ElevenLabs Scribe v3 ranks #2 on AA-WER speech benchmark
Third-party benchmarks like AA-WER are becoming the de facto shortlist for teams selecting ASR APIs, and this one shows accuracy gaps among leading models have narrowed to fractions of a percentage point. Developers building transcription features should weigh Scribe v3 against Alibaba's Fun-Realtime-ASR-preview and Microsoft's MAI-Transcribe-2 on latency, pricing and language support rather than WER alone. Enterprise buyers in multilingual ASEAN markets will want to check per-language accuracy before committing.