EchoChain - Criterion Pass Rate: leaderboard
Full-duplex voice assistants on EchoChain (2026): in 200 interrupted conversations the user barges in mid-response with information that changes the task, and human annotators grade the continuation blind against an instance-specific rubric. Share of individual rubric criteria satisfied, in percent; higher is better.
Metric: Mean criterion pass rate (%): share of individual rubric criteria satisfied, over 200 human-screened interrupted conversations in which the user barges in mid-response with new task information (pre-generated cloned-voice audio, fixed barge-in timing), each model response graded blind by human annotators against an instance-specific rubric; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok Voice Agent | 85.5 | |
| 2 | GPT Realtime 2025-08-28 | 81.1 | |
| 3 | Nova 2 Sonic | 78.1 | |
| 4 | Gemini Live 2.5 Flash Native Audio | 64.8 |
Interactive version: theaggregate.ai/benchmark?slug=echochain-criterion-pass-rate · How It Works · Data refreshed daily, snapshot 2026-11-22.