Microsoft AI has released MAI-Transcribe-2, a speech-to-text model it calls the fastest, most accurate, and cheapest in the world — and it put a price on that claim: $0.10 (around ₱6) per hour of audio, announced on the Microsoft AI newsroom on September 3. That is a 72% cut from the $0.36 (around ₱22) launch price of its predecessor, and it is the kind of number that matters in the Philippines, where transcription remains paid work inside the business process outsourcing (BPO) industry.
The benchmark claims
Microsoft says MAI-Transcribe-2 ranks first on the FLEURS multilingual benchmark across 60 languages, with an average word error rate of 5.2% — meaning about one word in twenty is transcribed wrong, on average, across those languages. It ranks second on the independent Artificial Analysis word-error-rate leaderboard. The company claims it outperforms Gemini 3.5 Transcribe, GPT-Transcribe, Whisper V3-Large, and ElevenLabs' Scribe v2 on accuracy, while also being much faster:
| Compared with | Microsoft's speed claim |
|---|
| OpenAI GPT-Transcribe | up to 10x faster |
| ElevenLabs Scribe v2 | up to 7x faster |
| Google Gemini 3.5 Transcribe | up to 5x faster |
The model supports speaker diarization (labeling who said what), word-level timestamps, keyword biasing, automatic language identification, and code-switching — handling conversations that mix languages mid-sentence, the way Taglish speakers do. It is available through Microsoft Foundry and the MAI Playground.
The $0.10-per-hour rate is a promotional price that runs through December 31, 2026; Microsoft has not said what the regular rate will be afterward. VentureBeat's reporting puts the economics in context: transcribing 100,000 hours of audio a year now costs about $10,000 (around ₱620,000), down from $36,000 (around ₱2.2 million) on the previous model. Microsoft AI chief Mustafa Suleyman credited a focused 10-person team for the model, saying it was trained at half the GPU cost of other state-of-the-art systems.
Why the Philippines should pay attention
Transcription is exactly the kind of routine, language-based work the Philippine outsourcing industry has long sold to the world — and the kind its own leaders say is most exposed to artificial intelligence. In July, the IT and Business Process Association of the Philippines (IBPAP) cut its 2028 workforce forecast to between 1.85 million and 2.14 million workers, down from an earlier target of 2.5 million, GMA News reported. Revenue targets fell too, from $59 billion (around ₱3.7 trillion) to between $43.3 billion and $50.5 billion (around ₱2.7-3.1 trillion). IBPAP president Jack Madrid named rapid AI advances among the main reasons, alongside slower client decisions and competition from other outsourcing countries.
When a global platform reprices transcription to roughly ₱6 per audio hour — less than a jeepney fare — the market for human-powered transcription services shrinks with it. The pressure lands on entry-level, volume-based work first: medical and legal transcription, call recordings, meeting notes. The counterpoint is that the same API becomes an input Filipino firms can build on, and the industry's own plan is to shift workers from that routine layer toward higher-value tasks — the "capacity to capability" shift Madrid has described. Autonomous AI agents are already moving up that value chain, as OpenAI's GPT-6 Astra launch showed in August.
The bigger Microsoft story
MAI-Transcribe-2 is Microsoft AI's third model release in five months, part of a deliberate push to build its own models and lean less on OpenAI after the two restructured their partnership in late 2025. Pricing a frontier-quality model at a loss-leader rate is a market-share play — and speech is a logical place to make it, because transcription is a high-volume commodity workload where price, not brand, wins the contract.
