Meta's Muse Voice Transcribe claims the top streaming accuracy spot at 3.1% WER
As seen on the 24/7 Wall St. homepage on September 1, 2026.
Meta Superintelligence Labs' first speech-to-text model tops an independent accuracy benchmark, arriving right as Google pushes Gemini 3.5 Transcribe. Voice is the next front in the AI capex race.
Meta has released Muse Voice Transcribe, taking the #1 spot for Final Transcript accuracy on AA-WER Streaming with 3.1% WER at 0.16s after end of speech Muse Voice Transcribe is the first streaming Speech to Text model developed by Meta Superintelligence Labs. Meta states that https://t.co/unTQQNP5t3
- Replies3
- Reposts10
- Likes47
Continue ReadingShow less
Meta Superintelligence Labs launched Muse Voice Transcribe, its first streaming speech-to-text model, and it immediately claimed first place on the AA-WER Streaming benchmark for Final Transcript accuracy with a 3.1% word error rate.
The AA-WER Streaming benchmark, tracked independently by Artificial Analysis, measures how accurately a model transcribes spoken audio in real time rather than after the full audio clip has finished. A lower word error rate means fewer mistakes, which is what matters most for live applications like voice assistants, meeting transcription, and real-time captioning.
Disclosure
*$149 for two years (or $1.43 per week) is an introductory promotion for new members only. 62% discount based on the current list price of Stock Advisor of $199/year. Membership will renew at the then-current list price at the end of the membership term. Stock Advisor returns are 930% as compared to the S&P 500 returns of 185% as of April 7, 2026.
The same investor newsletter that told subscribers to buy Amazon in 2002, Netflix in 2004, and Nvidia in 2005 still publishes two new stock picks every month. Over 23 years, Motley Fool's Stock Advisor has more than quadrupled the S&P 500. New members get this month's picks, the Top 10 Rankings, and a 30-day money-back guarantee. Click here to unlock their next top stocks while new members are still being accepted.
The commentary notes that Google has pushed Gemini 3.5 Transcribe into the same competitive space, meaning two of the largest AI spenders in the world are now publicly racing on voice accuracy. That raises the stakes for whoever holds the top benchmark position, because enterprise customers evaluating transcription providers often treat independent leaderboard rankings as a shortlist filter.
Meta framing this release under the Superintelligence Labs banner signals that voice is not a side project. For investors tracking where AI infrastructure spending translates into product, a benchmark-leading model shipped this quickly from a newly named research division is a data point worth noting.
Mentioned: META