Developers building voice agents that must react mid-sentence get fast transcription at an introductory $0.54 an hour until the end of 2026.
Kenning's take
What it does
MAI-Transcribe-2-Streaming is Microsoft AI's first streaming transcription model. It returns a first guess at the words about 100 milliseconds after hearing audio, then corrects it as more speech arrives, so voice agents can start acting before a speaker finishes. It supports 60 languages with automatic language detection and is available to developers in Microsoft Foundry, Azure Voice Live and the MAI Playground.
- Partial transcripts in just over 100ms
- 60 languages with continuous language detection
- Ranked first for accuracy on Artificial Analysis (Microsoft claim)
- Pairs with MAI-Voice-2.1-Flash for voice agents
Who it suits. Developers building voice agents, live captioning or dictation.
Key facts
- Category
- AI models
- Made by
- Microsoft AI
- Based in
- United States
- Released
- 1 October 2026
- How to get it
- through Microsoft Foundry, Azure Voice Live and the MAI Playground, with LiveKit support listed as coming soon
- Works on
- API
- API price
- $0.54 per hour of audio (introductory price through end of 2026)
FAQ
When was MAI-Transcribe-2-Streaming released?
Microsoft launched it on 1 October 2026.
How much does MAI-Transcribe-2-Streaming cost?
$0.54 per hour of audio at an introductory price through the end of 2026.
How many languages does it support?
60, with automatic language detection.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.