AI modelsOctober 03, 2026
34
š¶ļøš¶ļø
Microsoft's Streaming STT Crushes the Field
Microsoft's MAI-Transcribe-2-Streaming is regarded as the top real-time speech-to-text model.
#Microsoft#Speech-to-Text#Streaming#Real-Time#AI Models
š„ What happened
Microsoft dropped MAI-Transcribe-2-Streaming: 2.5% WER at 0.13s after speech ends, ranking #1 of 38 models. First partials land in just over 100ms ā faster than anything else shipping.
š” Why it matters
Agents can start reasoning mid-sentence. Grok needs 0.49s, Muse 0.16s ā Microsoft matches their accuracy at 4x the speed. Price: $0.54/hour, 5x the batch rate.
ā” Our take
Blazing fast, painfully priced. Voice-agent builders are now paying Microsoft's margin ā until open weights catch up.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions ā when in doubt, read the linked original source.
