Microsoft Releases MAI-Transcribe-2 Speech Recognition Model
Microsoft has shipped three transcription models since April, and MAI-Transcribe-2 claims the top FLEURS score while undercutting OpenAI, Google, and ElevenLabs on both speed and price.
Reporting from 1 source: GIGAZINE.
Microsoft released MAI-Transcribe-2, a speech recognition model that auto-detects 60 languages. The company says it is the fastest, most accurate, and cheapest transcription model available. On the FLEURS benchmark it recorded a 5.2% average word error rate and ranked first. Artificial Analysis verification found it 10 times faster than OpenAI's GPT-Transcribe, 7 times faster than ElevenLabs Scribe v2, and 5 times faster than Google's Gemini 3.5 Transcribe. Pricing is $0.1 per hour through the end of 2026.
MAI-Transcribe-2 follows MAI-Transcribe-1 from April and MAI-Transcribe-1.5 from June, and Microsoft says the new version cuts inference time and latency substantially. The model assigns each word to the correct speaker, tracks precise timestamps for every word, and recognizes domain-specific terms and abbreviations that are hard to resolve from context alone.
Users can choose whether fillers like "um" stay in the transcript or get removed. Language detection runs automatically without prior specification. Artificial Analysis puts the cost at $1.67 per 1,000 minutes, against $3.67 for Scribe v2 and $4 for Smallest AI Pulse Pro. The $0.1-per-hour rate runs through the end of 2026.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.