Microsoft introduced MAI-Transcribe-2, a speech-recognition model offered at $0.10 per audio hour through year-end.
Microsoft launched MAI-Transcribe-2 on Sept. 3 as a public-preview speech-to-text model in Azure Speech and Microsoft Foundry. It supports 60 languages, speaker diarization, word-level timestamps, automatic language detection and code-switching. Microsoft says it records a 5.2% average word-error rate on FLEURS and processes audio up to 10 times faster than OpenAI’s GPT-Transcribe. At $0.10 per audio hour through Dec. 31, 2026—about 72% below its predecessor’s $0.36 rate—the model could sharply reduce transcription costs, though Microsoft has not announced post-promotion pricing.