Microsoft's MAI-Transcribe-1 runs 2.5x faster than its predecessor at $0.36 per audio hour

Microsoft announced MAI-Transcribe-1, a speech-to-text model that supports 25 languages.

The model achieved the lowest word error rate on the FLEURS benchmark and outperformed several competitors.

MAI-Transcribe-1 runs 2.5x faster than Microsoft's previous Azure Fast offering and costs $0.36 per audio hour.

Microsoft is rolling the model out in Copilot Voice and Teams and offering a public preview via Foundry and AI Playground.