Microsoft MAI
Microsoft AI profile, latest version intelligence, capability movement, and source-aware release history.
Profile
Microsoft MAI is Microsoft AI’s first-party model ecosystem, including reasoning, coding, image, voice, and transcription models.
Latest Version
MAI-Transcribe-2
Sep 2026
Confidence
94%
extracted
Context
N/A
last known / not captured
Versions
7
documented releases
Current Release Intelligence
Latest published version and metadata capture state
Release
MAI-Transcribe-2
minorMicrosoft released MAI-Transcribe-2, a new speech recognition model that supersedes its predecessor. The model is claimed to be the fastest, most accurate, and cheapest speech recognition model in the world.
Decision Fit
Best For
Repository coding and rapid iteration
Avoid If
You need non-technical, no-setup workflows
Release Diff
Core fields and capture notes for this release
Version
Context Window
Not captured in this release; showing last known value when available
Not captured
Pricing
Not captured in this release; showing last known value when available
{"MAI-Image-2.5-Pro":{"text_input":"$5 per 1M text input tokens","image_input":"$8 per 1M image input tokens","image_output":"$106 per 1M image output tokens"},"MAI-Voice-2-Flash":"$15 per 1M characters"}
Current Capabilities
MAI-Transcribe-2 with previous-release movement
Previous score: 9
Version History
7 releases documented
MAI-Transcribe-2
September 5, 2026
Microsoft released MAI-Transcribe-2, a new speech recognition model that supersedes its predecessor. The model is claimed to be the fastest, most accurate, and cheapest speech recognition model in the world.
MAI-Image-2.6
August 11, 2026
Microsoft launched MAI-Image-2.6, a new image AI model that debuted at No. 2 on the Arena leaderboard, surpassing models from Google, Meta, and xAI. This represents a significant milestone for Microsoft's MAI image generation capabilities, positioning it as a top competitor in the image AI space.
MAI-Cyber-1-Flash
July 28, 2026
Microsoft introduced MAI-Cyber-1-Flash, a new cybersecurity-focused AI model integrated inside MDASH (Microsoft Defender Advanced Security Hub). This represents a specialized 'Flash' variant of the MAI-Cyber-1 model, optimized for speed and cybersecurity applications. The release marks a significant update to Microsoft's MAI family with a dedicated cybersecurity intelligence capability.
MAI-Image-2.5-Pro and MAI-Voice-2-Flash
July 24, 2026
Microsoft released two new model variants: MAI-Image-2.5-Pro, a high-fidelity image generation and editing model with precise in-image text rendering now in public preview, and MAI-Voice-2-Flash, a speed-optimized voice model that is 2x faster and 32% cheaper than MAI-Voice-2 also now in public preview. MAI-Image-2.5 is now the sole model powering Bing Image Creator end-to-end, and MAI-Voice-2-Flash now powers Dynamics 365 Contact Center for enterprise call center agents.
Hill-Climbing MAI models
July 24, 2026
Microsoft announced new Hill-Climbing MAI models being deployed for GitHub Copilot and Excel. This represents a significant update to the MAI model lineup with specialized versions optimized for coding assistance and spreadsheet productivity use cases.
MAI-Voice-2
June 3, 2026
Microsoft introduced MAI-Voice-2, a new expressive speech model supporting 10 languages. This represents a significant update to Microsoft's MAI voice capabilities with improved expressiveness across multiple languages.
MAI-Transcribe-1.5
June 3, 2026
Microsoft announced MAI-Transcribe-1.5, a new version of their transcription AI model. The update focuses on improved accuracy, context-awareness, and production-readiness compared to previous versions.