Aqua Voice's Avalon is #6 on OpenASR Leaderboard (aquavoice.com)

Aqua Voice’s Avalon has climbed the OpenASR leaderboard to become the #1 commercial speech-recognition model worldwide and #6 overall across open and proprietary systems, posting an average WER of 6.24 and beating commercial incumbents such as OpenAI Whisper-large-v3, ElevenLabs Scribe v1, and Rev AI Fusion, as well as strong open models like NVIDIA Canary 1B and CrisperWhisper. The OpenASR suite—seven public datasets evaluating word-error-rate against human labels—provides a standardized industry yardstick, so Avalon's placement signals broad, real-world accuracy rather than narrow benchmark tuning. Technically, Avalon was engineered for human–computer interaction and to reduce mistranscriptions of programming terms, company names, and other jargon. On Aqua Voice’s AISpeak benchmark it reached 97% accuracy on those domain-specific terms (vs. 65% for Whisper Large v3 and 78% for ElevenLabs Scribe v1), and those targeted improvements translated into lower aggregate WER on standard datasets. Avalon is available in the Aqua Voice app for English and via a free API through October 15th, making it immediately accessible for developers and teams looking to integrate a commercially competitive ASR tuned for conversational and technical vocabularies.