Voice Agents
The Voice AI People Prefer Still Fails One in Four Tasks
Speech Agent Arena: preference Elo leader vs task success rate
The Voice AI People Prefer Still Fails One in Four Tasks

Why it matters: Artificial Analysis’s new Speech Agent Arena finds Gemini 3.1 Flash Live Preview Minimal leads conversational preference at 1,046 Elo but completes only 74.6% of eligible agentic tasks, 20 points behind Grok Voice Think Fast 2.0 High at 94.7% task success.
View source →- Source
- Artificial Analysis
- Unit
- % task success (agentic)
- Data as of
- August 24, 2026
- Scope
- Artificial Analysis Speech Agent Arena across agentic voice scenarios