Fact · Voice and AI
26 to 38percent task completion
On the τ-Voice benchmark, voice agents completed 26 to 38 percent of tasks under realistic conditions, with noise, diverse accents and interruptions.
Checked against its source on October 17, 2026
In the source's own words
“and 26--38% under realistic conditions with noise and diverse accents”
τ-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains, Soham Ray et al., arXiv, 2026, Abstract.
Open the original- What it measures
- Realistic conditions add noise, accents and turn-taking, and the simulated caller can interrupt the agent.
- What it does not show
- Simulated callers, not real customers. The result is for this benchmark and its tasks.