Source: Access
Shutdown
None announced
no source

Further details

API compatibility
Compatible with the OpenAI Realtime API. A separate public API model ID is not specified in the reviewed sources.
Source: API compatibility
Weights and license
Proprietary hosted service. Public downloadable weights and a model license are not specified.
Source: Weights and license

Prices

Announced starting price
Microsoft lists $0.01 per audio minute for speech-to-speech as the starting price. Microsoft says customers in the limited Foundry preview will not be charged before public preview begins.Source
Announced reasoning price
Microsoft lists separate reasoning rates per 1 million tokens: $1 for input and $5 for output. Customers in the limited Foundry preview will not be charged before public preview begins.Source