Qwen-Audio-3.1-Realtime Report - tau-Voice Half-Duplex Adaptation - Telecom: leaderboard
Telecom domain of the report's own half-duplex speech-to-text adaptation of tau-Voice customer-service tasks, not Sierra's or Artificial Analysis's tau3-Voice runs. Score (%), higher is better. Qwen-Audio-3.1-Realtime technical report (Qwen team, arXiv 2609.25176, September 2026): the Qwen team's runs of Qwen-Audio-3.1-Realtime, Qwen-Audio-3.0-Realtime (Plus or Flash not stated), GPT-Realtime-2 (effort not stated) and ByteDance's SeedDuplex 1.2.6.1; the two unspecified comparators carry report-scoped labels.
Metric: Score (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen-Audio-3.1-Realtime | 93.9 | |
| 2 | Qwen-Audio-3.1-Realtime Report Qwen-Audio-3.0-Realtime (Plus or Flash unspecified) | 90.4 | |
| 3 | SeedDuplex 1.2.6.1 | 54.9 | |
| 4 | Qwen-Audio-3.1-Realtime Report GPT-Realtime-2 (effort unspecified) | 32.5 |
Interactive version: theaggregate.ai/benchmark?slug=qwen-audio-3-1-realtime-report-tau-voice-half-duplex-adaptation-telecom · How It Works · Data refreshed daily, snapshot 2026-11-05.