AA Speech to Speech Index: leaderboard
Artificial Analysis Speech to Speech Index (v2): AA's weighted average, for live speech-to-speech voice models, of Big Bench Audio speech reasoning, AA's own τ³-Voice agent runs, blind Speech Agent Arena preference and arena task success. Cascaded agent platforms without index results are not rated here. Index (0-100), higher is better.
Metric: Index (0-100): AA's weighted average of Big Bench Audio, AA-run tau3-Voice, Speech Agent Arena preference and task success. Source: artificialanalysis.ai. Saturation forecast: Rough model projection: around 2026. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phonic 1 (High) | 87.08 | |
| 2 | Grok Voice 3.0 | 86.89 | |
| 3 | GPT-Live-1 (Astra, medium) | 83.15 | |
| 4 | Gemini 3.8 Live Extended Thinking (High) | 82.58 | |
| 5 | Grok Voice 2.0 (High) | 81.34 | |
| 6 | GPT-Live-1 (Sol, low) | 80.08 | |
| 7 | Eleven Realtime v1 Max | 77.32 | |
| 8 | GPT Realtime 2.1 (High) | 75.16 | |
| 9 | Gemini 3.8 Live | 74.52 | |
| 10 | Grok Voice 1.0 | 73.73 |
Interactive version: theaggregate.ai/benchmark?slug=aa-speech-to-speech-index · How It Works · Data refreshed daily, snapshot 2026-11-05.