AssemblyAI's Universal-3.5 Pro Realtime model has achieved a significant milestone by being the sole entry within Coval's Human Parity Zone on an independent speech-to-text leaderboard. This zone signifies models that match or surpass human performance in both accuracy and speed. The model demonstrated a 3.40% word error rate, placing it lowest among 29 models, and also achieved the fastest time-to-first-token, indicating a natural conversational flow. This dual achievement in accuracy and speed is crucial for voice agents, enabling them to listen accurately and respond promptly. AI
IMPACT Sets a new benchmark for speech-to-text models, potentially driving faster adoption of advanced voice agents.
RANK_REASON Independent benchmark results showing a model achieving human parity on key metrics. [lever_c_demoted from research: ic=1 ai=1.0]
- Apache Software License 2.0
- AssemblyAI
- Coval
- Human Parity Zone
- Lampyridae
- LiveKit
- Universal-3.5 Pro Realtime
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
