Soniqo Voice Agent Evaluation
Compare three decisions a voice assistant makes: when to speak, which action to take, and how to deliver a reply.
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
Soniqo’s study, bylined Ivan and updated November 1, 2026, compares local and cloud configurations. It uses TurnBench development recordings, spoken commands from SLURP, and English and Chinese InstructTTSEval material.
Methodology
Turn detection measures timing and false alarms. Tool evaluation checks action selection and arguments. The speech test judges how well the voice follows a direction or role. The TurnBench development runs are separate from its official test leaderboard. Speech pilots used a different judge setting; some adapter and judge outcomes remain provisional. Soniqo ran this comparison; the underlying benchmarks have their own creators.