| |
Nari Labs' Qwen3-ASR and Qwen3-TTS models rank at the top of Coval's voice AI benchmark, achieving #1 for Speech-to-Text latency (44ms) and Text-to-Speech accuracy (3.8% WER), while offering significantly lower costs than competing models like ElevenLabs and Deepgram. The models occupy the quality-latency Pareto frontier, balancing performance with speed critical for responsive voice AI agents, with pricing starting at $0.06/hour for STT and $5 per million characters for TTS.
Read Full Article →
← More Tech news