Microsoft's MAI-Transcribe-2: Launch and Rapid Adoption

Episode 1 · Microsoft launches MAI-Transcribe-2: 411x realtime, lowest market price (2026-09-03, 13 posts)

Microsoft AI launched the speech-to-text model MAI-Transcribe-2 on Sept 3, announced by Microsoft AI chief Mustafa Suleyman, simultaneously appearing on Artificial Analysis' non-streaming ASR leaderboard. Available via Microsoft Foundry and OpenRouter, it enters the top tier on accuracy, speed and price, with the lowest pricing in the market, and is widely seen as a direct challenge to OpenAI, Google and ElevenLabs.

Confirmed

• Accuracy: AA-WER of 2.0%, second on the Artificial Analysis leaderboard behind Alibaba's Fun-Realtime-ASR-preview; per altryne and emmanuelvivier, it ranks first across 60 languages on FLEURS with an average WER of 5.2%.
• Speed: 411x realtime; Suleyman says it is 10x faster than GPT-Transcribe and 5x faster than Gemini 3.5. Altryne's hands-on test transcribed 90 minutes of audio in about 15 seconds.
• Pricing: $0.10/hour via Microsoft Foundry ($1.67 per 1,000 minutes), the lowest in the market, below competitors like ElevenLabs; per VentureBeat (via MichaelFNunez), the price holds through 2026 and Microsoft claims it outperforms Gemini 3.5 Transcribe and GPT-Transcribe.
• Availability: Beyond Microsoft Foundry, MAI-Transcribe-2 is on OpenRouter supporting 60 languages, automatic language detection, mixed Chinese-English, speaker diarization and word-level timestamps; altryne reports accuracy clearly better than most prior models.
• Background: Per luisdans, the model comes from Microsoft's internal MAI team and already powers products like BizChat; Microsoft executives amplified VentureBeat coverage highlighting its cost-performance balance, and the team is soliciting user feedback.

Why it matters

• The speech transcription market is heating up: Microsoft's combination of near-top accuracy, extreme inference speed and dominant pricing pressures OpenAI, Google and ElevenLabs, plus rivals like Alibaba's Fun-Realtime-ASR.
• At $1.67 per 1,000 minutes, it dramatically lowers the barrier for large-scale transcription (podcasts, meetings, subtitles) and may accelerate a price war.
• OpenRouter availability plus 60-language support, diarization and word-level timestamps lower developer friction and help it spread quickly across the ecosystem.

• Microsoft's MAI-Transcribe-2 hits 2.0% WER at 411x real time for $1.67 per 1,000 minutes — ArtificialAnlys · 2026-09-03
• Microsoft's MAI-Transcribe-2 hits ASR leaderboard at $0.10/hour, far below rivals — ArtificialAnlys · 2026-09-03
• Microsoft's MAI-Transcribe-2 hits 2.0% WER at 411x real-time, priced at $1.67 per 1,000 minutes — ArtificialAnlys · 2026-09-03
• Microsoft AI launches MAI-Transcribe-2: 10x faster than GPT-Transcribe at market-lowest pricing — mustafasuleyman · 2026-09-03
• Microsoft launches MAI-Transcribe-2, 10x faster than GPT-Transcribe at lowest market price — mustafasuleyman · 2026-09-03
• Microsoft's MAI-Transcribe-2 hits 2.0% WER (#2) at ~411x real time, $1.67 per 1,000 minutes — altryne · 2026-09-03
• Microsoft unveils MAI-Transcribe-2, claims it beats Gemini 3.5 and GPT at $0.10/audio hour — MichaelFNunez · 2026-09-04
• Microsoft's MAI-Transcribe-2 undercuts OpenAI, Google and ElevenLabs on price and speed — MichaelFNunez · 2026-09-04
• Microsoft's MAI-Transcribe 2 tops FLEURS, transcribes 90 minutes of audio in 15 seconds — altryne · 2026-09-04
• Microsoft launches MAI-Transcribe-2, claiming 10x speed of GPT-Transcribe — ZacharyHuang12 · 2026-09-04
• Microsoft's MAI-Transcribe-2 draws attention for cost-performance balance — luisdans · 2026-09-05
• Microsoft launches MAI-Transcribe-2, a faster, cheaper and more accurate speech model — emmanuelvivier · 2026-09-05
• Microsoft Launches MAI-Transcribe-2, Tops FLEURS Across 60 Languages With 5.2% WER — emmanuelvivier · 2026-09-05

Episode 2 · Microsoft's MAI-Transcribe-2 Hits 1M Requests in 5 Days on OpenRouter (2026-09-11, 2 posts)

Microsoft AI's MAI-Transcribe-2, a multilingual speech-to-text model with built-in speaker diarization and timestamps, processed 1 million requests within 5 days of launching on OpenRouter.

• Microsoft's MAI-Transcribe-2 hits 1M OpenRouter requests in 5 days at $0.10/hour — mustafasuleyman · 2026-09-11
• Microsoft launches MAI-Transcribe-2: single multilingual transcription model with diarisation and timestamps — mustafasuleyman · 2026-09-12