Microsoft

About Microsoft

Explore research at Microsoft, a site featuring the impact of research along with publications, products, downloads, and research careers.

Microsoft Research conducts fundamental science and technology research across a spectrum of research areas. With labs around the globe we pursue breakthroughs across the computing and AI stack to accelerate discovery across sciences and advance the frontiers of computing and AI to create new foundations for others to build upon.

Artificial intelligence | Search & information retrieval | Human language technologies | Human-computer interaction | Computer vision | Graphics & multimedia | Audio & acoustics

Medical, health & genomics | Economics | Social sciences | Technology for emerging markets | Ecology & environment

Website ↗Documentation ↗

Source ↗ · 2026-11-12T02:17:04.301031227Z

669 indexed models

669 matching models

MAI-Image-2e · azure-mai v2026-04-09

azureFoundry/azure-mai/MAI-Image-2e@2026-04-09

Built for creatives, delivering enhanced photorealism at scale.

Context window
131,072
Maximum output
4,096
Input modalities
text
Output modalities
image
Capabilities
imageOutput

MAI-Thinking-1 · azure-mai v2026-06-01

azureFoundry/azure-mai/MAI-Thinking-1@2026-06-01

MAI-Thinking-1 is Microsoft's reasoning model that delivers top-tier reasoning at the best price-to-performance ratio in its weight class.

Context window
256,000
Maximum output
64,000
Input modalities
text
Output modalities
text

MAI-Transcribe-1.5 · azureml-cogsvc v2026-06-02

azureFoundry/azureml-cogsvc/MAI-Transcribe-1.5@2026-06-02

MAI-Transcribe-1.5 is the second iteration of Microsoft's best-in-class speech-to-text model family. It delivers consistently strong transcription accuracy across 43 languages, accents, speaking styles, and noisy environments, with faster inference and now

Input modalities
audio
Output modalities
text
Capabilities
audioInput

MAI-Transcribe-1 · azureml-cogsvc v2026-01-23

azureFoundry/azureml-cogsvc/MAI-Transcribe-1@2026-01-23

MAI-Transcribe-1 is an ASR model built to deliver high quality batch transcription whenever the user speaks. It is designed to achieve high accuracy across 25 languages and to adapt seamlessly to diverse accents, dialects, and regional speech patterns.

Input modalities
audio
Output modalities
text
Capabilities
audioInput

MAI-Transcribe 2

vercelAIGatewayCatalog/microsoft/mai-transcribe-2

MAI-Transcribe 2 is a multilingual speech-to-text model from Microsoft AI, ranked #1 on the FLEURS multilingual benchmark. It supports 60 languages with automatic language identification, code switching for mixed-language speech, speaker diarization, word-level timestamps, keywor

Input modalities
audio
Output modalities
text
Capabilities
audioInput, transcription

MAI-Transcribe-2 · azureml-cogsvc v2026-09-03

azureFoundry/azureml-cogsvc/MAI-Transcribe-2@2026-09-03

The second generation of our speech-to-text model family more accurate, faster, and built to handle a wider range of real-world audio. MAI-Transcribe-2 transcribes reliably across accents, speaking styles, and noisy environments, and now covers 60 language

Input modalities
audio
Output modalities
text
Capabilities
audioInput

MAI-Transcribe-2-Streaming

vercelAIGatewayCatalog/microsoft/mai-transcribe-2-streaming

MAI-Transcribe-2-Streaming is a low-latency, "speech in, text out" realtime transcription API. Audio can be sent to the model as a continuous stream over a WebSocket connection and transcripts are returned incrementally as the speaker talks, with no need to wait for the utterance

Input modalities
audio
Output modalities
text

MAI-Transcribe-2-Streaming · azure-mai v2026-08-06

azureFoundry/azure-mai/MAI-Transcribe-2-Streaming@2026-08-06

MAI-Transcribe-2-Streaming is a low-latency streaming speech recognition model that turns live audio into text as it is spoken. It returns fast partial hypotheses followed by stable final results, enabling responsive speech experiences across 60 languages.

Input modalities
audio
Output modalities
text
Capabilities
audioInput

MAI-Voice-1 · azureml-cogsvc v2025-12-18

azureFoundry/azureml-cogsvc/MAI-Voice-1@2025-12-18

MAI-Voice-1 is a text-to-speech (TTS) model that generates high-quality single-speaker speech and, soon, multi-speaker speech for public preview. It produces audio that strictly follows the input transcript and supports per-turn emotion control as well as

Input modalities
text
Output modalities
audio
Capabilities
speechOutput

mai-voice-2

cheaperInference/mai-voice-2

Output modalities
audio
Capabilities
speechOutput

MAI-Voice-2

vercelAIGatewayCatalog/microsoft/mai-voice-2

Microsoft AI's highest-fidelity text-to-speech model, with expressive multilingual voices and consistent speakers across long-form narration. Served through Azure Speech (preview).

Input modalities
text
Output modalities
audio
Capabilities
audioOutput, speechOutput

mai-voice-2.1

cheaperInference/mai-voice-2.1

Output modalities
audio
Capabilities
speechOutput

MAI-Voice-2.1

vercelAIGatewayCatalog/microsoft/mai-voice-2.1

MAI-Voice-2.1 is our highest-fidelity, most expressive text-to-speech model, delivering rich, natural speech across 23 languages. It extends the MAI-Voice family with broad multilingual coverage, gated instant voice cloning, and strong long-form generation capabilities. With its

Input modalities
text
Output modalities
audio
Capabilities
audioOutput, speechOutput

MAI-Voice-2.1 · azureml-cogsvc v2026-09-19

azureFoundry/azureml-cogsvc/MAI-Voice-2.1@2026-09-19

MAI-Voice-2.1 is a prompted text-to-speech (TTS) model that generates high-fidelity, natural, and expressive speech across 10+ languages. It captures human-like intonation, rhythm, and emotional nuance for engaging conversational experiences.

Input modalities
text
Output modalities
audio
Capabilities
speechOutput

mai-voice-2.1-flash

cheaperInference/mai-voice-2.1-flash

Output modalities
audio
Capabilities
speechOutput

MAI-Voice-2.1-Flash

vercelAIGatewayCatalog/microsoft/mai-voice-2.1-flash

MAI-Voice-2.1-Flash is a text-to-speech model built for fast, low-latency generation. It produces high- fidelity, natural, and expressive speech across 23 languages and supports gated instant voice cloning, all while being optimized for real-time responsiveness. Its human-like in

Input modalities
text
Output modalities
audio
Capabilities
audioOutput, speechOutput

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace