Microsoft AI released MAI-Transcribe-2 on Thursday, pricing the speech-recognition model at $0.10 per hour of audio — a roughly 72% reduction from the $0.36 per hour it charged when the first model in the line shipped five months ago1. The early-bird rate is locked through 20262.
Microsoft says MAI-Transcribe-2 is faster, more accurate, and cheaper than current offerings from OpenAI, Google, and ElevenLabs. The company claims the model is 10 times faster than OpenAI's GPT-Transcribe, seven times faster than ElevenLabs' Scribe v2, and five times faster than Google's Gemini 3.5 Transcribe.
On benchmarks, Microsoft says MAI-Transcribe-2 ranks first on FLEURS across 60 languages with an average word error rate of 5.2%, and second on the Artificial Analysis word-error-rate leaderboard.
The model transcribes audio in 60 languages, up from 43 in June's MAI-Transcribe-1.5 and 25 in the original April release. Enterprise-oriented features include speaker diarization, word-level timestamps, keyword biasing, automatic language identification, and both verbatim and clean output modes. The model also handles code switching, including Hinglish and Spanglish.
MAI-Transcribe-2 is available now through Microsoft Foundry, the company's model marketplace, and in MAI Playground.
The release fits a broader pattern VentureBeat describes: Microsoft building its own frontier-class models modality by modality and swapping them into products that previously ran on OpenAI's technology. Transcription is the modality where that substitution has moved fastest. In July, Bloomberg reported that Microsoft had begun using MAI models to answer a portion of user prompts in Word and Excel.
The competitive dynamics sit against a shifting partnership backdrop. Microsoft invested more than $13 billion in OpenAI. Microsoft hired Mustafa Suleyman from Inflection AI in March 2024, along with most of Inflection's staff. Microsoft and OpenAI restructured their partnership in October 2025, then amended the deal again in April 2026, ending Microsoft's exclusive access to OpenAI's models and eliminating Microsoft's revenue-share payments.
ANALYSIS The pricing trajectory is steep: three releases in roughly five months drove the per-hour cost from $0.36 to $0.10 while expanding language coverage from 25 to 60. For an enterprise processing high volumes of call-center or meeting audio, the cost difference between MAI-Transcribe-2 and competitors' listed rates could shift procurement decisions away from OpenAI and Google endpoints.
The speed claims, if validated by independent benchmarks, compound the pricing pressure. A model that is both cheaper and faster changes the calculus for latency-sensitive applications such as live captioning or real-time agent assist in contact centers.
Microsoft owns Teams, Nuance, and Azure speech services. ANALYSIS Those distribution channels give MAI-Transcribe-2 a built-in deployment surface that standalone transcription vendors cannot match.