What is Benten Unlike standard LLM evaluation tools that only analyze text transcripts, Benten evaluates the complete acoustic and conversational performance of human-to-agent voice calls. When a customer speaks with a Voice AI agent, Benten measures key acoustic and speech metrics: Turn Latency: Response delay (in ms) from user speech termination to agent voice playback. Speech Quality (NISQA MOS): Deep learning Non-Intrusive Speech Quality Assessment scoring (1.0 to 5.0 MOS). Dead Air Percentage: Unintended silent pauses during the call session. User Interruptions: Occurrences where the customer interrupts or speaks over the agent. Health Score: Calculated conversation quality score (0 to 100) combining acoustic telemetry, sentiment, and flow.

Benten

Voice AI Agents Audio Evaluation & Telemetry Platform

Voice AI Agents Audio Evaluation & Telemetry Platform