$META Says Muse Voice Transcribe Leads OpenAI and $GOOGL on Speech Benchmark

Meta said its Muse Voice Transcribe model now leads OpenAI and Google on a key real-time speech benchmark, posting a 3.1% word error rate and supporting more than 70 languages.
The model can also distinguish more than 20 speakers and is available through the Meta Model API, Meta AI for Mac, and Muse Code at $3 per 1,000 audio minutes.