Model Tracker
MAI-Transcribe-2
MAI-Transcribe-2 is an AI model made by Microsoft. Microsoft released it on 3 September 2026. It takes audio and returns text.
In inglese
- Released
- 3 September 2026 — microsoft.ai, read 16 October 2026
- Status
- Not confirmed: two readings disagree: preview and ga
- Takes
- Audio — ai.azure.com, read 16 October 2026
- Returns
- Text — ai.azure.com, read 16 October 2026
- Open weights
- Not confirmed: The maker's model card and catalog do not state whether the model has open weights or provide downloadable weights.
- Licence
- Not confirmed: The maker's model card and catalog do not print a licence name for this model.
- Context window
- Not confirmed: The maker's pages do not print a context window in tokens for this model.
- Longest answer
- Not confirmed: The maker's pages do not print a standard maximum output limit in tokens for this model.
- Knowledge cutoff
- Not confirmed: The maker's pages do not print a knowledge-cutoff date for this model.
- API price, input
- Not confirmed: The maker lists an introductory price per hour of audio, not a standard API input price per 1,000,000 tokens; converting units is not permitted.
- API price, output
- Not confirmed: The maker lists an introductory price per hour of audio, not a standard API output price per 1,000,000 tokens; converting units is not permitted.
- Where
- API — microsoft.ai, read 16 October 2026
- Parameters
- Not confirmed: The maker's pages do not publish a parameter count for this model.
- Model card
- https://microsoft.ai/pdf/MAI-Transcribe-2-Model-Card.pdf — microsoft.ai, read 16 October 2026
Questions
When was MAI-Transcribe-2 released?
Microsoft released MAI-Transcribe-2 on 3 September 2026.
What does MAI-Transcribe-2 take and return?
MAI-Transcribe-2 takes audio and returns text.
More from Microsoft
- Aurora
- Fara1.5-27B
- Fara1.5-4B
- Fara1.5-9B
- MAI-DS-R1
- MAI-Image-2
- MAI-Image-2-Efficient
- MAI-Image-2.5
- MAI-Image-2.5-Flash
- MAI-Image-2.5-Pro
- MAI-Image-2.6
- MAI-Image-2.6-Flash
- MAI-Thinking-1
- MAI-Transcribe-1
- MAI-Voice-1
- Orca 2
- Phi Silica
- Phi-3-mini-4k-instruct
- Phi-3-small-128k-instruct
- Phi-3-small-8k-instruct
- Phi-4-reasoning-plus
- Phi-4-reasoning-vision-15B
Sources
- MAI-Transcribe-2 | Model Catalog | Microsoft Foundry (read 16 October 2026)
- MAI-Transcribe-2 is the fastest, most accurate and cheapest speech recognition model in the world (read 16 October 2026)
- MAI-Transcribe-2 (read 16 October 2026)
- MAI-Transcribe-2 Model Card (read 16 October 2026)