Skip to main content

MicrosoftAudio modelPreview

MAI-Transcribe-2-Streaming

Popularityno data yet

n/ano search or usage data

Summary

MAI-Transcribe-2-Streaming accepts live audio and returns real-time transcripts. Microsoft describes continuous language detection across 60 languages, so language changes do not need to wait for the end of a recording.

The variant is available as a Microsoft Foundry preview under mai-transcribe-2-streaming. Its announced introductory price is $0.54 per hour of audio through the end of 2026.

GraduallyBench Audio

MAI-Transcribe-2-Streaming

Partially tested
  • The test coverage is incomplete.
  • The route does not confirm the exact model version.
  • A regular connection close was not confirmed.
  • The streaming test has limited coverage.
  • Some cases were answered after a retry.
Show test details

Speech recognition

Gateway · German, English

35 valid outputs from 36 attempted and 36 planned test slots.

Word error rate (WER)
4.55 %(25 / 550 reference words)German
Word error rate (WER)
2.39 %(11 / 461 reference words)English
Word error rate (WER)
3.56 %(36 / 1,011 reference words)German, English
Silence cases with text during the stream
0 %(0 / 4 assessable silence cases)German, EnglishIntermediate text fragments and punctuation count even if they are absent from the final transcript.

WER compares words and CER compares characters with the reference. The speech output was re-transcribed with GPT Transcribe. Its exact backend revision is unconfirmed. Read speech and digital silence do not cover spontaneous speech or real background noise. Error rates can exceed 100 %. Where used, the test references were not independently reviewed by humans. Test coverage and methods differ by model. Each metric stands on its own; there is no combined overall score across tests. Data compiled on 11/10/2026. How we test · Download all Decision and Audio results as JSON

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 10 published benchmark results

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

Omi realtime medical WER, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
1.32
Metric
Medical Word Error Rate; source row gives 38 errors / 2,872 medical words
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Sample
2872
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime overall WER, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
5.73
Metric
Word Error Rate (WER), shared recording set; reference-word denominator not stated
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime drug errors, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
1.81
Metric
Drug errors; source row gives 8 / 442 occurrences
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Sample
442
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime dose/unit matches, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
85.11
Metric
Dose / unit matches; source row gives 80 / 94 reference events
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Higher is better
Sample
94
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime dose/unit errors, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
14.89
Metric
Dose / unit errors; source row gives 14 / 94 reference events
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Sample
94
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime dose precision, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
87.91
Metric
Dose precision
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Higher is better
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime dose F1, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
86.49
Metric
Dose F1
Unit
%
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Higher is better
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Artificial Analysis final transcription WER (Show measurement, test conditions, and source)
Source value
2.5
Metric
word error rate
Unit
%
Category
audio
Direction
Lower is better
Source type
vendor-reported
Evaluator
MAI-Transcribe-2-Streaming Model Card
Status
active
Retrieved at
2026-10-04
Methodology
Vendor-reported result. Final-transcription AA-WER Streaming after detected End-of-Speech; September 28, 2026 leaderboard snapshot in Microsoft's model card, page 2. Published evaluator method: AA-AgentTalk 50%, VoxPopuli-Cleaned-AA 25%, Earnings22-Cleaned-AA 25%; WER duration-weighted within each dataset.
Context
Microsoft reproduces an Artificial Analysis result; the evaluator's own model record was not retrieved directly. Exact serving configuration, sample count for this run, benchmark version in the memo, and run date are unspecified. Snapshot date is not run date; 60 supported languages do not establish test language coverage.
Source MAI-Transcribe-2-Streaming Model Card
Omi realtime median final latency, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
137
Metric
Median final latency, milliseconds
Unit
ms
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Sample
199
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark
Omi realtime p95 final latency, MAI-Transcribe-2 Streaming (Show measurement, test conditions, and source)
Source value
206
Metric
P95 final latency, milliseconds
Unit
ms
Benchmark version
Omi realtime transcription, edition 2026-10-18.1
Category
audio
Direction
Lower is better
Sample
199
Source type
community-reported
Evaluator
Omi Health Medical Speech-to-Text Benchmark
Status
active
Retrieved at
2026-10-19
Methodology
Community-reported result. Exact realtime-table row label Microsoft MAI-Transcribe-2 Streaming. The source describes a shared set of 1,513 English recordings (7.2 hours); every model uses the same audio and scorer. Medical WER covers 2,872 medical words, drug errors 442 occurrences, and dose/unit matching 94 reference events. The source says the MAI streaming row was added in October 2026 but does not state its exact accuracy-run date or immutable serving revision. For speed, the page reports a 19 October 2026 US AWS Ohio client, Azure Sweden Central deployment, shared end-of-speech signal, 200 short clips under 30 seconds, one pass, and timing n=199 after exclusions.
Context
Omi runs and publishes this benchmark while operating its own competing medical transcription product, so retain sourceType community-reported rather than treating it as a neutral independent replication. The source says other providers' accuracy comes from September 2026 and this MAI row was added in October, but it does not give this row's exact accuracy run date. No confidence interval or immutable MAI serving revision is published; the source warns that small differences do not establish statistical significance and scores do not establish clinical safety. Overall WER's reference-word denominator is not stated; 1,513 is the recording count. Precision/F1 denominators are not separately stated.
Source Omi Health Medical Speech-to-Text Benchmark

All specifications

Architecture

Model class
Streaming speech recognition
Source: Model class

Input and output

Input
Live audio
Source: Input
Output
Streaming transcript
Source: Output
Languages
60 languages with continuous language detection
Source: Languages

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
Microsoft Foundry API
Source: Access
Shutdown
None announced
no source

API

API model ID
mai-transcribe-2-streaming
Source: API model ID

Further details

Streaming
Real-time transcription for ongoing audio input
Source: Streaming

Prices

Introductory price
$0.54 per hour of audio through the end of 2026Source

Sources and data date

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

TypeEvidence and data date
Research statusResearch date October 4, 2026. 5 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published

evaluation-protocol

Note

The model card and evaluator methodology establish final-transcription WER, snapshot date, and dataset weighting. Exact serving configuration, run-specific sample count, run date, and a directly retrieved model record remain unavailable.

Additional sourceMicrosoft streaming transcription announcement (retrieved October 3, 2026; October 4, 2026) · Microsoft streaming transcription announcement · Editorial description reviewed October 3, 2026
Additional sourceMicrosoft MAI-Transcribe-2 model page (retrieved October 3, 2026; October 4, 2026) · Microsoft MAI-Transcribe-2 model page · Editorial description reviewed October 3, 2026
Additional sourceMAI-Transcribe-2-Streaming Model Card (retrieved October 4, 2026)
Additional sourceOmi Health Medical Speech-to-Text Benchmark (retrieved October 19, 2026)
Additional sourceModel metadata (retrieved October 4, 2026)