Microsoft claims cheapest, fastest speech recognition model
Microsoft released MAI-Transcribe-2, which it claims is the fastest, most accurate, and cheapest speech recognition model available, undercutting Gemini 3.5 Transcribe, GPT-Transcribe, Whisper V3-Large, and ElevenLabs Scribe. New features include diarization, configurable transcription styles, and word-level timestamps, at $0.10 per hour of audio.
Why it matters
๐ป Developer ยท Word-level timestamps and diarization out of the box make this worth benchmarking against Whisper for transcription pipelines.
๐ฆ Product ยท Transcription costs keep falling, which changes the math for any product built around call or meeting transcripts.
๐จ Design ยท Configurable transcription styles could simplify building readable transcripts without extra post-processing.
๐ Business ยท At 10 cents an hour, transcription is now cheap enough to add to almost any customer-facing workflow.
๐ค Just Curious ยท Microsoft says its new speech-to-text model is both cheaper and more accurate than Google's and OpenAI's.
Sources: MAI-Transcribe-2