MAI-Transcribe-2 is the fastest, most accurate and cheapest speech recognition model in the world | Microsoft AI
The multilingual model supports 60 languages and launches with a promotional price of $0.10 per hour through the end of the year.
- Includes speaker diarization, word-level timestamps, keyword biasing, and automatic language identification.
- Supports code-switching for blended conversational languages such as Hinglish and Spanglish.
- Provides configurable output modes, including verbatim capture with filler words and a clean mode for readability.
- Averages a 5.2% word-error-rate across 60 languages on the FLEURS benchmark.
- Runs up to 10x faster than OpenAI's GPT-Transcribe and 5x faster than Gemini 3.5 Transcribe in Artificial Analysis benchmarks.
Developers transcribing multilingual and noisy audio gain lower inference latency and built-in diarization at reduced operating costs.
